PXT · AI Consulting Built by engineers since 2007
All articles
Strategy 10 min read

Company AI readiness checklist: how to audit your AI stack and maturity

A scored company AI readiness checklist across five dimensions, how to audit your AI stack, and how to measure AI maturity before you spend on a build.

A company AI readiness checklist scores five things: your data, your infrastructure, your people, your use cases, and your governance. The point is to find out, before you spend, whether you are set up to get value from AI or whether you would be building on sand.

Most AI projects fail on execution rather than ideas, and the failures cluster at the same few points every time. Readiness is where you catch them while they are still cheap to fix. Below is a scored 15-question version you can run yourself in about an hour, plus how to audit the AI tools you already own and what to do with the result.

What AI readiness actually means

AI readiness is not about whether you have bought AI tools. It is about whether your organization can turn AI into results that stick.

A company can have a data science team, a budget and executive enthusiasm and still be unready, because the data is a mess, the systems do not integrate, or nobody owns the outcome. Readiness measures the conditions that let AI projects ship and survive contact with production.

It breaks into five areas, and weakness in any one can sink a project no matter how strong the others are. That last part is why a single overall score is close to useless. The pattern of scores is the signal.

The scored readiness checklist

Score each question 0, 1 or 2. Be harsh. The value of this exercise is entirely in whether you answer it honestly, and the most expensive mistake is giving yourself a 2 on data because the data exists somewhere.

#DimensionQuestion0 = not ready1 = partial2 = ready
1DataIs the data this project needs reachable?Locked in silos, no clean way inReachable with manual export or effortAvailable via API or a queryable store
2DataIs it clean and consistent enough to learn from?Gaps, contradictions, no standard formatUsable after known cleanupConsistent and reliable today
3DataDo you have enough history for the problem?Little or noneSome, thin in placesYears of relevant, labeled examples
4DataCould a new hire learn the task from your records alone?No, it lives in people’s headsMostly, with questionsYes, the records are the process
5InfrastructureCan your systems integrate with AI services?Closed boxes, no APIPartial or legacy APIsModern, documented APIs
6InfrastructureDo you have the environment for the planned workloads?No, and no planSome capacity, unclear fitProvisioned and understood
7InfrastructureIf your data is sensitive, can you deploy compliantly?Unknown or blockedPossible, unprovenPath is known and available
8PeopleIs there a named owner whose job improves if this works?No oneSomeone informallyNamed, accountable, resourced
9PeopleIs there enough literacy to evaluate a partner’s work?NoOne or two peopleEnough to make good calls
10PeopleWill the team change how they work?Active resistance expectedMixed, needs managingTeam is asking for it
11Use caseCan you name the specific workflow?”We should do AI”A general areaOne workflow, precisely
12Use caseCan you name the number you want to move?NoDirectionallyA metric with a baseline
13Use caseDo you know the cost of a wrong output?Never consideredRoughlyQuantified, with a review plan
14GovernanceIs security, privacy and compliance handled for AI use?No policyDraft or informalClear and applied
15GovernanceIs it clear who owns AI decisions and outputs?NoAmbiguousDocumented

Reading the result. Total out of 30, but read the dimensions before the total.

Any dimension averaging under 1 is a blocker: fix it before building, because it will otherwise surface halfway through a project at ten times the cost. A score of 20-plus with no dimension under 1 means you are ready to build something real. Between 12 and 20 usually means you are ready for one carefully chosen workflow, not a program. Under 12 means the first project is a data or process project, not an AI project, and treating it as one will save you a year.

Data is usually the lowest-scoring dimension and the most common blocker. Question 4 is the one that catches people: if a new hire could not learn the task from your records, an AI cannot either, because those records are the training material.

How to audit your AI stack

Auditing your AI stack means taking inventory of what you already run and how it fits together. Most companies accumulate AI tools without a plan and end up with sprawl they have never looked at in one place.

List every AI tool, subscription and integration in use across the company, including the ones individual teams adopted on their own. For each, note what it does, who uses it, what data it touches, what it costs, and whether it connects to anything else.

Then look for the three problems that show up in almost every stack.

Sprawl. A pile of disconnected tools that do not talk to each other, often with overlapping capabilities nobody has compared. Include your RPA estate in this inventory; brittle bots are frequently a sign of a workflow that needed AI rather than rules. The cost is rarely the subscriptions; it is that no single tool has enough context to do anything valuable.

Shadow AI. Tools teams adopted without security or compliance review, frequently touching customer or employee data. This is the one that turns into an incident. Ask specifically what data each tool sees, not just what it does.

Gaps. High-value workflows with no coverage while low-value ones have three overlapping tools. This pattern is extremely common and is usually a symptom of tools being adopted by whoever was most enthusiastic rather than wherever the hours were.

The audit turns “we use a lot of AI” into a clear picture of what is actually helping, what is a risk, and where the real opportunities sit.

How to measure AI maturity

Maturity models rate an organization from ad-hoc experimentation to AI embedded in core operations. The ladder looks like this.

LevelWhat it looks likeRealistic next step
1. Ad-hocIndividuals experimenting, nothing coordinated, no measurementPick one workflow and ship it properly
2. PilotingIsolated pilots, some working, none integratedTake one pilot all the way to production
3. IntegratedAI in specific workflows, delivering measured value, governance in placeExtend to adjacent workflows, build internal capability
4. EmbeddedAI in core operations and part of how the company competesContinuous improvement, own the roadmap

Knowing your level matters because it sets a realistic next step. A company at level 1 should not be trying to embed AI across the business. It should ship one integrated workflow, prove the value, and earn the mandate for the next one. Maturity models are most useful as a check on ambition.

What to do with the results

A readiness assessment is only useful if it changes what you do next. The output should be a short prioritized list: the two or three opportunities where readiness and business value are both high, plus the specific gaps to fix before anything else. Not an 80-page report. A decision.

This is what a focused strategy engagement produces. In a two-week sprint for a Series A fintech with board pressure to do AI and three stalled pilots, the audit surfaced roughly $250,000 of misdirected spend to stop and a $400,000 priority backlog worth building, for a $6,000 fixed fee. Three projects worth running, two pilots worth killing, and a roadmap that survived the next board meeting. The full engagement is here.

The value was not a maturity score. It was knowing what to build, what to kill, and in what order.

The four gaps that block most companies

Run enough of these and the same gaps show up. Knowing them in advance tells you where to look hardest.

Data access is the most common blocker by far. The data exists, but it is trapped in silos, locked behind systems with no clean way in, or scattered across formats that do not line up. Companies routinely think they have a modeling problem when they have a plumbing problem. The fix is unglamorous and it pays for itself across every AI project that follows.

No clear owner. AI efforts get funded and then orphaned, with nobody whose job improves when the project works. Without an owner, projects drift, stall, and quietly die at the first hard decision. Name one before anything is built.

Vague use cases. “We should use AI” is not a project. A workflow, an owner and a number to move is a project. Questions 11 to 13 above force this into the open, which is most of their value.

Governance left until later. For regulated data especially, an unresolved compliance question can block a working system from ever reaching production. Surfacing it during readiness rather than after the build is dramatically cheaper. Our guide to HIPAA-compliant AI covers what that looks like when the data is health information.

The encouraging part is that none of these require better AI. They require clean data access, a named owner, a specific target and an early compliance answer. Fix those and a company that scored badly becomes ready quickly, because the technology was never the thing standing in the way.

AI readiness for enterprises specifically

Larger organizations have the same five dimensions in harder versions. Data lives across more silos and more systems of record. Infrastructure includes legacy platforms that resist integration and cannot be replaced on a project timeline. Governance carries real regulatory weight rather than good intentions. And use-case clarity is harder precisely because more stakeholders want more things, so the prioritization step does more work than the scoring step.

For enterprises the checklist above is still the right frame, but run it per business unit rather than company-wide. An average across six divisions hides exactly the variation you need to see, and the right first project is almost always in the division that scores highest, not the one with the loudest sponsor.

Frequently asked questions

What are the essential steps in a company AI readiness checklist? Score data, infrastructure, people and skills, use-case clarity and governance from 0 to 2 across the 15 questions above. Audit the AI tools you already run for sprawl, shadow AI and gaps. Place yourself on the maturity ladder. Then turn it into two or three high-value, high-readiness opportunities plus the gaps to fix first.

Where can I find a detailed AI readiness checklist tailored for enterprises? The scored checklist above is the version you can run yourself, and the interactive readiness checklist on this site gives you a snapshot in a few minutes. For an enterprise-specific read, a strategy audit assesses your actual data, systems and workflows and returns a tailored, prioritized map instead of a generic score.

How do I evaluate my company’s infrastructure for AI readiness? Questions 5 to 7 above. Whether data is reachable through APIs rather than trapped in silos, whether your systems can integrate with AI services, whether you have the environment and security for the planned workloads, and whether you can deploy compliantly if the data is sensitive. Clean access and integration matter far more than raw compute.

What tools help measure organizational AI maturity and readiness? Interactive checklists and structured maturity frameworks give you a snapshot across data, talent, governance and adoption. A tailored strategy audit goes further, assessing your specific situation and returning a roadmap to close the gaps rather than a rating.

Should we fix the gaps first or start a project anyway? Both, in the right order. Fix any dimension scoring under 1, because it will block the project regardless. Everything else can be fixed alongside a carefully chosen first build, and often gets fixed faster with a real project pulling on it than as a standalone cleanup nobody is motivated to finish.


Run the 15 questions above in the next hour and you will know more about your position than most companies know before they sign a build contract. If you would rather have the scored version done for you, the interactive readiness checklist takes a few minutes and emails you the result, which also gives you something to argue about with your team.

If the result says you are ready and you want the prioritized version, that is what our two-week strategy sprint delivers, starting at $3,500. Sometimes the honest finding is that the first job is fixing data access rather than building AI, and we will tell you that before you spend.

Recognise your own problem in this?

Enough reading. Let's scope it.

Bring us the workflow. We'll tell you what it would take, what it would cost, and whether it's worth doing at all.

No deck · No demo · No sales pressure