Is my company ready for AI? Six checks before you spend

You are not ready for AI because you bought Copilot licences. You are ready when six things hold up under pressure: your data, your processes, your people, your technology, your strategy and your governance. Tooling is the last of those to matter and the first thing everyone buys.

The fastest honest answer is to ask yourself six questions, one per area, and refuse to grade generously. If you needed a clean, accurate list of your best 500 customers right now, how confident are you in what you would get? Could a new starter follow your core process without asking anyone? Would your team be capable of using an AI tool tomorrow? Can your key systems share data without someone exporting a spreadsheet? Does your AI strategy name a business outcome and a number? And who gets called when an AI tool causes a compliance failure? If any of those made you hesitate, that hesitation is your answer, and it is more useful than any vendor assessment.

Why the tooling question is the wrong one

RAND found that 80.3% of enterprise AI projects fail to deliver the business value promised: 33.8% abandoned before reaching production, 28.4% reaching production but failing to deliver, 18.1% running but never recovering the investment (aiassemblylines.com). You will have seen the MIT 95% figure quoted more often. Treat it carefully. That analysis was preliminary, not peer-reviewed, relied on self-reported data and used a roughly six-month window to judge success (MIT NANDA report). The RAND number is less dramatic and more defensible, and it still says four in five projects fail.

Notice what is missing from the failure modes. None of them are “the model was not good enough”. They are abandonment, non-delivery and unrecovered spend. Those are organisational failures wearing a technology costume. I have watched the same five patterns repeat for 25 years across ERP, CRM and now this.

So the readiness question is not which tool. It is whether the six things underneath the tool will hold.

1. Data: can you produce a clean list of your best 500 customers today?

Ask it exactly that way, because it is specific and it is checkable. Do not ask whether your data is “good”. Everyone says yes to that.

If the honest answer involves someone in finance pulling an export, someone in sales pulling another, and a third person reconciling the two in a spreadsheet by Thursday, you have your answer. An AI tool pointed at that data will produce confident, plausible, wrong output, and it will produce it faster than anyone can check it.

The second data question is harder and better: who in your business is responsible for the quality and accuracy of your data? Not who uses it. Who is accountable when it is wrong. In most companies between 50 and 10,000 people, that name does not exist. Data quality is owned by nobody and blamed on everybody. That is the cheapest thing on this list to fix and the one most reliably skipped, which is why dirty data is the cheapest AI mistake a business can address before spending anything.

The check: name the person accountable for data accuracy. If you cannot name one, you are not ready.

2. Process: could a new starter run it without asking anyone?

AI automates process. If the process is not documented, there is nothing to automate. You are not deploying AI, you are asking a language model to guess what your people do and then paying for the guess.

The practical test takes an afternoon. Give the same routine task to three people in the business and compare how they approach it. The gap between those three approaches is the gap between what you think your process is and what it actually is. Every step one person does and another skips is a step no automation can reliably replicate.

There is a related question worth asking at the same time: how aware are you of where the bottlenecks sit in your most important processes? Leaders usually have a theory. The theory is usually about a department they already dislike.

The check: pick your most important process. Could someone who joined on Monday run it on Tuesday from the documentation alone?

3. People: capability, attitude and whether you are using it yourself

There are three separate questions hiding in the people pillar, and leaders tend to answer only the first.

The first is attitude. How would you honestly describe your team’s attitude toward AI right now? Curious, anxious, resentful, indifferent, quietly already using it on personal accounts. Those lead to very different adoption outcomes and you cannot tell which you have from a townhall.

The second is capability. If you asked your team to use an AI tool in their day-to-day work tomorrow, how capable would they be? Not willing. Capable. These are not the same and conflating them is how pilots die in month three.

The third is the one that stings. How actively are you modelling AI adoption yourself, trying it, talking about it, normalising it? If the answer is that you sponsor the programme but do not use the tools, your team has already drawn the correct conclusion about how serious this is. Smaller teams without technical staff can still move here: there is plenty a small business can do first without a tech team.

The check: has the business made any formal investment in helping the team understand AI, or has it only bought licences?

4. Technology: count the spreadsheets between your two most important systems

This is the only pillar where the question sounds technical, and even here the honest version is mundane.

Can your key business systems share data with each other without someone manually exporting and importing spreadsheets? Ask how many manual exports sit between your two most important systems. Every one of them is a point where an AI tool loses context, inherits yesterday’s numbers, or quietly breaks when the person who maintains the macro leaves.

You do not need a modern stack to be ready. You need to know where the seams are. A business with three systems and two documented integrations is in better shape than one with fifteen systems and a shared drive full of files named final_v3_USE_THIS.

The check: how many manual exports and imports sit between your two most critical systems? Get the number, not an impression.

5. Strategy: is the document for guidance or for the board?

Here is the most uncomfortable statistic of the set. 75% of executives admit their company’s AI strategy is “more for show” than actual internal guidance (Writer, 2026). The same research found 48% describing adoption as a massive disappointment.

That is not a failure of strategy writing. It is a response to pressure. 62% of CEOs report their boards are actively applying pressure for measurable AI-driven outcomes (Dataiku/Harris Poll of 900 CEOs), and 61% say their boards are rushing it (Yahoo Finance). Meanwhile only 45% of surveyed finance executives could quantify the return on their AI investments, with median reported ROI at 10% against a 20% target (Lozen Advisory).

Set those side by side. The board wants outcomes. Finance cannot evidence them. So a document gets written that satisfies the first problem without solving the second. If you want to break that loop, you need a measurement position before you need a strategy document, which is the argument in how to prove AI ROI to your board.

The check: does your AI strategy name a specific business outcome and a number attached to it, or does it name technologies?

6. Governance: who gets named when something goes wrong?

Grant Thornton’s 2026 AI Impact Survey found 78% of business executives lack strong confidence they could pass an independent AI governance audit within 90 days (Grant Thornton). Not that they would fail. That they do not know.

Governance is not a policy PDF. The test is a single question: if an AI tool used in your business produced a decision that breached GDPR, or leaked client data, or discriminated in a hiring shortlist, whose name goes on the incident report? If the honest answer is a committee, or IT by default, or nobody, you have an ownership gap rather than a governance framework. It is also worth being clear that governance and compliance are not the same thing, and treating them as interchangeable is how companies end up with a compliance checklist and no actual control.

The check: name the single person accountable for AI outcomes. One name, not a function.

What to do with your six answers

You now have six answers, and the pattern matters more than the score. Three rough shapes come up repeatedly.

If data and process are weak but people and strategy are strong, you have an enthusiastic organisation about to automate chaos. Fix the foundations first. The enthusiasm will survive a quarter of groundwork; it will not survive a failed deployment.

If technology and data are strong but people and governance are weak, you are the company that buys well and adopts badly. The tools will be in place and unused, and somebody in the business will be pasting client data into a personal account because the sanctioned tool is awkward.

If everything is middling, that is normal and it is workable. Pick the weakest pillar, fix one thing in it, and do one use case properly rather than five badly.

What you should not do is conclude you are ready because five of six felt fine. Governance failures do not average out against good data.

The structured version of this check

Six questions you ask yourself in a meeting will give you a direction. They will not give you a comparable position, and they are vulnerable to the generosity you apply when you are the one marking.

The free AI readiness audit is the same check run properly: 30 questions, five on each of the six pillars, 7 minutes. You get a score out of 120, a band for each pillar and a short synopsis of where you stand. Every answer is self-reported, so it is only as honest as you are. No call is booked at the end unless you ask for one.

If the score comes back lower than you expected, that is the audit working. The companies in RAND’s 80.3% did not fail because they were unready. They failed because nobody checked.