Cisco surveyed 2,511 chief executives across 23 countries in January 2026 and found that 91 per cent had become more optimistic about AI over the previous twelve months. In the same body of research, only 13 per cent of organisations were judged able to run AI safely, reliably and at scale. (Cisco AI Readiness Index)
The distance between those two figures is why the phrase AI readiness assessment exists. Intent is not the constraint. Very few leadership teams are still undecided about whether AI matters. What most of them lack is a reliable way to check whether the ground underneath that intent will hold weight.
This article sets out what an AI readiness assessment measures, how it differs from an AI maturity assessment, six questions that can be answered by observation rather than opinion, and how the established frameworks compare.
Quick Answer: An AI readiness assessment is a structured check of whether an organisation has the foundations in place to adopt AI successfully. It examines strategy, data, governance, skills, capacity and measurement before any system is built. It differs from an AI maturity assessment, which measures progress already made rather than whether the conditions to begin exist at all.
What an AI readiness assessment measures
An AI readiness assessment looks at conditions rather than outputs. It asks whether the things an AI deployment will depend on are already present, and where they are not, what would have to change first. The output is a diagnosis and an ordered list of gaps, not a score kept for its own sake.
Most published frameworks converge on a similar set of areas. Strategy, data, infrastructure, people and governance appear in almost all of them, with the differences sitting in how those areas are grouped and named rather than in what is examined.
That convergence is genuine and worth knowing. It is also only half useful on its own. A list of pillars tells you where to look. It does not tell you what a problem looks like once you are looking at it, which is the part most teams struggle with.
The scope also shifts depending on what is being adopted. Readiness for assistive tools that draft and summarise is a lower bar than readiness for systems that take actions on their own. The difference between agentic and generative AI changes what has to be in place beforehand, particularly around oversight.
AI readiness assessment vs AI maturity assessment
The two terms are often used interchangeably. They measure different things.
An AI maturity assessment measures how far an organisation has already travelled. It looks at what is deployed, how widely it is used and what it has returned, then places the organisation at a stage on a defined path. Our AI maturity model sets out how those stages progress. An AI readiness assessment measures whether the foundations exist to set off at all. It is run earlier, and it looks at conditions rather than achievements.
The distinction matters because the two problems can look identical from the outside and need opposite responses. BCG surveyed 1,250 senior executives across nine industries and assessed maturity across 41 foundational capabilities. It found that 5 per cent of companies were generating substantial value from AI, 35 per cent were scaling and beginning to see returns, and the remaining 60 per cent reported minimal gains and did not yet have the capabilities in place to scale. (BCG)
That last clause is the important one. Those organisations are not stalled part-way along a journey. Many are attempting the journey without the foundations, which is a readiness problem presenting as a maturity problem. Running a maturity assessment on them would produce an accurate low score and no useful instruction.
| AI readiness assessment | AI maturity assessment | |
| Question it answers | Do the foundations exist to start | How far have we already progressed |
| When it is run | Before or at the start of adoption | Once AI is in use across teams |
| What it examines | Conditions, capabilities and gaps | Deployment, adoption and outcomes |
| Typical output | A gap list and a sequence of fixes | A stage, grade or score with a trend |
| What it prompts | Remove blockers before building | Move to the next stage of adoption |
| Repeat frequency | Once, then after major change | Continuously or at set intervals |
Six questions that reveal whether you are ready
Pillar lists are a reasonable way to organise an assessment and a poor way to conduct one, because a pillar can be marked amber by anyone with an opinion. The six questions below are framed so the answer can be observed instead. Each has a version that is visibly true and a version that is visibly not.
They are not a replacement for the established frameworks. They are a faster way to find out whether running one is worth the effort yet.
Can you name the work you would hand over first?
Readiness begins with a specific candidate, not a category. Naming a department is not naming work. A useful answer identifies a process, who currently performs it, how often it runs and roughly how long it takes. That is the level of detail an AI agent needs before it can be built at all.
- Ready: a named process with a known owner, a known frequency and a rough time cost
- Not ready: a function or a department, with the actual task still to be identified
Can the data the system needs actually be reached?
Cisco found that 19 per cent of organisations have fully centralised data, against 76 per cent of the most AI-ready. The gap is rarely about whether data exists. It is about whether anything other than a person can find it.
- Ready: the data sits somewhere a system can query without a manual export
- Not ready: the data exists, and one person knows which folder
Does anyone own the decision?
In the same research, 58 per cent of organisations had a well-defined AI strategy, against 99 per cent of the most ready. Strategy here is less about a document than about authority. Someone has to be able to approve a deployment, and to stop one.
- Ready: a named individual can say yes or no and make it stick
- Not ready: approval is spread across several people, which usually means it rests with none
The stakes rise where systems act rather than advise. If you are assessing readiness for autonomous AI systems, ownership needs settling before anything is built.
Would you know if something went wrong?
Cisco reports that 24 per cent of organisations can control agent actions with proper guardrails and live monitoring, against 84 per cent of the most ready. Oversight is where readiness gaps stay hidden longest, because nothing appears to be wrong until it very publicly is.
- Ready: actions are logged, and a named person reviews the log on a schedule
- Not ready: the first signal that something had failed would come from a customer
Do the people who do the work have capacity to be involved?
Cisco found comprehensive change management plans at 35 per cent of organisations, against 91 per cent of the most ready. Readiness assessments tend to be scoped as technical exercises and then delivered by people who already have full workloads.
- Ready: time is formally allocated and the day job is backfilled
- Not ready: the work depends on goodwill and evenings
Could you prove it worked in six months?
The last question is the one most often skipped, because it has to be answered before anything changes rather than after. If no baseline is captured now, the before state will later be reconstructed from memory, and memory is generous.
- Ready: current time, cost or volume is measured and written down
- Not ready: the improvement will be estimated retrospectively
How established AI readiness frameworks compare
Several organisations publish readiness frameworks and most are free to use. They differ in depth and in what they are built to support, rather than in fundamentals.
Cisco’s index is the largest ongoing study of the subject and is most useful for benchmarking, since it reports how organisations compare globally as well as scoring an individual one. Microsoft’s assessment is the quickest route to a structured answer and takes around 45 minutes. TDWI’s model goes deeper, with roughly 75 questions producing a score per dimension. BCG’s work is research rather than a self-assessment tool, though its capability list is detailed enough to be read as one.
The overlap between them is substantial. Choosing between them is mostly a question of how much time you have, and whether you want a benchmark or a diagnosis.
| Framework | Structure | Format | Access |
| Cisco AI Readiness Index | Six pillars: strategy, infrastructure, data, governance, talent, culture | Annual global index plus a self-assessment tool | Free |
| Microsoft AI Readiness Assessment | Seven pillars: business strategy, AI governance and security, data foundations, AI strategy and experience, organisation and culture, infrastructure, model management | Self-service questionnaire, around 45 minutes | Free |
| TDWI AI Readiness Model | Five categories, roughly 75 questions | Interactive assessment with scores per dimension | Free, registration |
| BCG AI maturity research | 41 foundational capabilities across strategy, technology, people, innovation and outcomes | Published research based on an executive survey | Free to read |
| GrowthNation | Interview-led discovery across every team, scored on an AI maturity scorecard | Voice or text interviews, output includes a scorecard and agents | Waitlist |
How to run an AI readiness assessment
Whichever framework you use, the sequence matters more than the instrument.
Start by narrowing the scope. One team is a better starting point than an entire organisation. A narrow assessment produces a finding that can be acted on within a quarter, where an organisation-wide one tends to produce a document.
Gather evidence from the people doing the work, not from leadership alone. Leadership can describe the intended process accurately. Only the team knows the version that actually runs, including the workarounds that never made it into any documentation. This is where most readiness assessments quietly go wrong, because the assessed process and the real process are not the same thing. Some assessments address this by interviewing every team directly rather than running a leadership workshop, which is the approach GrowthNation takes.
Record the evidence behind each answer, not just the answer itself. A rating with no supporting observation cannot be revisited later, and readiness assessments are usually revisited.
Then sequence the gaps rather than listing them. Some block everything downstream and some are merely inconvenient. Fixing them in the order they were discovered, rather than the order in which they constrain, is the most common way a well-run assessment fails to produce a result.
What to do with the results
A readiness assessment produces a gap list, and a gap list only earns its cost if it changes the order of work.
Two responses are legitimate. The first is to fix the blocking gaps before building anything, which is the right call when ownership or oversight is missing. The second is to start narrow, on the single process where the foundations already hold, and use it to build the case for fixing the rest. That second route is often faster, because a working example makes the remaining gaps concrete.
What does not work is proceeding as though the assessment had not happened. If you are weighing up which processes to take first, our guide to AI agents for business covers how those candidates are usually chosen.
GrowthNation runs readiness as an interview rather than a workshop. It maps how each team actually works, turns the repeatable parts into agents the organisation owns centrally, and tracks the hours saved. You can see how the process works from first interview through to measurement.
Frequently asked questions
What is an AI readiness assessment?
An AI readiness assessment is a structured review of whether an organisation has the foundations required to adopt AI successfully. It examines strategy and ownership, data accessibility, infrastructure, skills, capacity and measurement, then reports where the gaps are and which of them block progress. It is run before or at the very start of adoption, and its purpose is to change what gets built first.
What is the difference between AI readiness and AI maturity?
Readiness asks whether the foundations exist to begin. Maturity asks how far you have already progressed. A readiness assessment is run early and looks at conditions, while a maturity assessment is run once AI is in use and looks at deployment, adoption and returns. Organisations that score poorly against an AI maturity model sometimes have a readiness problem rather than a progress problem, and that needs a different response.
How long does an AI readiness assessment take?
It depends far more on scope than on method. A self-service questionnaire such as Microsoft’s takes around 45 minutes and gives a structured starting point. A single-team assessment that gathers evidence from the people doing the work typically runs over one to three weeks. Organisation-wide assessments can take several months, which is one reason narrowing the scope is usually worth doing.
Who should be involved in an AI readiness assessment?
The people who perform the work, alongside whoever holds decision-making authority. Assessments built only from leadership input describe the intended process rather than the real one, and the difference between those two is often exactly where the readiness gap sits. A named decision-maker also needs to take part, since several readiness questions cannot be answered without one.
How often should you reassess AI readiness?
Once at the outset, and then after any material change to systems, structure or data. Readiness is not a continuous metric in the way maturity is. Once the foundations are in place, tracking shifts to measuring progress rather than rechecking conditions, unless something significant changes underneath.
