AI Readiness Assessment: What It Covers and How to Run One

What an AI readiness assessment covers

Most frameworks group readiness into a handful of dimensions. The names vary; the substance does not. A workflow is only as ready as its weakest dimension, so an assessment should score each one separately rather than average them into a single maturity number.

• Business case: which workflows cost the most today, in hours, errors, delay or lost revenue, and what a realistic improvement is worth.
• Process: whether the workflow is stable and documented, how many exceptions it has, and whether it should be redesigned before it is automated.
• Data: whether the inputs the AI needs exist, are accessible, are of usable quality, and can lawfully be processed for this purpose.
• Systems and integration: which systems the workflow touches, whether they have APIs, and how results get written back without someone re-keying them.
• People: who owns the workflow, who reviews AI output, and whether the team has the time and skills to run it once it is live.
• Governance and risk: what the AI is allowed to decide on its own, where a human must approve, how decisions are logged, and which regulations apply.

A 12-question self-check

You can answer these for one workflow in under an hour. Every "no" or "don't know" is a finding: it is either a blocker to clear before a pilot or a question the assessment needs to answer.

• Can you name the one or two workflows where AI would save the most time or money?
• Do you know their monthly volume and how long each item takes to handle?
• Do you know the current error or rework rate?
• Is the workflow done the same way by everyone who does it?
• Are the inputs (documents, emails, records) available digitally?
• Could you pull a representative sample of 50 to 100 real cases this week?
• Do the systems involved have APIs, or a supported way to read and write data?
• Is there a named owner who can make decisions about the workflow?
• Is someone available to review AI output during a pilot?
• Have you decided which decisions AI may take alone and which need a person to approve?
• Do you know which data protection, sector or AI regulations apply to this data?
• Would you be able to measure the result against today's baseline?

How to read the result

Score each workflow, not the organisation. It is normal for one workflow to be ready for a pilot while another needs months of data or process work first.

• Ready for a pilot: you can answer yes to the volume, data, sample, owner and reviewer questions. Start a time-boxed pilot on real cases with a baseline to measure against.
• Ready after groundwork: the business case is clear but data access, integration or process consistency is missing. Fix the named blocker first; it is usually weeks of work, not a transformation programme.
• Not a candidate yet: volume is low, the process changes every month, or no one owns it. Automating it would lock in a process that should be redesigned, or save too little to matter.

What a good assessment delivers

• A ranked shortlist of use cases, each scored on value and feasibility.
• A measured baseline for the top candidates: volume, handling time, cost and error rate.
• The specific blockers for each use case, with who needs to clear them.
• A recommended approach and architecture, including where humans stay in the loop.
• A governance outline: approvals, audit logging and the policies AI must follow.
• A costed, time-boxed first step, so the decision to proceed is a decision about a known amount of money.

Common mistakes

• Assessing the whole organisation at once. Company-wide maturity scores rarely say what to do next; workflow-level findings do.
• Relying on self-reported surveys. People overestimate data quality and underestimate exceptions. Look at real cases.
• Starting with the model or vendor. Which AI to use is the last question, after the workflow, data and integration are understood.
• Leaving governance until after the pilot. Approval rules and audit requirements change the design, so they belong in the assessment.
• Skipping the baseline. Without today's cost and error rate, nobody can prove the pilot worked.

How New Odyssey runs one

There are two routes, depending on how much you already know. The Process Assessment takes one workflow you have already chosen: we map it, measure what it costs today, assess feasibility and risk, and finish with a fixed implementation proposal. The fee is fixed and credited against a Sprint if you go ahead.

The AI Integration Diagnostic is for organisations that have not yet chosen a workflow. Over one to two weeks it maps your systems, scores automation opportunities across them, and produces board-ready material with a recommended path. Where governance is the main concern, the plan includes how agents will be controlled and audited, which is what our CIP control plane is built for.

Frequently Asked Questions

How long does an AI readiness assessment take?

A single-workflow assessment can be done in days if the data is available. An assessment across several systems and teams usually takes one to two weeks. Anything much longer is usually trying to assess the whole organisation at once.

Is there a free AI readiness assessment?

Online quizzes are free and can be a useful prompt, but they score your answers rather than your data, so they cannot tell you whether a specific workflow is ready. The 12-question self-check on this page is free to use. To get a measured answer for your own workflow, someone has to look at real cases, systems and volumes.

Who should be involved?

The owner of the workflow, one or two people who do the work every day, someone who knows the systems and data involved, and whoever signs off on risk or compliance. Bringing in the people who do the work matters most, because they know where the exceptions are.

What is the difference between AI readiness and digital maturity?

Digital maturity describes how far an organisation has digitised overall. AI readiness is narrower and more practical: whether particular workflows have the data, integration, ownership and governance that AI needs to run on them in production. A digitally mature organisation can still have workflows that are not ready for AI.