An AI readiness assessment is a decision instrument for a growing business. It should show where to start, what to fix first and which ideas should wait. I use four layers, intent, foundations, enablement and execution, because a strong score in one layer cannot compensate for a missing condition underneath it.
Foundations before scale
Readiness is not a permanent label; it changes as the business, data and decisions change. That is the question I would put in front of a leadership team before approving an AI investment.
Bring one AI decision and the people responsible for it into the room. Set out the facts, name the constraint and agree the evidence needed to choose a path. Then test the effect on managers, customers, data, culture and the board before scoring maturity.
An AI readiness assessment should start with a live workflow, not a maturity questionnaire. The first pass traces where work slows, where judgement is repeated and where information is re-entered. That map reveals whether the first opportunity is automation, better data, a clearer decision or a different role boundary.
The intent layer tests whether the leadership team can name the business constraint AI is meant to improve. Foundations then examine data access, process quality and the technical environment. A business can be excited about a tool while still lacking the conditions for a safe, useful experiment.
Enablement is where many assessments become too abstract. I look for a named process owner, a person who can challenge an output, and a short explanation that a new starter could follow. If those pieces are missing, training alone will not create confidence.
Execution requires a review rhythm that can change course. The owner should know what evidence moves a pilot forward, what evidence stops it and which risk requires a different control. That makes the plan a sequence of decisions rather than a catalogue of vendors.
The most valuable result may be a deliberate “not yet”. A narrow use case with clean evidence is often a better starting point than a broad programme that exposes the business to avoidable risk.
Readiness is earned through improved foundations and repeated judgement. The result is a current decision, not a permanent score attached to the organisation.
Watch a live workflow before assigning a readiness score. Name who owns it and capture one piece of evidence from the next cycle. The useful observation is a changed decision or handoff that another leader can inspect without being briefed first.
The constraint to name early: What Is an AI Readiness Assessment
The book includes a technology investment where leaders could not agree between building internal capability and adopting a commercial platform. The due-diligence work became useful only after the team agreed design principles around integration, customer experience and future change. My working method is that sequence in readiness work now: define the conditions first, then discuss tools.
- Intent: Name the business constraint before choosing a tool.
- Foundations: Check data, workflow and access conditions.
- Enablement: Give people boundaries, confidence and a route to challenge outputs.
- Execution: Set an owner, evidence threshold and stop rule.
Readiness is visible in what happens when the first use case is inconvenient. Does the data owner answer quickly? Can a manager challenge an output without being labelled resistant? Is there a route for reporting a harmful result? These are not side questions. They show whether the organisation can learn safely. I would run a short rehearsal before a wider launch: use a realistic workflow, invite challenge and record the points that need policy or process changes. The rehearsal produces a more honest decision than a confident slide about future potential.
What the assessment should change this month
A useful AI readiness assessment ends with a decision that someone can see in the next four weeks. That might be a change to the claims-handling queue, a new rule for checking customer correspondence, or a decision to leave a proposed use case alone. The subject matters less than the test. The owner should be able to point to the old route, the new route and the evidence that will decide whether the change stays.
Start with one workflow that already consumes leadership attention. Ask the person who does the work to draw it from memory, then compare that sketch with the records in the system. The gaps are useful. They show where a handoff is invisible, where a policy lives only in one person’s head and where an AI suggestion would arrive without enough context. I would rather find three small breaks in a live process than receive a confident score from a questionnaire.
The assessment also needs a stopping conversation. A pilot can produce a plausible answer and still be unsafe to scale. Set the boundary before the first test: which decisions remain with a named person, which data cannot be entered, and what evidence would make the team stop. This is a management decision, not a software setting. It belongs in the same review where the team discusses service levels, cash and customer risk.
For a growing business, readiness is often a question of sequence. The organisation may have enough clean information for one narrow task while lacking the role clarity for a larger programme. That is not a failure of ambition. It is a useful order of work. Fix the owner, make the evidence visible and run the smallest test that can change the next decision.
I would leave the assessment with four lines in the operating plan: the workflow being tested, the person accountable, the evidence due and the date of the stop or scale decision. If those lines cannot be written plainly, the organisation is not ready to spend more on tools. It is ready to make the operating problem visible.
The board paper should make the assessment falsifiable. Write down the expected change in plain terms, such as fewer manual checks in a named queue or a shorter response time for a defined customer group. Record the baseline and the person who owns the measure. Avoid a target that can be met by moving work somewhere else. If the queue falls but complaints rise, the test has not worked. That simple comparison keeps enthusiasm from becoming the only evidence.
There is a human question as well. Who will carry the awkward cases when the first answer is wrong? Give that person a route to report an error, a way to pause the trial and enough context to explain the decision to a colleague. A readiness assessment that ignores this work leaves the risk with the most junior person in the process. That is a design choice, even when nobody wrote it down. Put the responsibility in the operating rhythm before the tool arrives.
The final page of the assessment should state what the team will learn, not what it hopes to prove. Name the first review date, the person who can stop the test and the record that will be checked. That small discipline makes the work legible to a board and fair to the people doing it. It also leaves room to say no when the evidence is thin, which is often the most responsible decision a growing business can make.
The assessment becomes useful when it changes the order of work. A growing business may have strong executive enthusiasm but weak data stewardship, or a promising use case but no time for frontline practice. Those findings are not a grade. They are a sequence. Start with the condition that would make the next decision safer and more valuable, assign an owner, then review the evidence before adding another use case.
The final question is whether the organisation can make a better decision because of the assessment. If the answer is no, the instrument is measuring interest rather than readiness. A useful result gives leaders a shared view of the next constraint, a safe experiment and a reason to revisit the decision. It leaves the business more capable of choosing, even before an AI system is introduced.
A growing company should repeat the assessment when the business changes shape. A new market, a new data boundary, a merger or a material change in customer expectation can alter the sequence even when the technology has not changed. Treat the result as a conversation about conditions and choices. That keeps the assessment connected to strategy rather than turning it into an annual compliance exercise.
The assessment should also make room for different levels of readiness inside the same business. A customer-service workflow may be ready for a bounded experiment while a high-consequence decision needs stronger controls and more human review. Treating the organisation as one score hides that distinction. I prefer a small map of use cases, conditions and owners, with a clear reason for the order. Leaders can then invest in the foundation that makes the next safe decision possible, instead of waiting for every part of the business to reach the same level at once.
The assessment earns its place when it makes the next decision safer and more deliberate, not when it produces the most impressive score.
The practical output is a sequence that a growing team can discuss and revise. It should identify the decision to make now, the condition that must be strengthened first, the person accountable for the work and the evidence that will be reviewed. That structure lets leaders move with care without waiting for a perfect picture. It also makes it easier to explain to colleagues why one promising idea is being supported while another is deliberately held back.
Turn readiness into a sequence
An assessment should change the order of decisions. Compare the result with why digital transformation programmes fail, three actions for digital transformation and measuring leadership capability across an organisation to decide which foundation, use case or capability needs attention first.
- Intent: Clarify the business decision and the value at stake.
- Foundations: Check data, ownership, risk and the time available to learn.
- Execution: Run a bounded test and review what changed in the work.
Three useful comparisons
Three related decisions sit close to this one and are worth reading before you act on it. The human and AI decision ladder is the sharpest comparison, and leading organisational transformation successfully shows what happens once readiness turns into a live programme. Neither piece replaces the judgement this assessment demands; they simply keep the question specific, which is what stops a readiness score from drifting into a generic maturity checklist.
Readiness becomes useful when it changes the order of decisions. A growing business might be ready to test one narrow workflow but not ready to connect a model to sensitive data or a customer promise. Record that distinction plainly. The assessment is doing its job when it gives the owner a safe next test and a reason to leave a larger ambition alone for now.
The review should include the people who carry the workflow, not only the person sponsoring the idea. They know where the hand-offs fail, which exceptions are common and what a safe human check would look like. Their evidence can lower the scope of the first experiment. That is not a lack of ambition. It is how a business learns without making the whole operation carry an untested bet.