# Metis by Orama > Metis is the operating system an enterprise runs its intelligence on: the applications that create data, the pipelines and warehouses that make it available, and the models, agents, and dashboards that turn it into decisions, outcomes, and returns. Built by Orama on Zetta; runs in your cloud, on-prem, or dedicated. --- # Metis · The Enterprise Intelligence OS URL: https://orama.com/ > Metis is the operating system an enterprise runs its intelligence on: the applications that create data, the pipelines and warehouses that make it available, and the models, agents, and dashboards that turn it into decisions, outcomes, and returns. Built by Orama on Zetta; runs in your cloud, on-prem, or dedicated. # The Enterprise Intelligence OS Investable, fiscally responsible AI and intelligence needs the right data. The Metis enterprise intelligence operating system: applications that create the right data, the pipelines and warehouses that make it available, and the models, agents, and dashboards that turn it into decisions, outcomes, and returns. Built by Orama on Zetta. Talk to sales See the platform OramaJS on GitHub · 10.6k stars · 5.8M monthly downloads Diagram: Four previews of Orama Studio, the surface of Metis, turning business questions into models, forecasts, explanations, and next actions Orama Studio, the surface of Metis, turning business questions into explained models, forecasts, and next actions. What is intelligence? ## Intelligence is high-quality decisions, made fast, made many For an organization, intelligence is not a model or a tool it has licensed. It is the capacity to know what is happening, why, and what will happen next, to decide what to do, and then to do it. The Greeks called the practical kind mētis: wisdom applied to the situation in front of you. Metis is named for it. ### High-quality decisions A sound process: the right frame, real alternatives, reliable information, and clear trade-offs, so the choice it produces is a sound one. Limited by whether the data is true, complete, and understood. ### Fast decisions Made at the speed the world changes. A weekly report cannot run an intra-day pricing decision. Limited by whether the data is current when the decision is made. ### Many decisions Made at scale and repeatably, up to and including acting without a person in the loop. Limited by whether the data can be reached, is permitted for the use, and is in a form software can use. A company that owns a forecasting platform but does not trust its forecasts has no predictive capability. Intelligence is what an organization can actually act on. The big idea ## AI is only as good as the data it can reach No model, agent, or platform can see past the data underneath it. That makes data availability the place to start, and how far your data can carry each decision can be measured before anything is built. ### The argument you have been sitting in - Which model is smarter - Which agent framework is newer - Which platform has more features - Whose AI strategy is thicker Every one of them argues about what to build on top of the data. Every one of them assumes the data is there. ### The question that has a number What can your data actually carry, for the decisions you have to make? Scored per decision on ten dimensions, combined as a weakest link, and summarized as a unit's Corporate Intelligence Capability (CIC) score from 0 to 10. Nothing above the data raises it, so it is measured first. Built around the data you already own, and the data you do not have yet - Application data - Warehouses and lakes - Documents and records - Decisions and outcomes - Sensors, partners, and the web Grounded in reality ## Data availability first, then decisions, outcomes, and returns Every report, model, and agent is downstream of the data it can reach: you cannot have AI, or intelligence, without data availability, and nothing built above the data raises that ceiling. Metis measures availability first, before anything is built, and scores it per decision, because relevance is a relation between data and a decision: the decision says which data has to be available. That is decision-driven, not data-driven, which starts from what exists and answers only the questions the data happens to be able to answer. Metis raises availability where it is short and logs every decision so outcomes and returns are measured, not assumed. Intelligence and AI investments, grounded in return on investment (ROI). Data availability Decisions Outcomes and returns Outcomes and returns: every decision is logged with what was expected, outcomes are measured at the horizon, and returns are stated as what was invested, how far availability rose, and what the business metric did: return on investment (ROI), in the unit's own metric. Diagram: How Metis grounds intelligence, in order: your data is scored for availability against a decision, an instrument runs only when the data can carry it, the decision is logged, the outcome is measured at the horizon, and the returns are stated in the unit's own metric Nothing above the data can add information the data does not contain. That is why Metis starts with availability, not with a model. What the OS spans ## Application development, data engineering, and AI, as one system ### Applications that create data The one thing nothing downstream can repair is data that was never captured. Metis builds the applications that create it: capture at the point of decision, workflow applications in place of spreadsheets, extractors, experiments, and collectors. ### Pipelines and warehouses that make it available Zetta compiles intent into verified pipelines and holds the data on one distributed platform, whether it is the store, runs over your warehouse, or becomes the store over time. ### Models, agents, and dashboards that use it Models trained on your data with calibrated confidence and explained drivers, grounded answers, agents under per-action permission, and dashboards over one definition per metric. AI is the engine of these, and a small part of the whole. ### Landed by forward-deployed engineers Metis is landed inside your organization: inventory the decisions, score the data, raise what is short, run what the data can carry, and measure the outcome on a metric you already report. In plain numbers ## Built in the open, sold in plain words 10.6k GitHub stars on OramaJS, our open-source search library 5.8M Monthly downloads of OramaJS on npm 30 days to 11 Sep 2026 One belief The right data driving measurable outcomes Open source Read OramaJS before you commit to anything Your cloud, on-prem, or dedicated Metis runs in your cloud, on-prem, or dedicated What Metis is made of ## One operating system, three ways in Understand the platform, bring enterprise decisions to Orama Studio, or build directly with the open-source OramaJS library. The Enterprise Intelligence OS ### Metis The operating system itself: what Zetta makes possible, how the OS is grounded in reality, and how it lands. See the platform The surface of Metis ### Orama Studio Turn business questions into verified production models, pipelines, and agents on your infrastructure. Explore Studio Open-source search ### OramaJS A free search library for applications that want to own and operate their index. Explore OramaJS How it lands ## Forward-deployed, on infrastructure you choose Metis runs in your cloud, on-prem, or dedicated, and it meets your data where it is. All three are supported, chosen with your platform team. Talk to sales Explore Studio - Zetta as the store, your warehouses as sources - Zetta over the warehouse you already have - Zetta becoming the store over time Proof you can inspect ## Start with the platform See what Zetta makes possible, dashboards, pipelines, warehouses, applications, agents, models and model training, and how the OS is grounded in data availability, decisions, outcomes, and returns. Product truth is more useful than an unsupported customer quote. See the platform ## Start with your decisions Bring the decisions that matter and the data you have. The first thing we do is score them; the second is show you what moves the score. That is how AI spending stays fiscally responsible. Talk to sales Why it works --- # OramaJS · Open-source search library URL: https://orama.com/products/oramajs > Full-text, vector, and hybrid search in an open-source JavaScript library. Runs in the browser, on your server, or at the edge. Install with npm and start searching. OramaJS · open source # OramaJS: the open-source search library Full-text, vector, and hybrid search in one library that runs in the browser, on your server, or at the edge. Free and open source, from the team behind Orama. The library has its own capability set; the architecture page lists what the managed Orama service ships. bash ``` npm i @orama/orama ``` View on GitHub Read the docs Open source 10.6k GitHub stars OramaJS docs open source - Quickstart - Schema - Documents - Search - API reference Guides Search the docs Getting started ## Quickstart Install OramaJS, define the fields you want to search, add documents, and run a full-text query. ### Install OramaJS Add the package with npm and keep the index inside your application. ### Define a schema Describe the document fields and types that belong in the index. ### Insert documents Add your own records without sending them to a managed service. ### Run full-text search Query the index with ranking, filters, and the result fields your UI needs. On this page - Install - Define a schema - Insert documents - Search OramaJS · managed path ## When you outgrow running it yourself OramaJS is free and open source, and it stays that way. When search is one part of a larger question, Metis, the Enterprise Intelligence OS from the same team, runs it on your data, in your environment, with everything above it. Talk to sales ## Start with the docs Follow the quickstart, inspect the open-source implementation, and choose the path that fits your application. Read the docs Browse the source --- # Orama Studio · The surface of Metis URL: https://orama.com/products/studio > Orama Studio is the surface of Metis, the Enterprise Intelligence OS: turn a plain-language business goal into a verified production pipeline with models trained on your data, deployed to infrastructure you control. Orama Studio · the surface of Metis # Turn a business question into production AI Orama Studio is the surface of Metis, the Enterprise Intelligence OS. Describe the outcome in plain language and connect the data you already have. Studio scores what that data can carry, builds the pipeline, trains models on that data, verifies the system, and deploys it to infrastructure you control. Talk to sales See how it works Diagram: Orama Studio connects your data to a verified pipeline, trains models, and serves running results on infrastructure you control.. Your data feeds a verified pipeline that trains models and serves running results on infrastructure you control. Your data feeds a verified pipeline that trains models and serves running results on infrastructure you control. Why a platform ## A useful model still needs a production system A model that works in a notebook still needs source connections, point-in-time features, leakage checks, serving, monitoring, and governance. Orama Studio builds those parts as one verified pipeline, without requiring the customer to assemble a separate data engineering or data science function. ## What your teams get ### From question to running pipeline Start with a plain-language goal. Orama Studio connects the sources, builds the features, trains the model, and creates the serving path. ### Measured against real outcomes Models train on your data and are evaluated against what happened. Each score includes the drivers behind it, so teams can inspect the result. ### Governed from the center Teams build and operate verified pipelines while architecture, access policy, deployment, and audit remain centrally owned. ## Go deeper Review the workflow, product architecture, deployment model, and operating model. How it works From a stated problem to a running, validated system in four stages. Read more → Architecture One engine as control plane and data substrate, in a distributed, self-healing cluster. Read more → Deployment Your cloud, on-prem, or dedicated. One true-hybrid cluster across vendors and clouds. Read more → For teams What changes for platform leaders, data scientists, and the teams waiting on answers. Read more → ## Your platform, under your brand Deploy in your own cloud, on-premises, or on dedicated infrastructure. Run Orama Studio under your own brand. A forward-deployed engineer helps your team select use cases, prove return, and build a center of excellence that can ship more projects without permanent operators. Talk to sales ## Start with the outcome Tell us what you want to predict or automate and where the system has to run. We will scope the first production use case with you. Talk to sales --- # Orama Studio · How it works URL: https://orama.com/products/studio/how-it-works > From a plain-language business goal to a running data and AI system in four stages. Every pipeline passes verification before production. Orama Studio · how it works # From a plain-language question to a running system Orama Studio turns a business goal, or an artifact you already have, into a deployed data and AI system. Each stage is visible, and every pipeline passes verification before production. The arc ## Connect, model, deploy, and learn as one flow. Diagram: The Orama Studio workflow: intent and existing sources are compiled into a validated graph, deployed in your cloud, on-prem, or dedicated, then executed to return answers with lineage that the next run learns from.. Intent and sources are architected into a validated graph, deployed in your cloud, on-prem, or dedicated, then run, returning answers with lineage that feed back into the next build. Intent and sources are architected into a validated graph, deployed in your cloud, on-prem, or dedicated, then run, returning answers with lineage that feed back into the next build. The four stages ## Every stage is visible and open to inspection. 01 · Ingest ### Plain language, or what already exists A project begins with a goal expressed in words, such as predicting which accounts are likely to churn, or with an artifact already in hand: a workflow, SQL script, notebook, or schema. Orama Studio reconciles the goal and existing logic into one problem definition. - Natural-language intent - Existing workflows & scripts - Notebooks & schemas - Source profiles 02 · Architect ### A correct graph, validated before it runs The problem is compiled into an execution graph (connect, clean, model, index, serve) laid out to professional standard. Schemas are type-checked against the real sources, and modeling steps are checked for leakage and point-in-time discipline. The graph is visible and inspectable, and gaps are surfaced as gaps, not papered over. - Connect → clean → model → index → serve - Schema type-checking - Leakage & point-in-time checks - Visible, editable graph 03 · Deploy ### One action, onto infrastructure you run A validated graph is provisioned as Orama Studio nodes and workflows onto the chosen target: a local environment, a Kubernetes cluster, your own cloud account, or an on-prem host. The work deploys into that environment and runs there. - Local → cluster → cloud → on-prem - Node provisioning & reconciliation - Runs in the environment you choose - Self-healing runtime 04 · Run & learn ### Results with lineage and explained drivers The graph executes on the nodes: workflows run, indexes build, and models score. Results include the sources and steps used, freshness, confidence, and the drivers behind each score. Observed outcomes become evidence for the next training run. - Live execution & monitoring - Source-level lineage - Freshness & confidence signals - Drivers behind each score 05 · Verification ## Verification is the path to production Before a pipeline is production-ready, it must pass schema, leakage, point-in-time, and entity-resolution checks. A failed check blocks the pipeline and explains why. ### Schema & type checks Every step is validated against the real source profiles. A reference to data that isn't there is flagged, not silently generated. ### Leakage & point-in-time Modeling steps use point-in-time feature cuts so training only sees information available at prediction time. ### Entity resolution Results resolve to the right entity: a single subsidiary, not several blended into an averaged, flattering blur. Talk to sales ## Start with the outcome Tell us what you want to predict or automate and where the system has to run. We will scope the first production use case with you. Talk to sales --- # Orama Studio · Architecture URL: https://orama.com/products/studio/architecture > Inside the adaptive database: one engine as control plane and data substrate, compute near the data, cost-managed storage, and a distributed, self-healing cluster. Orama Studio · architecture # Inside the engine: an adaptive database Orama Studio is built on an adaptive database that combines the data store, AI compute, model training, and orchestration. The store, training loop, model execution, and agent runtime are one product architecture, designed to run as a distributed system. 01 · Spine & substrate ## One engine, both control plane and substrate. The engine manages platform state, deployments, storage, workflows, models, and the agent runtime. Data and compute share the same distributed substrate, so teams operate one product architecture instead of assembling the AI system from separate frameworks. Diagram: The engine as one plane: a control plane and a data substrate share the same replicated store, deployed in your cloud, on-prem, or dedicated.. One engine holds the platform's own state (control) and serves the data and compute (store, transform, model), deployed in your cloud, on-prem, or dedicated. One engine holds the platform's own state (control) and serves the data and compute (store, transform, model), deployed in your cloud, on-prem, or dedicated. 02 · Why it's fast ## The work comes to the data. ### Compute next to the data Rather than pulling data out to a warehouse, processing it, and pushing it back, the work runs on the node where the data sits. Generated and hand-written logic executes next to the bytes, for lower latency and no warehouse round trip. ### Storage that manages its own cost Frequently used data stays on fast storage and colder data moves to cheaper storage automatically, so repeated questions stop carrying warehouse-scale cost every time they are asked. ### A distributed cluster Nodes form a replicated, self-healing, distributed cluster, where a laptop, a server, or a cloud instance are the same kind of node. Adding capacity is joining a node, and the topology spans from a single machine to a private cloud. Native compute spans CPUs, every major GPU, and ML ASICs, across operating systems and architectures, from a laptop to a data center, the same kind of node. One cluster, every vendor: see deployment → 03 · Learning, not inference ## Train models on your outcomes A general-purpose model can provide an interface, but a call per row does not train a model on your business outcomes. Orama Studio trains the models and pipelines the problem needs on your data. Each model has measured accuracy and explained drivers, and several models can work together when the use case requires an ensemble. General-purpose model ### Provides an interface Useful for language tasks and interaction, without learning a business-specific outcome from each request. Trained on your data ### Learns from outcomes Trained on your signals and evaluated against what happened, with measured accuracy and explained drivers. 04 · Personalized intelligence ## It knows you and your data. Training on your data is the hard part. It has always needed a scarce expert in the loop, and it is where most AI projects quietly stall. Orama Studio closes that gap with automation: it studies you and your data, proposes what is worth building, and trains the agent to do it, so you don't have to already know the answer. ### Auto-suggest You don't have to know what to do with your data. Orama Studio studies it, and how you work, and hands back ranked, high-value use cases, ready to build. ### Automated training It trains an agent on your data automatically: the 'trained on your data' hard part done by the platform rather than by hand. The agent arrives already fluent in your systems. ### Personalized interface Reach Orama Studio over a standard agent interface, and it adapts to you. The same endpoint that serves your agents shapes its tools and responses to your context, so every surface is personalized, not one-size-fits-all. 05 · The name ## Zetta: the platform Metis is built on. Zetta is the platform Metis is built on and Orama Studio runs on: dashboards, pipelines, warehouses, applications, agents, models and model training. The name is a statement about the ceiling, not your data. Scale runs one way: going bigger is the hard direction, and going smaller is free. An architecture built for the extreme, with data where it lives, work next to it, and storage managed for cost, is exactly the architecture that stays fast and economical at the scale you actually run. Solve the largest data problems, and the everyday ones come for free. Talk to sales ## Start with the outcome Tell us what you want to predict or automate and where the system has to run. We will scope the first production use case with you. Talk to sales --- # Orama Studio · Deployment URL: https://orama.com/products/studio/deployment > Deploy in your own cloud, on-premises, or on dedicated infrastructure. One true-hybrid cluster across CPUs, GPUs, and ASICs from any vendor, multiple clouds, and your own hardware. Orama Studio · deployment # Your cloud, on-prem, or dedicated Orama Studio deploys in your own cloud, on-premises, or on dedicated infrastructure. Wherever it runs, the cluster underneath is true hybrid: CPUs, GPUs, and ASICs from any vendor, across operating systems, architectures, and multiple clouds at once, squeezing the most out of whatever hardware you give it. 01 · Where it runs ## Deployed onto infrastructure you run. ### Your cloud Provision Orama Studio nodes into your own AWS, Azure, GCP, or Oracle Cloud account, or your Kubernetes cluster. The platform runs under your identity provider and your controls. ### On-premises Run entirely within your own data center for the most regulated environments, with the same workflow and the same governance, and no dependency on a public cloud. ### Dedicated Run on dedicated infrastructure isolated to your organization. Start on a single machine to validate a use case end-to-end, then promote the same graph to a shared target. Diagram: A single Orama Studio cluster spans your cloud, an on-prem data center, and dedicated infrastructure as one distributed cluster across all three targets.. A control plane coordinates distributed nodes running across your cloud, on-prem, and dedicated targets, one cluster across all three. A control plane coordinates distributed nodes running across your cloud, on-prem, and dedicated targets, one cluster across all three. 02 · The payoff ## Control over cost, flexibility, and resilience. Because the cluster is never bound to one vendor's hardware, prices, or availability. ### Cost Run each workload on the cheapest hardware that fits, and spend capacity you already own or have idle before renting more. Burst into a cloud only when needed, under a cost cap. No premium for standardizing on one vendor. ### Flexibility Mix any vendor, architecture, and cloud in a single cluster. Adopt new silicon as a backend, not a re-architecture, and place each workload where it runs best. ### Resilience An outage or slowdown in one cloud, region, or GPU vendor does not take the cluster down. The cluster self-heals, healthy nodes pick up the pending work, and the job keeps running on whatever is still up. 03 · What it spans ## Every major vendor, in one cluster. Heterogeneous compute: native execution on every backend for high performance at lower cost Table: Supported compute backends and the hardware each targets | Backend | Hardware | | --- | --- | | CPU | x86-64 · ARM | | CUDA | NVIDIA GPU | | ROCm | AMD GPU | | MLX | Apple silicon | | Vulkan | Cross-vendor GPU | | SYCL | Intel · oneAPI / cross-vendor | | GCP TPU | Google ASIC | | AWS Neuron | Trainium · Inferentia | Platforms & architectures macOS Linux Windows x86 ARM Clouds & deployment AWS GCP Azure Oracle Cloud Kubernetes Bare metal Talk to sales ## Start with the outcome Tell us what you want to predict or automate and where the system has to run. We will scope the first production use case with you. Talk to sales --- # Orama Studio · For teams URL: https://orama.com/products/studio/for-teams > Run a white-labelled internal AI platform with centralized governance, verified pipelines, and adoption support from a forward-deployed engineer. Orama Studio · for teams # Build an AI capability across the company Run Orama Studio as your internal AI platform, under your brand and central controls. A forward-deployed engineer helps the first teams choose use cases, ship verified systems, and establish a center of excellence that can support the projects that follow. The roles ## Value by role. 01 · Platform & AI leaders ### Build the center of excellence An internal AI platform has to support many projects without turning the central team into permanent operators for each one. Orama Studio gives project teams a governed path from question to production. The forward-deployed engineer supports adoption and return, while your center of excellence keeps architecture, access policy, and audit consistent. White-labelled internal platform Governance stays centralized Adoption support for early projects 02 · Data scientists & ML engineers ### Ship the whole production pipeline A useful model still needs source connections, leakage-free features, verification, serving, monitoring, and governance. Orama Studio builds the pipeline, applies leakage and point-in-time checks, and deploys it onto governed infrastructure. Data scientists can inspect and revise the graph without owning a separate platform project. A correct pipeline scaffolded, not hand-built Leakage & point-in-time checks built in Deployed without a separate platform effort 03 · Business & analytics teams ### Grounded answers, within governed boundaries. Getting an answer that isn't already on a dashboard usually means asking a technical team and waiting. Decisions get made on stale reports while the answer sits in data nobody had time to query. Once a platform is set up, questions are asked in plain language and answered from live data, scoped to the access each person already holds, and returned with their sources, freshness, and confidence shown. A recurring check can be scheduled and delivered where teams already work. Plain-language questions, grounded answers Scoped to existing access Sources, freshness & confidence shown Talk to sales ## Start with the outcome Tell us what you want to predict or automate and where the system has to run. We will scope the first production use case with you. Talk to sales --- # Company · Orama URL: https://orama.com/company > Orama builds Metis, the Enterprise Intelligence OS, on one belief: you cannot have AI, or intelligence, without data availability. Who we are, how we work, and how to reach us. Company # The company behind the Enterprise Intelligence OS Orama builds Metis, the Enterprise Intelligence OS. We are the team behind Orama and the open-source OramaJS library, and we started with search because search is where an organization first discovers whether its data is actually available. Everything since follows one belief: you cannot have AI, or intelligence, without data availability. Talk to sales See the platform orama · the suite - 01OramaJS The open-source search library the team started from. - 02Metis The Enterprise Intelligence OS, built by Orama on Zetta. - 03Orama Studio The surface of Metis: business questions into production models, pipelines, and agents on your infrastructure. ## What we build Metis is built on Zetta, one platform for everything that runs on your data: dashboards, pipelines, warehouses, applications, agents, models and model training, on one binary that runs as a node or a cluster. Orama Studio is the surface over it. OramaJS is the free, open-source search library it grew from. Metis runs in your cloud, on-prem, or dedicated. ## Honest about our size We would rather do a few things properly than many things loudly: one operating system, and no theater about scale. You do not have to take that on faith. The core library's code is open, the platform is inspectable, and how we sell is written on the site in plain words. ## How we sell The open-source library is free. Metis is landed by forward-deployed engineers, inside your organization: we inventory the decisions that matter, score the data each one needs, raise what is short, run what the data can carry, and measure the returns on a metric you already report. Where deployments differ, we scope them on a call: book a call and we start with your decisions. ## Legal identity Orama is operated by OramaSearch, Inc., a Delaware corporation. ## Contact Sales and everything else starts at the contact page: a short form, read by a person. For legal, privacy, or security matters: legal@orama.com. Talk to sales ## Talk to a person Book a call, or read how the platform is built. Talk to sales See the platform --- # Talk to sales · Metis URL: https://orama.com/contact/sales > Tell us what you need and book a call with the Orama team. Sales for Orama and enterprise deployments; questions welcome. Contact sales # Talk to us about Metis Tell us what you are trying to do. Your request goes straight to the team. Step 1 of 2 ## Start with the basics Name Work email Company Continue What you share here is used to respond to you. Details in our privacy policy. orama · your request read by a person - Request received Your form reaches the Orama team through the same secure submission flow. - Reviewed by the team A person reads every request and routes it to someone who can help. - A contextual reply We respond to what you shared, without an automated sales sequence. What happens first ## Start with your decisions The first conversation is about your decisions, not our product. Bring the three to seven that matter most in one unit, and the data you have. - 01 Inventory the decisions The decisions with the most at stake, written down with their owners, alternatives, objective, frequency, and the metric each one moves. - 02 Score the data Each decision's availability profile on the ten dimensions, with evidence, against the instrument it needs; the binding gate named, with the discipline that owns the fix. - 03 State the unit's score Its Corporate Intelligence Capability (CIC) score, 0 to 10, with the profiles behind it, and what the next score admits. - 04 Write the returns chain The investment, the score to reach, the instruments that reaches, the metric it moves, and the baseline it moves from, agreed with finance before any work: return on investment (ROI), written down first. What the unit sees at the end: its decisions, named and owned; a profile per decision; its first CIC score; and a returns statement it can hold us to. ## Prefer to explore first? See how the platform is built, or start with the open-source OramaJS library and its documentation. See the platform Explore OramaJS --- # Privacy policy — Orama URL: https://orama.com/legal/privacy > How Orama collects, uses, and protects your data. # Privacy policy The plain-language version: we collect what we need to run the site, answer you, and provide the Services; we use analytics but no advertising tags; we do not sell your personal information; and you can ask us about or delete your data at any time. Last updated September 14, 2026 ## Who is responsible for your data OramaSearch, Inc., a Delaware corporation, is the entity responsible for the personal data described in this policy. This policy explains how Orama collects, uses, and shares personal information when you visit orama.com, contact us, create an account, or use Metis, Orama Studio, Zetta, and our other services (the “Services”). When a customer uses the Services to process data about its own customers, employees, or other people, Orama processes that data on the customer’s behalf, as a processor or service provider, under our agreement with that customer. The customer’s own privacy policy governs that data, and requests about it should go to the customer first. ## What we collect Information you give us: - Sales and contact requests: your name, work email, and company, and, if you choose to share them, your company size, what you want Metis to do, and when you want it live. - Account information: your name, email address, organization, and sign-in details when you create or use an account. - Billing information: billing contact and payment details if you buy online. Payments are handled by our payment provider, and we do not store full card numbers. - Communications: the content of emails, support requests, and notes from calls and meetings with us. Information collected automatically: - Site analytics: the pages you view, the links and buttons you click, the page that referred you, your browser and device type, and approximate location derived from your IP address. This is collected through a third-party analytics service, which uses cookies and similar technologies. - Logs: IP address, browser user agent, request details, and timestamps recorded by our servers and hosting providers to operate and secure the site and the Services. Customer Data: the data you or your organization connect to or create in the Services. We process it to provide the Services, as described in our terms of service and your agreement with us. ## How we use it - To provide, operate, maintain, and support the Services, including creating and managing accounts. - To respond to your requests and prepare for conversations you ask to have with us. - To process payments and manage billing. - To understand how the site is used and improve it. - To protect the security and integrity of the site and the Services, and to detect and prevent fraud and abuse. - To send you information about the Services you asked for or that relates to your account. You can opt out of marketing emails at any time. - To comply with legal obligations and enforce our terms. Where the GDPR or similar laws apply, we rely on these legal bases: performing a contract with you or taking steps you request before one; our legitimate interests in running, securing, and improving our business, balanced against your rights; your consent, where we ask for it; and compliance with legal obligations. ## Cookies and analytics The site uses a third-party analytics service to measure visits and how people use the site. It sets cookies and receives the analytics information described above on our behalf. The site does not run advertising or retargeting tags, and we do not use analytics data to show you ads. You can block or delete cookies in your browser settings. The site continues to work if you do. ## Who we share it with We run the site and the Services on third-party providers that process data on our behalf for hosting and content delivery, sign-in, payments, and site analytics. They may use it only to provide their services to Orama. We also share personal information: - With service providers that help us operate the business, such as email, customer support, and productivity tools, under contracts that limit their use of it to serving Orama. - When required by law, legal process, or a lawful request from public authorities, or when necessary to protect the rights, property, or safety of Orama, our customers, or others. - In connection with a merger, acquisition, financing, or sale of all or part of our business, subject to this policy. - With your consent or at your direction. We do not sell personal information, and we do not share it for cross-context behavioral advertising. ## International transfers Orama is based in the United States, and our providers may process data in the United States and other countries. Where the law requires it, we protect personal information transferred out of the European Economic Area, the United Kingdom, or Switzerland with appropriate safeguards, such as the European Commission’s standard contractual clauses. ## How long we keep it We keep your data while your account is active. When you close your account or ask us to delete your data, it is deleted within 30 days. We keep sales and contact requests for as long as we are in conversation with you and for a reasonable period afterwards, unless you ask us to delete them sooner. We keep information longer where the law requires it, such as billing records, or where it is needed to resolve disputes or enforce our agreements. ## How we protect it We use administrative, technical, and physical safeguards designed to protect personal information, including encryption in transit and access controls that limit who can reach it. No method of transmission or storage is completely secure, so we cannot guarantee absolute security. ## Your rights Depending on where you live, you may have legal rights over your personal data — including access, correction, deletion, portability, and objection to certain processing (for example under the GDPR or the CCPA). Email legal@orama.com to exercise them; we respond to every verified request. If you are in the European Economic Area or the United Kingdom, you may also lodge a complaint with your local data protection authority. If you are a California resident, you have the right to know what personal information we collect and how we use and disclose it, to delete and correct it, and not to be discriminated against for exercising these rights. You may use an authorized agent, and we may ask you to verify your identity before acting on a request. ## Children The site and the Services are for businesses and are not directed to children. We do not knowingly collect personal information from anyone under 16. If you believe a child has given us personal information, contact us and we will delete it. ## Changes to this policy We may update this policy from time to time. We will post the updated version on this page and change the date at the top, and if a change is material, we will give more prominent notice. ## How to reach us about your data Email legal@orama.com for any request about your data — access, correction, deletion, or a question this policy does not answer. A person reads every request. --- # Terms of service — Orama URL: https://orama.com/legal/terms > The terms that govern your use of Orama products and services. # Terms of service The plain-language version: your data stays yours, you use Orama lawfully and within the acceptable use policy, a signed agreement with us takes priority over these terms, and neither side is liable beyond the limits below. Last updated September 14, 2026 ## Who this agreement is between This agreement is between you and OramaSearch, Inc., a Delaware corporation ("Orama", "we"). These terms govern your access to and use of the Orama website at orama.com, Metis, Orama Studio, Zetta, our hosted services, APIs, and documentation, and any related services we provide (together, the “Services”). By accessing or using the Services, you agree to these terms. If you use the Services on behalf of a company or other organization, you confirm that you are authorized to accept these terms for it, and “you” means that organization. You must be at least 18 years old to use the Services. ## How these terms fit with other agreements Most customers buy Metis under a signed agreement with Orama, such as a master services agreement and an order form. Where a signed agreement covers the same subject as these terms, the signed agreement controls. OramaJS and the other open-source software we publish are licensed under the open-source license that accompanies each project. Nothing in these terms limits the rights that license gives you. Your use of the Services is also governed by our acceptable use policy (orama.com/legal/acceptable-use) and our privacy policy (orama.com/legal/privacy), which form part of these terms. ## Accounts Some Services require an account. You agree to give accurate information, keep your sign-in credentials confidential, and tell us promptly at legal@orama.com if you believe your account has been accessed without authorization. You are responsible for all activity under your account and for the people you allow to use the Services, including making sure they follow these terms. ## Your data “Customer Data” means the data, content, and files you or your users connect to, upload to, or create in the Services, and the outputs the Services produce from them. As between you and Orama, you own your Customer Data, and we claim no ownership of it. You grant Orama a worldwide, non-exclusive, limited license to host, copy, process, transmit, and display Customer Data only as needed to provide, secure, and support the Services for you, and as required by law. You are responsible for your Customer Data: that you have the rights, permissions, and any notices or consents needed to connect it to the Services and to have us process it, and that your use of it complies with applicable law. When you close your account or ask us to delete your data, deletion follows the retention terms in our privacy policy. Before closing an account, you are responsible for exporting any Customer Data you want to keep. ## AI outputs and decisions The Services use models and automated analysis to produce outputs such as answers, scores, forecasts, explanations, and recommended actions. Outputs are generated from the data available to the Services and can be incomplete or wrong. You are responsible for reviewing outputs before relying on them and for the decisions you make with them. Outputs are not professional legal, financial, medical, or other regulated advice. Any score, return, or target we discuss with you, including a return-on-investment target for an engagement, is an estimate and not a guarantee, unless a signed agreement expressly says otherwise. ## Usage data and feedback We collect information about how the Services are used and perform, such as feature usage, errors, and system metrics. We use it to operate, secure, support, and improve the Services, and we only share it outside Orama in aggregated or de-identified form that does not identify you or your users or include Customer Data. If you send us suggestions or feedback, we may use them without restriction or obligation to you. ## Fees and payment Fees for the Services are set out in your order form, or shown to you before you buy if you buy online. Unless your order says otherwise, fees are stated in US dollars, exclude taxes, and are payable in advance. You are responsible for any sales, use, value-added, or similar taxes, other than taxes on Orama’s income. If an undisputed amount is overdue, we may suspend the affected Services after giving you notice and a reasonable opportunity to pay. ## Plans purchased online If you buy a plan online rather than under an order form, the following also applies. There is no free trial. You may request a refund of the Orama base fee within 30 days of your first payment. Metered usage charges are final and are not refundable. You can cancel at any time. Cancellation stops future base and usage billing at the end of the current paid period, and access continues through that period. The base fee is refundable only under the 30-day first-payment policy; usage charges are final. Other payments are not refunded or prorated. ## Third-party services The Services can connect to systems you choose, such as your databases, warehouses, and business applications, and to third-party models and tools. Your use of a third-party service is governed by your agreement with its provider. Orama is not responsible for third-party services, and when you connect one, you authorize us to exchange data with it as needed to provide the Services. ## Our intellectual property Orama and its licensors own the Services, including the software, models we build, documentation, and all related intellectual property. Subject to these terms and any order, we grant you a limited, non-exclusive, non-transferable right to use the Services for your internal business purposes during your subscription or engagement. Except as allowed by an open-source license or applicable law, you may not copy, modify, or create derivative works of the Services; reverse engineer, decompile, or attempt to extract the source code of any part of them that is not open source; or sell, resell, sublicense, or provide the Services to third parties as a service. The Orama, Metis, Orama Studio, Zetta, and OramaJS names and logos are our trademarks. You may not use them in a way that suggests we endorse you without our written permission. ## Confidentiality Each of us may receive non-public information from the other that is marked confidential or that a reasonable person would understand to be confidential, including Customer Data and non-public details of the Services. The receiving party will use it only to perform under these terms, protect it with at least reasonable care, and disclose it only to people who need to know it and are bound by similar obligations. These obligations do not apply to information that is or becomes public through no fault of the receiving party, was already known to it, is independently developed, or is received lawfully from a third party. A party may disclose confidential information when required by law, after giving the other party notice where legally permitted. ## Security We maintain administrative, technical, and physical safeguards designed to protect the Services and Customer Data against unauthorized access, loss, and disclosure. No system is perfectly secure, and you are responsible for configuring and using the Services securely on your side, including managing who has access. ## Suspension We may suspend access to all or part of the Services if we reasonably believe it is necessary to prevent harm to the Services, to other customers, or to third parties, to respond to a violation of these terms or the acceptable use policy, or to comply with law. Where the situation allows, we will give you notice first, limit the suspension to what is necessary, and restore access once the issue is resolved. ## Termination You can stop using the service and cancel at any time. We may suspend or terminate an account that violates these terms or the acceptable use policy; when reasonably possible we contact you first. Termination for a violation does not create a refund right beyond the 30-day base-fee policy. Sections of these terms that by their nature should survive termination survive it, including those on your data, fees owed, intellectual property, confidentiality, disclaimers, limitation of liability, indemnification, and governing law. ## Warranties and disclaimers Each of us confirms that it has the authority to enter into these terms. Any warranty for a paid engagement is the one stated in your signed agreement. Except as expressly stated in these terms or a signed agreement, the Services are provided “as is” and “as available.” To the fullest extent permitted by law, Orama disclaims all other warranties, express or implied, including warranties of merchantability, fitness for a particular purpose, title, non-infringement, and that the Services or their outputs will be uninterrupted, error-free, or accurate. ## Limitation of liability To the fullest extent permitted by law, neither party will be liable for any indirect, incidental, special, consequential, or punitive damages, or for lost profits, revenue, goodwill, or data, even if it was advised that they were possible. To the fullest extent permitted by law, each party’s total liability arising out of or relating to these terms or the Services will not exceed the amounts you paid to Orama for the Services in the twelve months before the event giving rise to the claim. These limits do not apply to your payment obligations, to either party’s indemnification obligations, or to liability that cannot be limited under applicable law. ## Indemnification You will defend Orama and its affiliates, officers, and employees against any third-party claim arising from your Customer Data or from your use of the Services in violation of these terms, the acceptable use policy, or applicable law, and pay the resulting damages, costs, and reasonable attorneys’ fees finally awarded or agreed in settlement. Orama will give you prompt notice of the claim, reasonable cooperation, and control of its defense, and you will not settle a claim that imposes an obligation on Orama without its consent. ## Governing law and venue This agreement is governed by the laws of the State of Delaware, without regard to its conflict-of-law rules. Disputes are resolved in the state or federal courts located in Delaware, and both sides consent to their jurisdiction. ## Changes to these terms We may update these terms from time to time. We will post the updated version on this page and change the date at the top. If a change is material, we will tell you in advance, by email or in the Services, before it takes effect. By continuing to use the Services after a change takes effect, you accept the updated terms. Changes do not alter a signed agreement unless both parties agree in writing. ## General - Entire agreement: these terms, together with the policies they reference and any signed agreement, are the entire agreement between you and Orama about the Services. - Assignment: you may not assign these terms without our written consent, except to a successor to substantially all of your business. We may assign them to an affiliate or to a successor in a merger, acquisition, or sale of assets. - Force majeure: neither party is liable for a delay or failure caused by events beyond its reasonable control, other than a failure to pay. - Export and sanctions: you will comply with applicable export control and sanctions laws, and you will not use the Services in, or make them available to people in, a country or region, or to a person, subject to comprehensive sanctions. - Independent parties: the parties are independent contractors, and nothing in these terms creates a partnership, joint venture, or employment relationship. - Severability and waiver: if a provision is found unenforceable, the rest of these terms stay in effect. A failure to enforce a provision is not a waiver of it. - Notices: we may send notices to the email address associated with your account or order. You may send notices to legal@orama.com. ## Contact Questions about these terms can be sent to legal@orama.com. --- # Acceptable use policy — Orama URL: https://orama.com/legal/acceptable-use > The rules for acceptable use of Orama products and services. # Acceptable use policy The plain-language version: use Orama lawfully, only with data you have the right to use, keep people in charge of consequential decisions, and do not attack the Services or use them to harm anyone. Last updated September 14, 2026 ## Who this policy applies to This policy applies to everyone who uses the Orama website, Metis, Orama Studio, Zetta, our hosted services, APIs, and related services (the “Services”), including the people a customer allows to use its account. It is part of our terms of service. If you are a customer, you are responsible for making sure your users follow it. The examples below are not a complete list. We may treat any use that is similar in purpose or effect as a violation. ## Unlawful and harmful activity Do not use the Services to: - Break any law or regulation, or help anyone else do so. - Create, store, or share child sexual abuse material, or any content that sexualizes minors. - Promote or incite violence, terrorism, or violent extremism, or threaten, harass, stalk, or intimidate anyone. - Commit fraud, run scams or phishing, impersonate a person or organization, or misrepresent where content comes from. - Send spam or other unsolicited bulk messages. - Infringe or misappropriate intellectual property, trade secrets, or other proprietary rights. - Develop, produce, or acquire biological, chemical, nuclear, or radiological weapons, or other weapons intended to cause mass harm. - Distribute malware, or create content designed to deceive people about elections or civic processes. ## Data you connect to the Services - Only connect, upload, or process data you have the legal right and any required authorization to use in the Services. - When the data includes personal information, have a lawful basis for processing it, give any notices and obtain any consents the law requires, and honor the rights of the people it describes. - Do not process special categories of personal data, such as health, biometric, genetic, or financial account data, or government identifiers, unless the law allows it and your agreement with Orama expressly covers it. - Do not collect data by scraping or other automated means in violation of a website’s terms or applicable law, and do not use the Services to access data you are not authorized to access. ## AI and automated decisions - Do not use outputs to make decisions that have legal or similarly significant effects on people, such as decisions about employment, credit, housing, insurance, education, healthcare, or access to essential services, without meaningful human review and compliance with the laws that apply to those decisions. - Do not use the Services for unlawful surveillance or tracking of individuals, or to identify people from biometric data without a lawful basis. - Do not use the Services to discriminate unlawfully against anyone based on a protected characteristic. - Do not present outputs as human-made where the law requires disclosure that they were generated by AI, and do not use the Services to create deceptive impersonations of real people. - Do not attempt to bypass the safeguards built into the Services or the models they use. ## Security and integrity of the Services Do not: - Access, or attempt to access, the Services, other accounts, or Orama systems without authorization, or probe, scan, or test their vulnerabilities, except under our coordinated disclosure process below. - Circumvent authentication, access controls, usage limits, or billing. - Interfere with or disrupt the Services, including by overloading them or launching denial-of-service attacks. - Use the Services to attack, probe, or gain unauthorized access to any other system or network. - Share credentials, or resell or provide access to the Services, except as your agreement with Orama allows. - Reverse engineer any part of the Services that is not open source, except where the law expressly permits it. ## What happens on a violation Violations can lead to content removal, suspension, or termination of the account, depending on severity. Where the situation allows it, we contact you before acting. Unlawful activity may be reported to the relevant authorities. We may investigate a suspected violation, and we may remove or disable access to content that violates this policy. You agree to cooperate with a reasonable investigation. ## Reporting abuse and security issues Report abuse, security issues, or content that violates this policy to legal@orama.com. If you find a security vulnerability, tell us before disclosing it publicly and give us a reasonable time to fix it. Please do not access or change data that is not yours, degrade the Services, or keep any data you encounter while testing. We will not pursue legal action against good-faith research that follows these rules. ## Changes to this policy We may update this policy as the Services and the law change. We will post the updated version on this page and change the date at the top. --- # Zetta, the platform · Metis URL: https://orama.com/architecture > Zetta is the software side of Metis, one platform for everything that runs on your data: dashboards, pipelines, warehouses, applications, agents, models and model training, on one distributed platform that runs in your cloud, on-prem, or dedicated. Zetta · the platform # One platform for everything that runs on your data. Metis is built on Zetta Metis is the Enterprise Intelligence OS. Zetta is its software side: one distributed platform, one binary, that scales from a single node to a cluster and runs where your data is. Talk to sales Batteries Included ## Six instruments, one platform. Zetta holds the data, runs the pipelines, trains the models, serves the agents, and renders the dashboards, on the same nodes. Nothing has to leave the platform to become intelligence, and nothing on it can outrun the data it can reach. Diagram: Zetta, one platform for everything that runs on your data: dashboards, pipelines, warehouses, applications, agents, and models and model training, on one platform. - 01 Dashboards What happened and why, over conformed data with one definition per metric, for the decisions a team actually makes. The descriptive and diagnostic tiers of the operating system. - 02 Pipelines Intent compiled into a typed pipeline, connect, clean, model, index, serve, verified before it runs. The pipeline is how raw capture becomes conformed, governed, consumption-ready data. - 03 Warehouses The store, in whichever of three ways you choose: Zetta as the store with your warehouses connected as sources, Zetta running over the warehouse you already have, or Zetta becoming the store over time. - 04 Applications The applications an enterprise runs on its intelligence, and the applications that create the data a decision lacks: capture at the point of decision, workflow applications in place of spreadsheets, extractors, experiments, and collectors. - 05 Agents Software that plans and acts across systems, under per-action permission and full provenance, with every action recorded as a decision. The tier with the least tolerance for missing data, because its failure is an action, not a belief. - 06 Models and model training Models trained natively on your data, inside your environment, with calibrated confidence and explained drivers. The predictive tier, and the engine underneath grounded answers and agents. Two sides ## Software, and how it is executed. The software side of Metis is Zetta. The execution side is forward-deployed engineering and the theory the operating system is built on. Neither works without the other. Diagram: The two sides of Metis: the software side, Zetta, and the execution side, forward-deployed engineering and the theory. The execution side lands the software inside an organization. ### The software side: Zetta - One binary, one cluster A single native executable is a node. Nodes join into one replicated, self-healing cluster that spans a laptop, a data center, and any cloud. - Compute next to the data Pipelines, training, indexing, and serving run on the nodes that hold the data, rather than pulling it out to be processed elsewhere. - Verified before production Schema, leakage, point-in-time, and entity-resolution checks gate every pipeline. A failed check blocks it and says why. - Driven programmatically, observed visually Every action a person can take in Orama Studio is a tool an agent can call, identically. The surface exists to make the platform's work legible. - Governed by construction Identity, scoped keys, consent for anything that touches a cloud provider, and an append-only audit of every call, allow and deny alike. ### The execution side: forward-deployed - Decisions first Inventory the decisions that matter, with their owners, alternatives, and the metric the unit already reports. - Data availability, scored Each decision's data scored on ten dimensions against the instrument it needs, with evidence, and the weakest gap named. - Raise what is short Connect what exists, conform and govern it, and create the data that was never captured, inside your tools. - Run what the data can carry Reports, models, grounded answers, and agents, one decision at a time, each admitted when its data meets the bar. - Measure outcomes and returns Every decision logged with its expectation; outcomes measured at the horizon; returns stated in your metric against a signed baseline. Orama Studio is the surface over Zetta. OramaJS is the open-source search library the team started from. Where your data lives ## Meets your data where it is. All three are supported, chosen per customer with your platform team. The method is the same in each; what differs is who owns the store and how currency and access are measured. Diagram: The three ways Metis meets a customer's data: Zetta as the store with warehouses connected as sources; Zetta running over the warehouse the customer already has; and Zetta becoming the store over time. ### Zetta as the store Your warehouses, databases, and applications connect as sources. Zetta holds the intelligence data and runs everything above it. Scoped with sales. Sales-led ### Zetta over your warehouse Your warehouse remains the store. Zetta runs pipelines, models, and agents over the tables where they sit, without copying them. Scoped with sales. Sales-led ### Zetta becomes the store Start in either of the first two and let Zetta take over the store progressively, at your pace, as the warehouse's job moves onto the platform. Scoped with sales. Sales-led Zetta runs as one node or a cluster, in your cloud, on-prem, or dedicated. Grounded in reality ## Why an operating system, and not another model. You cannot have AI, or intelligence, without data availability. Everything on this page exists to make that sentence operational. ### Data availability Does the data a decision needs exist, can it be reached and lawfully used, is it true, current, and understood, and is it fit for the instrument. Ten dimensions, scored per decision, combined as a weakest link. ### Decisions The unit everything hangs off. Metis is decision-driven, not data-driven: availability is scored against a decision, never in the abstract, and an instrument is run for a decision only when its data can carry it. ### Outcomes Every decision is logged with what was expected. Outcomes are measured at the horizon and calibrated over many decisions, so skill is separated from luck. ### Returns Return on investment (ROI), stated in the unit's own metric against a signed baseline: what was invested, how far availability rose, which instruments that made possible, and what came back. Every figure we publish is a verified claim in our registry. Where a number is not verified, we state the method and publish nothing. ## See it on your own decisions Bring the decisions that matter and the data you have. We score availability first, then build what the data can carry. Talk to sales --- # Why it works · Metis URL: https://orama.com/theory > You cannot have AI, or intelligence, without data availability. How Metis is grounded in reality: data availability scored per decision, decisions as the unit, instruments the data can carry, outcomes measured at the horizon, and returns stated in your own metric. Why it works # Successful AI and intelligence needs the right data. Every report, model, agent, and dashboard is downstream of the data it can reach, and nothing built above the data can add information the data does not contain. Metis is the Enterprise Intelligence OS built on that belief: it measures the data first, raises it where it's short, and only then runs what the data can carry. Talk to sales See the platform How we found out ## Years of building AI taught us where it breaks. For years, we built AI systems and agents for organizations, noting where they succeeded and where they failed. Better models didn't fix it. Better data did. Grounded answers are capped by what retrieval can find. An index over a stale corpus answers fluently about a world that no longer exists. A question about a fact nobody recorded has no answer, however good the search. Every fix on offer sat above the data: a bigger model, a different embedding, an agent framework. Some helped at the margin. Few of them gave the system what it was missing. “Successful AI projects need a real problem to solve, and the right data.” — Andrew Ng. Fiscally responsible AI is anchored to ROI, and AI investments that are not will fail. No processing can make an output carry more information about the world than the data it came from. And because the same data is available for one decision and not another, that ceiling can be scored, per decision, before anything is built. We stopped asking which AI to buy. We started asking what problem it solves, whether the data can carry it, and what it returns. Metis is designed around this central principle. We target a 400% ROI (over 3 years) on every engagement, and we measure it in the unit's own metric, with a signed chain from the investment to the returns. The chain ## From data you have to returns you can measure. Intelligence is a chain, and it breaks at its weakest link. Data availability comes first because it is the ceiling; decisions come next because availability is only ever measured for a decision. Metis makes every link a thing you can see and score. Diagram: The chain from data to returns: your data is scored for availability against a decision, an instrument runs only when the data can carry it, the decision is logged, and the outcome and the returns are measured. - 01 Data availability Does the data a decision needs exist, can it be reached and lawfully used, is it true, current, and understood, and is it fit for the instrument. Scored per decision, on ten dimensions. - 02 Decisions The unit everything hangs off. The same data is available for one decision and not for another, so availability is always scored against a decision, never in the abstract. - 03 Instruments Reports, diagnostics, predictive models, recommendations, grounded answers, and agents. Each tier demands more of the data than the one before it, and the tolerance for missing data falls toward zero as instruments move from describing to acting. - 04 Outcomes Every decision is logged with what was expected. Outcomes are measured at the horizon and compared, in aggregate, so skill is separated from luck. - 05 Returns Return on investment (ROI), stated in the unit's own metric: what was invested, how far availability rose, which instruments that made possible, what the metric did, and what came back. 01 · Data availability ## Ten dimensions, three questions. Availability is not a single score and it is not on or off. It is ten measurable dimensions in three bands, and a decision's data is only as available as its weakest one. Diagram: The ten availability dimensions in three bands, in dependency order: reach, then fidelity, then fitness. A decision's data is only as available as its weakest dimension. ### Can the instrument get it, lawfully Existence: was it recorded at all, at the grain the decision needs. Accessibility: can the systems and people that serve the decision reach it. Permission: may it be used for this purpose, at this grain. ### Can the instrument trust and understand it Quality: is it true to what it claims. Currency: is it current when the decision is made. Provenance: where it came from, who owns it, what happened to it. Meaning: does everyone, human and machine, compute the metric the same way. ### Does it serve this decision through this instrument Relevance: would knowing it change the choice. Sufficiency: is there enough history, enough labeled outcomes, enough of the rare cases. Shape: can the instrument consume it without a preparation project. Scores combine as a weakest link, never an average. A dataset with perfect quality that the instrument may not lawfully use has an availability of zero for that decision. 02 · Decisions ## Find data for a purpose, not a purpose for data. Relevance is not a property of data; it is a relation between data and a decision. So the ceiling stays where the theory puts it, on data availability, and the decision is what availability is measured for: Metis inventories the decisions an organization has to make, derives what each one needs, and scores the data against that. That is decision-driven, not data-driven. Data-driven in the wrong sense starts from what exists, looks for uses, and ends up answering only the questions the data happens to be able to answer. ### Inventory the decisions For each unit, the few decisions with the largest value at stake: alternatives, objective, frequency, the owner, and the metric the unit already reports. ### Derive the variables What, if known, would change the choice. Priced by the value of information, so data that would not change anything is not collected. ### Score and admit Each decision's data is scored on the ten dimensions against the instrument it needs. An instrument runs when the data can carry it, and the gap is named when it cannot. This is why the pivot away from starting with a model matters: a decision two gaps from working beats a model with no data underneath it. 03 · Instruments ## Six tiers, tightening toward action. Every instrument answers a different question and demands a different profile of the data. The action-facing demands rise monotonically. Diagram: The six instrument tiers, from Tier 1 Descriptive to Tier 6 Agentic: descriptive, diagnostic, predictive, prescriptive, generative, agentic. The demands on the data tighten at every step, and the failure of the last tier is an action, not a belief. ### Descriptive What happened. Reports and dashboards over conformed data with one definition per metric. ### Diagnostic Why. Joining across domains, with the causes recorded and not only the effects. ### Predictive What will happen. Models trained on representative, labeled, sufficient history, with calibrated confidence and explained drivers. ### Prescriptive What to do. Recommendations that need something the others do not: the record of past decisions and their outcomes. ### Generative Explain and answer. Grounded answers whose accuracy is capped by what retrieval can find, not by the size of the model. ### Agentic Agents that execute. The most demanding profile: current data, per-action permission, and full provenance, because the failure is an action, not a belief. AI is the engine of the last four tiers. It is a small part of the operating system, and it is the part most likely to be funded above data that cannot support it. 04 · Outcomes ## Quality decisions, not lucky ones. A decision is a process. A choice is its output. An outcome is what the world does next, and it includes chance. Metis scores the decision, logs the choice, and measures outcomes only in aggregate. Higher quality decisions mean better outcomes and returns, over many decisions, and that is the reason to score them. Diagram: A decision is a process that produces a choice; an outcome is what the world does next and includes chance. The decision and its expected outcome are logged at commitment; the outcome is measured at the horizon. ### Logged at commitment Alternatives considered, the choice, the rationale, the expected outcome, and who decided, human or agent. The log is itself data, and it is the data that prescriptive instruments need. ### Measured at the horizon Outcomes are joined to their decisions and reviewed by cause: was the frame wrong, the data unavailable, the reasoning flawed, or the commitment unmet. ### Calibrated over many Across many decisions of the same kind, stated confidence is compared with what happened. Calibration is a measurable, improvable skill, and it is how an organization knows its instruments are telling the truth. A single outcome says almost nothing about a decision. A hundred say a great deal. 05 · Returns ## Invested this much, moved this metric, returned this. Returns are return on investment (ROI): intelligence and AI investments stated in the metric the unit already reports, with the chain written down before the work starts: the investment, how far availability rose, which instruments that made possible, what the metric did against its baseline, and what came back. ### A baseline and a counterfactual The metric's value and trend before the work, and an agreed way to say what would have happened anyway. Before and after alone is not attribution. ### A fixed window and an owner The measurement period is set at the start, and a named executive defends the metric and the value per unit that finance signed. ### Adoption, not deployment An instrument counts only for the share of decisions actually taken with it. A forecast nobody uses returns nothing, however accurate. This is what makes outcome-based commitments possible: baseline, counterfactual, window, and validation are objects in the system, not a negotiation afterwards. What you might be thinking ## The objections, answered from the theory. Each of these is reasonable, and each has an answer that follows from the ceiling. We already have a warehouse. Our data is fine. A warehouse raises accessibility and quality. It does not raise relevance, and it can be built without a single decision written down. Scored against the decisions that matter, the usual surprises are meaning (how many definitions of one metric exist across teams) and existence (whether the unit records its own decisions and what happened next). We will use a bigger model. A bigger model approaches the floor the data sets and cannot go below it. Nothing built above the data can add information the data does not contain. Once a model clears adequacy for the decision, the next dollar goes to the weakest availability dimension, not to a larger model. Our data is not ready, so we cannot start. Scoring is how you start. Reports and diagnosis work at the Foundational tier, and a decision two gaps from working beats a decision six gaps from working, even if the second is more valuable. Run what the data can carry today, and close the named gate for what it cannot. Agents will handle the integration. Agents are the tier with the least tolerance for missing data: currency, permission, provenance, and accessibility all near their maximum at once, because the failure is an action, not a belief. A standard tool interface in front of an ungoverned system gives an agent fast access to bad data. This is data governance by another name. Governance is a function of availability, not overhead: for an agent, lineage and permission are how it knows what it may do and whether its inputs are true. But the deliverable is never a catalog. It is a changed decision with a measured consequence, stated as returns in a metric the unit already reports. We do not have time for an assessment. The assessment is three to seven decisions per unit, scored with evidence, on a fixed cadence. The alternative is finding out after the spend which instruments the data could not carry. Is this for finance, or supply chain, or service? The journey is measured on the metric the unit already reports: close-cycle days, stock-out rate, cost per interaction, forecast accuracy. The binding gates cluster by function, and they are known. How do we know the score is not gamed? Evidence per anchor, sampled verification, weakest-gate scoring, an independent scorer for anything tied to money, and instrument outcomes as ground truth: if the score says model-ready and the forecast is not trusted, the score is wrong and is re-anchored. How it lands ## Forward-deployed, inside your organization. The method cannot be run from outside, and seeing the map is not walking the path: scoring ten dimensions with evidence, against the column each instrument requires, and closing the binding gate in the right discipline is the engagement, not the reading. Orama's forward-deployed engineers land Metis with your decision owners, on your systems, on infrastructure you choose: your cloud, on-prem, or dedicated. - 01 Inventory the decisions that matter and who owns them - 02 Score each decision's data on the ten dimensions, with evidence - 03 Write the returns chain and get it signed - 04 Raise availability: connect what exists, and create the data that was never captured - 05 Run the instruments the data can carry - 06 Log every decision, measure outcomes, re-score on a cadence, and hand the loop to your own teams Every engagement leaves the platform stronger for the next one, so the second organization's first step is faster than the first's. ## See it on your own decisions Bring the decisions that matter and the data you have. We score availability first, then build what the data can carry. Talk to sales See the platform --- # Your journey · Metis URL: https://orama.com/journey > What Metis means for an organizational unit: the decisions that matter get scored, the data underneath them gets raised, the instruments the data can carry get run, and the score moves. The journey in six steps, and what changes by role. Your journey # What Metis means for your organization Metis raises an organization's ability to act intelligently, one unit and one decision at a time, and shows the score moving as it does. Data availability comes first, and it is scored decision-driven, not data-driven: for the decisions that matter, not for the data that happens to exist. This is what the journey looks like, what is measured along the way, and what comes back as return on investment (ROI). Talk to sales See the CIC scale Where it happens ## The unit is where the score lives Metis works with the parts of an organization that own decisions and a metric: a function, a business unit, a region. Each one gets a score, and the journey is raising it. ### A unit owns decisions Finance closes the books, sales forecasts the quarter, supply chain replenishes the shelves, service resolves the ticket. Each of those is a decision with an owner, alternatives, and a consequence. ### A unit reports a metric Close-cycle days, forecast accuracy, stock-outs, cost per interaction. The metric the unit already reports is the one the journey is measured on. Nothing new is invented for the purpose. ### A unit has a score The Corporate Intelligence Capability (CIC) score, 0 to 10, says how intelligently the unit can act today: which instruments its data can reliably carry, from reporting through to agents that act. Raising it is the journey. You cannot have AI, or intelligence, without data availability. The CIC score is where that sentence becomes a number a unit can move. What CIC means ## How intelligently a unit can act Corporate Intelligence Capability (CIC) is the ability of an organization, or a unit inside it, to act intelligently: to make quality decisions, make them fast, and make many of them. Picture someone in a dark room with a flashlight. How well they can act in that room depends first on what the light reaches, and a bigger brain does not help with the corner it never lit. CIC is how intelligently a unit can act. The flashlight is its data. ### Quality decisions The decision process is sound: the right frame, real alternatives, reliable information, clear trade-offs, sound reasoning, and commitment. Limited by whether the data is true, complete, and understood. ### Fast decisions Made at the speed the world changes. A weekly report cannot run an intra-day pricing decision. Limited by whether the data is current when the decision is made. ### Many decisions Made at scale and repeatably, up to and including without a person in the loop. Limited by whether the data can be reached, is permitted for the purpose, and is in a form software can use. The CIC score, 0 to 10, is roughly how far down the list of instruments a unit can reliably go for the decisions that matter to it: say what happened, say why, say what is next, pick the best move, answer questions about it, and act on it. Each step needs a brighter, steadier, more trustworthy light than the one before. The journey ## Six steps, each one visible Forward-deployed engineers land Metis inside the unit. Every step ends with something the unit can see in the system, not a slide. - 01 Inventory the decisions The few decisions with the most at stake, written down with their owners, alternatives, objective, frequency, and the metric they move. What the unit sees Its decisions, named and owned. - 02 Score the data Each decision's data is scored on the ten availability dimensions against the instrument it needs, with evidence, and the weakest gap is named. What the unit sees An availability profile per decision, and the unit's first CIC score. - 03 Write the returns chain The investment, the CIC score to reach, the instruments that reaches, the metric it moves, and the baseline it moves from, agreed with finance before any work starts. Return on investment (ROI), written down first. What the unit sees A returns statement it can hold us to. - 04 Raise what is short Connect what exists, conform and govern it, and create the data that was never captured, inside the unit's own tools. What the unit sees The gaps closing, dimension by dimension. - 05 Run what the data can carry Reports, models, grounded answers, and agents, one decision at a time, each admitted when its data meets the bar for that instrument. What the unit sees Instruments in use, with every score carrying its confidence. - 06 Measure, re-score, hand over Every decision is logged with what was expected; outcomes are measured at the horizon; the profile is re-scored on a cadence; the unit's own teams run the loop. What the unit sees The CIC score moving, the metric moving, and the returns stated. By role ## What changes for the people in the unit The same journey lands differently depending on where you sit. ### Platform and AI leaders A measured ceiling before anything is funded, so the change effort goes to instruments that can work. Governance, access, deployment, and audit stay centrally owned; every team gets a governed path from question to production. ### Data and engineering teams Work orders with a named cause. A binding gate says which dimension is short for which decision and which discipline owns the fix, so pipelines, sources, and models are built for decisions, not for a backlog. ### Business and analytics teams Answers from live data in plain language, scoped to the access you already hold, with sources, freshness, and confidence shown, and with your decisions recorded so the next answer is better than the last. Where you start, where you go ## Four named tiers on the CIC scale Every unit starts somewhere on the CIC scale, 0 to 10, and the named tiers say plainly what is reliable at each score. The full scale is on the CIC scale page. ### Incapable, 0 to 2 No instrument is reliable. The work is capture and centralization, not AI. ### Foundational, 3 to 5 Reporting and diagnosis work. Prediction and generation will not be reliable until quality, meaning, and governance reach the model-ready profile. ### Analytical, 6 to 7 Predictive, prescriptive, and grounded generative instruments are trustworthy. Agent pilots still need heavy supervision. ### Autonomous, 8 to 10 Agents can act, under per-action permission and full provenance. A small minority of organizations are here today. ## Start with your decisions Bring the decisions that matter and the data you have. The first thing we do is score them; the second is show you what moves the score. Talk to sales Why it works --- # The scores · Metis URL: https://orama.com/journey/scores > Every step of the journey is measured: the availability profile on ten dimensions, the Corporate Intelligence Capability score per unit, the Decision Quality Score with velocity and scale, and calibration over many decisions. What each measures, what moves it, and how Metis raises it. The scores # Every step of the journey is measured Metis uses four measurements, and each one answers a different question: how available is the data, how capable is the unit, how sound are its decisions, and are its instruments telling the truth over time. Talk to sales See the CIC scale Diagram: The ten availability dimensions in three bands, in dependency order: reach, then fidelity, then fitness. A decision's data is only as available as its weakest dimension. Four measurements ## What is scored, and on what Two describe the ceiling, two describe what is realized under it. - 01 The availability profile Scored on One decision's data, against one instrument What it measures Whether the data a decision needs exists, can be reached and lawfully used, is true, current, and understood, and is fit for the instrument: ten dimensions in three bands. How it is scored Each dimension 0 to 10 with evidence, against the level the instrument requires; combined as a weakest link, never an average; the largest gap is the binding gate. What moves it Closing the binding gate: connecting a source, conforming it, agreeing a definition, capturing what was never recorded. - 02 The Corporate Intelligence Capability (CIC) score Scored on An organizational unit What it measures How intelligently the unit can act today: the highest instrument tier its priority decisions reliably support, from descriptive reporting to agents that act. How it is scored 0 to 10, anchored to typical availability profiles; the unit's score is the one that holds for the majority of its priority decisions, never an average across them. What moves it Raising the profiles of the decisions that matter until the next tier's requirements are met for most of them. - 03 The Decision Quality Score, with decision velocity and decision scale Scored on One decision, before it is committed What it measures Whether the decision process was sound: the right frame, real alternatives, reliable information, clear trade-offs, sound reasoning, and commitment to act. Decision velocity adds how fast; decision scale adds how many, how consistently. How it is scored Six links each 0 to 10, combined as the minimum; latency against the decision's required latency; throughput, coverage, and noise across the people who make it. What moves it The information link is bounded by availability, so it moves with the profile; the others move with how the decision is framed, argued, and adopted. - 04 Outcome calibration Scored on Many decisions of one kind What it measures Whether stated confidence matches what happened: when an instrument says a thing is likely, is it. How it is scored Over a population of logged decisions, the stated probabilities are compared with the outcomes at the horizon. A single outcome says almost nothing; a hundred say a great deal. What moves it Logging every decision with its expectation, measuring at the horizon, and reviewing gaps by cause. The ten dimensions ## Three questions the data has to pass The availability profile scores ten dimensions in three bands, in dependency order: reach, then fidelity, then fitness. ### Can the instrument get it, lawfully Existence: was it recorded at all, at the grain the decision needs. Accessibility: can the systems and people serving the decision reach it. Permission: may it be used for this purpose, at this grain. ### Can the instrument trust and understand it Quality: is it true to what it claims. Currency: is it current when the decision is made. Provenance: where it came from, who owns it, what happened to it. Meaning: does everyone, human and machine, compute the metric the same way. ### Does it serve this decision through this instrument Relevance: would knowing it change the choice. Sufficiency: enough history, enough labeled outcomes, enough of the rare cases. Shape: consumable without a preparation project. A dataset with perfect quality that the instrument may not lawfully use has an availability of zero for that decision. That is what weakest link means. How they relate ## The ceiling, and what is realized under it The four scores are not independent. Two set the ceiling; two show what the unit actually achieves under it. And higher quality decisions mean better outcomes and returns. ### The profile bounds the CIC score A unit cannot be model-ready if its priority decisions' data does not meet the model-ready profile. The CIC score summarizes; the profiles explain. Metis never shows one without the other. ### The profile bounds decision quality The information link of a decision can be no better than the availability of its data, and a decision is no better than its weakest link. So decision quality has a ceiling too, and it is the same ceiling. ### Outcome calibration shows the truth over time Whether the instruments the unit runs are actually right, at the confidence they claim, only shows across many decisions. Outcome calibration is the lagging measure that keeps the other three honest. Higher quality decisions mean better outcomes and returns. A decision that scores higher on its six links is, over many decisions, right more often at the horizon, and the returns are stated in what those outcomes do to the unit's own metric. Raise availability and decision quality rises with it; raise decision quality and the returns follow. How Metis raises them ## A gap names its owner When a decision's profile falls short of the instrument it needs, the binding gate says which discipline owns the fix. That is the whole coordination model: the score dispatches the work. Table: A gap names its owner | Binding gate | Who owns the fix | The usual remedy | | --- | --- | --- | | Existence | Application development | Build the application that creates the data: capture at the point of decision, a workflow application in place of a spreadsheet, an extractor, an experiment, a collector. | | Accessibility | Data engineering | Connect the source, or run over it where it sits. | | Permission | Decision analysis with legal | A purpose register, a consent flow, a contract, per-action policy for agents. | | Quality | Data engineering | Conformance, a contract with the producer, monitors on freshness, volume, and schema. | | Currency | Data engineering | Streaming or change capture; a freshness target set by the decision's clock. | | Provenance | Data engineering | A named owner, end-to-end lineage, a trust grade on the source. | | Meaning | Decision analysis, then data engineering | One agreed definition per metric, encoded in the semantic layer so people and machines compute it the same way. | | Relevance | Decision analysis | Re-derive what would actually change the choice; price it by the value of information; stop collecting what would not. | | Sufficiency | Data science with application development | Capture outcomes and labels in the flow of work; collect the rare cases; run experiments for the interventions history never tried. | | Shape | Data science with data engineering | Consumption-ready marts, features, indexes, and tool interfaces, so no instrument starts with a preparation project. | Every arrow is an object in the system. Nothing in the loop lives in a slide or a ticket outside Metis. ## Start with your decisions Bring the decisions that matter and the data you have. The first thing we do is score them; the second is show you what moves the score. Talk to sales Why it works --- # The CIC scale · Metis URL: https://orama.com/journey/levels > The Corporate Intelligence Capability (CIC) score, 0 to 10: the availability profile behind each score, which instruments it reliably supports, the four named tiers from incapable to autonomous, and why the step to acting is the steep one. The CIC scale # The Corporate Intelligence Capability score, 0 to 10 Corporate Intelligence Capability (CIC) is how intelligently a unit can act: quality decisions, made fast, and made at scale. The CIC score, 0 to 10, says which instruments the unit's data can reliably carry today, and it is built so that the score predicts what will actually work. Here is each score on the scale, the availability profile that typically sits behind it, and what it reliably supports. Talk to sales See the scores Diagram: The six instrument tiers, from Tier 1 Descriptive to Tier 6 Agentic: descriptive, diagnostic, predictive, prescriptive, generative, agentic. The demands on the data tighten at every step, and the failure of the last tier is an action, not a belief. Why raise it ## What a higher CIC score means for you Every score on the scale admits a kind of instrument the unit could not trust before. Raising the score is not collecting more data or buying more tools; it is widening what the unit can decide well, fast, and at scale. ### Decisions you could not make before At each score, decisions that used to run on memory and anecdote get a reliable report, then a diagnosis, then a forecast, then a recommendation, then an answer, then an agent. The unit's own metric moves with them. ### Spend that lands The score says which instruments will work today and which will not, so intelligence and AI investments go where the data can carry them, and the return on investment (ROI) is stated before the work rather than hoped for after it. ### A ceiling you can see Nothing done with the data can show what the data did not reach: bigger models and better agents do not lift the ceiling. The score makes the ceiling visible and names the gap that raises it. Two things the score is not. Not a list of tools: a forecasting platform whose forecasts nobody trusts is zero predictive capability. Not a property of the data alone: the same data can support a report and fail a model, so capability is always about a decision. The CIC scale ## One scale, 0 to 10, four named tiers Scores are cumulative in the dimensions that matter for acting: accessibility, permission, currency, provenance tighten as the scale rises. The profile column describes the priority decisions' data at that score. Table: One scale, 0 to 10, four named tiers | Score | Label | Typical availability profile | Reliably supports | Tier | | --- | --- | --- | --- | --- | | 0 | Non-existent | The decision's variables are not recorded anywhere. | Nothing beyond memory and anecdote. | Incapable | | 1 | Ad hoc | Data exists in spreadsheets, paper, and local files. | Manual reporting by individuals, not systems. | Incapable | | 2 | Fragmented | Several systems capture data; none are integrated; no shared definitions. | Departmental reports that disagree with each other. | Incapable | | 3 | Centralizing | Raw data lands in one place; quality and meaning are still low. | Descriptive reporting, with manual cleanup every cycle. | Foundational | | 4 | Standardized | Core entities are conformed and deduplicated; consistent definitions for the basics; owners known. | Reliable descriptive reporting; early diagnosis. | Foundational | | 5 | Governed | Consumption-ready data with lineage, ownership, quality rules, and a decision inventory behind it. | Diagnosis that holds across the whole unit. | Foundational | | 6 | Model-ready | Representative, labeled history at scale for the priority targets; features reusable; permission covers analytical use. | Predictive models that are trusted for forecasting and scoring. | Analytical | | 7 | Context-enriched | One governed definition per metric; passages indexed and permissioned for retrieval; decisions and outcomes logged. | Grounded generative answers; prescriptive recommendations where decisions are logged. | Analytical | | 8 | Real-time capable | Near-real-time freshness and unified cross-system access for the operational decisions. | Agent pilots in narrow, supervised workflows. | Autonomous | | 9 | Autonomous-ready | Per-action permission, action-level audit, governed memory. | Agents in production for defined workflows, with people on the exceptions. | Autonomous | | 10 | Self-optimizing | Every dimension monitored and self-healing; new sources onboarded, governed, and shaped within a release; investment ranked by decision value. | Multiple coordinating agents across units; the ceiling rises as data arrives. | Autonomous | The named tiers ## What to tell the executive The four named tiers compress the scale into the sentence a leader needs. ### Incapable, 0 to 2 No instrument tier is reliable. Spend on capture and centralization. Do not spend on AI. ### Foundational, 3 to 5 Descriptive and diagnostic instruments work. Predictive and generative will be unreliable until quality, meaning, and governance reach the model-ready profile. ### Analytical, 6 to 7 Predictive, prescriptive where decisions are logged, and grounded generative instruments are trustworthy. Agent pilots will fail or need heavy supervision. ### Autonomous, 8 to 10 Agents can act. Only a small minority of organizations are here, and most of them got here one decision at a time. The steep step ## Why 7 to 8 is the hard one Scores 0 to 7 raise the data's interpretability: it can be found, trusted, understood, and retrieved. Score 8 is where the data must become actionable in real time. ### Currency and access jump together An agent acts on the state of the world now, across every system it touches. Freshness and unified access both have to be near their maximum at once. ### Then permission and provenance follow From 8 to 9, an agent must be authorized per action and every action must be auditable. A wrong action is a different kind of failure from a wrong report. ### The same asymmetry everywhere Every published maturity model treats its top stage as rare and hard-won. This scale is explicit about where the difficulty is, and why. By function ## Where units usually get stuck The binding gates cluster by function. These are the usual ones, and the metrics the journey is measured on. Table: Where units usually get stuck | Function | Metric the unit already reports | Instruments that move it | Usual binding gates | | --- | --- | --- | --- | | Finance | Close-cycle days; forecast variance; audit findings | Close automation; cash forecasting | Meaning; provenance | | Sales | Forecast accuracy; win rate; cycle time | Propensity models; account research | Quality of the customer record; outcomes logged | | Service | Cost per interaction; first-contact resolution; satisfaction | Assistants over the knowledge base; ticket resolution agents | Currency and permission of the knowledge base | | Supply chain | Stock-out rate; inventory days; on-time delivery | Demand forecasting; order optimization | Currency; shape; sufficiency | | Marketing | Acquisition cost; conversion; retention | Churn prediction; spend allocation | Existence of attribution events; meaning | | People | Time to fill; attrition; engagement | Attrition prediction; policy assistants | Permission over personal data; meaning | | Operations | Downtime; yield; safety incidents | Predictive maintenance; dispatch agents | Existence of sensor data; currency | ## Start with your decisions Bring the decisions that matter and the data you have. The first thing we do is score them; the second is show you what moves the score. Talk to sales Why it works --- # The returns · Metis URL: https://orama.com/journey/returns > How raising a unit's CIC score turns into returns, return on investment (ROI) in the unit's own metric: the chain written before the work, what to expect for each kind of work, what separates good returns from great ones, and the measurement discipline that makes the claim defensible. The returns # Invested this much, raised the CIC score, moved the metric, returned this Returns are return on investment (ROI): fiscally responsible intelligence and AI investments, stated in the metric the unit already reports, with the whole chain written down before the work starts. Every clause of that sentence is a measurement set up in advance, which is why it can be defended afterwards. Talk to sales See the scores Diagram: The chain from data to returns: your data is scored for availability against a decision, an instrument runs only when the data can carry it, the decision is logged, and the outcome and the returns are measured. The chain ## Written before the work, defended after it The sentence a sponsor wants to say afterwards, and what each clause requires to be true. Table: Written before the work, defended after it | The clause | What it requires | | --- | --- | | We invested this much in the unit | A cost baseline that includes platform, people, and the run cost after go-live. | | It raised the unit's CIC score from here to there | A scored availability profile per priority decision, before and after, by the same rubric, with an independent scorer for anything tied to money. | | by closing these gaps for these decisions | The gap analysis, with each binding gate named and its owner. | | That admitted these instruments | The instruments in production and used; adoption measured as the share of decisions actually taken with them. | | which moved the unit's own metric by this much | A metric the unit already reports, a baseline, an agreed counterfactual, and a measurement window fixed at the start. | | worth this much a year, so the returns are this | A value per unit of the metric signed off by finance before the work, and the formula written down with its assumptions. | What to expect ## Bad, good, and great, by kind of work Returns differ by an order of magnitude depending on the kind of work and on whether the data was scored first. Stated as shapes, not figures: every number Orama publishes is a verified claim, and these are the patterns behind the numbers. Table: Bad, good, and great, by kind of work | Kind of work | Bad | Good | Great | | --- | --- | --- | --- | | Raising the data: connecting, conforming, governing | A lake-first build with no decision attached; nothing above it ever ships. | Instruments attached from the first wave; payback inside the planning horizon. | The platform work amortized across many decisions, each one cheaper than the last. | | Reporting and diagnosis | Reports built and ignored; no decision changes. | Cycle times fall; one definition per metric ends the reconciliation meeting. | Decision latency drops and shows in the unit's metric. | | Prediction | A model built and never used, or trusted and systematically wrong. | Positive, attributable returns on a decision the unit makes every week. | A decision with a high value of information and a redesigned process around the forecast. | | Grounded answers | A pilot with no result, the modal outcome for programs that skipped the scoring. | A production use case in a workflow, with adoption measured. | Embedded in a redesigned workflow, with an outcome the unit's leader defends. | | Agents that act | Cancelled, or an incident. | A narrow, supervised workflow with measured cycle-time and error reduction. | Agents operating for defined workflows with people on the exceptions, every action logged. | The bad column is not rare. It is the base rate for work started without scoring the data first. What makes it great ## Two things beyond the data The data sets the ceiling, and higher quality decisions mean better outcomes and returns. Two things decide how close a unit gets to the ceiling. ### A decision worth the information Some decisions have a high value of information: knowing more would change what you do, and the difference is large. Those are the ones to raise first. The scoring finds them. ### A process that uses the instrument An accurate forecast nobody acts on returns nothing. The great outcomes are the ones where the unit redesigned how it decides, and adoption was measured, not assumed. ### A journey that compounds Each closed gap lowers the cost of the next instrument on the same data, and every logged decision raises the next round's sufficiency. The second wave is cheaper than the first. The measurement discipline ## Seven rules that make the claim defensible These separate the minority of programs that can prove their returns from the majority that cannot. ### An instrumented baseline before any work The metric's value, variance, and trend for at least four quarters. ### A counterfactual agreed in advance Trend extrapolation, a holdout, or a phased rollout. Before and after alone is not attribution. ### A named executive owner Someone who defends the metric and the value per unit. ### A fixed measurement window Stated at kickoff, with the ramp assumption written down. ### Independent validation Of both the scores and the attribution, for any figure that goes to a board. ### Adoption measured, not assumed The share of decision volume actually taken with the instrument. ### Re-score and re-attribute on a cadence A level that stops moving while spend continues is the earliest warning that a wave is building platform without instruments. ## Start with your decisions Bring the decisions that matter and the data you have. The first thing we do is score them; the second is show you what moves the score. Talk to sales Why it works