> Site index: [llms.txt](https://ortelian.com/llms.txt) with all pages and descriptions. # AI readiness assessment: a checklist for complex companies. Source: https://ortelian.com/ai-readiness-assessment/ Markdown: https://ortelian.com/ai-readiness-assessment.md An AI readiness assessment checklist based on how we audit work on site: the work, the owner, data and access, the standard, risk and who runs it after launch. An AI readiness assessment checks whether a company can put AI to work on a specific piece of work, and what has to change first. Most readiness frameworks score the whole company on maturity. We think readiness only means something for one piece of work, so this checklist starts there. It follows the two-week audit we run on site at the start of every engagement. You can work through it on your own first. [Talk to us](https://ortelian.com/contact/) [How we work](https://ortelian.com/how-we-work/) Readiness is per piece of work ## A maturity score won’t tell you where to start. AI maturity models rate a company on data, technology, skills and culture. That can help a board see the overall picture. It doesn’t tell you which work to change first, or whether that work is ready. A company with a low overall score can have one process that is ready for agents next month, and a company with a high one can have none. So ask the questions of one piece of work that matters: building the account list, preparing a planning round, checking invoices against orders. The sections below are what we look at. 01 · The work ## Can you describe the work and where it gets stuck? - One piece of work Can you name one piece of work that matters to results, with a clear output at the end? - How it runs Can the people who do it describe the inputs, decisions, handoffs and exceptions, including the parts nobody wrote down? - Where it is held up Do you know where the work waits, gets redone or depends on one person? - Judgement across sources Does it need judgement across several sources? If every step has a fixed input and a fixed output, it is a job for ordinary software, not agents. - Volume Does it happen often enough that doing it better changes something? 02 · The owner and the team ## Is someone able to decide, and can you get time with the team? - An owner Is there an internal owner with the authority to make decisions and drive implementation, ideally a founder or someone on the leadership team? - Time with the team Can the people who do the work spend time showing it and testing what gets built with them? - Appetite for change Do they want the work to change, or only to go faster? Agents pointed at the current process speed up the waste too. 03 · Data and access ## Do you know which sources the work depends on? - The real sources Which systems, files and inboxes does the work depend on, including the spreadsheets the org chart doesn’t mention? - Outside sources Which outside sources does it rely on, such as registries, partner lists or supplier portals? - One identity per thing When two systems disagree about a customer, supplier or order, is it clear which one is right? - Access Can access be agreed per system, with reading and writing separated? - Data handling Is it agreed where data may be hosted, which model providers may be used and what they may keep? 04 · The standard ## Can you say what good looks like? This is the section most companies skip, and the reason most pilots end without anyone able to say whether they worked. More in [agree the standard before you build](https://ortelian.com/notes/agree-the-standard-before-you-build/). - Real examples Can you point to real past cases and say what the right result was? - A baseline Do you know, or can you measure, quality, cost, time and human effort for the work today? - Stop-and-ask cases Do you know which cases need a person, where the right outcome for an agent is to stop and ask? 05 · Risk and control ## Do you know what must wait for a person? The controls are set out in [AI agent governance](https://ortelian.com/ai-agent-governance/). - What leaves the company Which steps send something outside, such as emails, invoices or orders, and should wait for approval? - Decisions about people Does any step rank, score or allocate people? Treat that as high-risk until a lawyer says otherwise. See what the EU AI Act asks of B2B companies, linked below. - A way to stop it Who can pause the work if something goes wrong? 06 · After launch ## Who keeps it running? - Who runs it Will your team run the work after launch, or do you want a partner to keep running it? - Keeping the standard How will you know in six months that it still meets the standard you agreed? Reading your answers ## What your answers mean, and what to do next. There is no score here, on purpose. The gaps matter more than the total. Option What it means What to do next Work, owner and standard are clear The work is ready for a first build. Agree access and start narrow: one job, in two-week cycles, measured against the standard. The work is clear, the sources aren’t Data and access are the first job. Map which sources the work really depends on and agree access before building. No owner Not ready, and no tool fixes it. Find the person who can decide how the work runs. Nobody can say what good looks like A pilot would end without an answer. Collect real past cases and agree the right result for each one. Every step is fixed input, fixed output This is an automation job. Use ordinary software or a workflow. Keep agents for work that needs judgement. The two-week audit ## What the audit does with the same questions. A checklist shows the gaps. The audit closes them on site, with the people who do the work, because the answers are rarely written down. It runs for about two weeks. 1. 01 Week one: observe and map We observe the work and talk it through with the people who do it, and read the records in your systems and your written procedures. That traces the inputs, decisions, handoffs and exceptions, and shows which systems and information people actually rely on. 2. 02 Week two: redesign and rank We find the constraint and redesign the process with your team around what AI can now do, deciding what belongs with agents, software or people. Then we rank the opportunities and check what the first build requires. 3. 03 What you leave with A map of the work, the first build and the standard it must meet, and a baseline to measure against. After the audit ## From the audit into the build. The audit is a fixed fee and is credited against the first build if you go ahead. Then we build, test and deploy in two-week cycles until the first job meets the standard. Full detail, including pricing, is on [how we work](https://ortelian.com/how-we-work/). Ortelian is a platform plus a forward-deployed team. We work on site not because our product needs it, but because your company does. After launch, your team runs the work or we keep running it for you. Either way, we host and operate the [platform](https://ortelian.com/platform/), in the EU, and the evaluations keep running. Your own agents can use the platform headless through scoped tools, including from [ChatGPT Enterprise](https://ortelian.com/ortelian-vs-chatgpt-enterprise/) or [Microsoft Copilot](https://ortelian.com/ortelian-vs-microsoft-copilot/) over MCP, set up inside the engagement. For what agents typically take on once the work is ready, see [AI agent examples by function](https://ortelian.com/ai-agent-examples/), [AI agents for sales](https://ortelian.com/ai-agents-for-sales/) and [AI agents for operations](https://ortelian.com/ai-agents-for-operations/). Deciding who should do it? Compare [building in-house](https://ortelian.com/ortelian-vs-building-in-house/) with [hiring an AI consultancy](https://ortelian.com/ortelian-vs-ai-consultancies/), or read what an [AI-native deployment partner](https://ortelian.com/ai-deployment-partner/) does. For anything that decides about people, read [what the EU AI Act asks of B2B companies](https://ortelian.com/notes/what-the-eu-ai-act-asks-of-b2b-companies/). Questions ## Questions people ask about AI readiness. **What is an AI readiness assessment?** An AI readiness assessment checks whether a company can put AI to work on a specific piece of work, and what has to change first. It looks at the work itself, the owner and the team, data and access, the standard the work must meet, risk and control, and who runs it after launch. **How is an AI readiness assessment different from an AI maturity model?** A maturity model scores the whole company on a scale. A readiness assessment, as we run it, looks at one piece of work and ends with what to build first, the standard it must meet and a baseline to measure against. **How long does an AI readiness assessment take?** Our audit takes about two weeks on site. You can work through the checklist on this page in an afternoon, but the answers that matter usually come from watching the work, not from a questionnaire. **What do we get at the end?** A map of the work, the first build and the standard it must meet, and a baseline to measure against. **Do we need clean data before we start?** No. You need agreed access to the sources one piece of work depends on. Messy data is normal. The audit shows which of it matters, and the world model we build keeps a source on every fact. --- ## Bring one piece of work. On a call, we work out whether an audit is worth it, who should be involved and how to start. [Talk to us](https://ortelian.com/contact/) [Or email Laurens](mailto:laurens@ortelian.com)