How to Know If You Are Ready for AI

Businesses ask whether their systems are modern enough, whether their data is clean enough, whether they need somebody technical first.

Those are rarely what decides it. The things that actually predict whether a project works are mostly organisational, and you can assess all six in an afternoon.

Why readiness matters more than capability

MIT’s Project NANDA published The GenAI Divide in July 2025, reviewing over 300 publicly disclosed AI initiatives. Against an estimated $30 to $40 billion of enterprise investment, roughly 95% of generative AI projects produced no measurable return.

Their study covers large organisations rather than businesses your size, and no measurable return frequently means nobody built a way to measure. What the figure establishes is that failure is normal and that it is not a technology problem, because the technology was the same for the 5% that worked.

The difference is in the conditions around the project.

The six

One process everybody describes the same way. If two experienced people describe it differently, you are not ready for that process yet. Settle it in a room first, which takes an hour and frequently delivers most of the value on its own.

A number you can state today. Elapsed time, hours, error rate, proportion handled within a day. Without a baseline the result will be a matter of opinion, and opinions do not survive a difficult year.

A named owner who is not you. The person accountable for the outcome before the system existed. Not IT, not a committee, and not whoever is most interested.

Willingness to switch the old route off. This is the one that separates the 5%. If both paths stay open, people use whichever is familiar under pressure, nothing is decommissioned, and nothing is ever measured.

The knowledge written down somewhere. The client who always needs a purchase order number, the supplier who says two weeks and means four. If it lives only in one person’s head, no system can use it.

Somebody with access to the systems. Unglamorous and the most common cause of a stalled first month. Sort the logins in week one.

The two most businesses fail on

Consistently, it is the fourth and the sixth.

Switching the old route off feels risky, so it gets deferred, and deferring it is how a working system quietly becomes an optional one. The decision is a management decision rather than a technical one, and nobody but the owner can make it.

Access is simply forgotten. Somebody has to be able to connect to the accounting package, the CRM, the job system. That request goes to a vendor, or an ex-employee’s account, and three weeks disappear before anything has been built.

Neither is difficult. Both are invisible until they are blocking.

What does not stop you

Worth listing, because these are the reasons businesses give for waiting and none of them hold.

Messy data. Handling messy input is the case AI is best at. Structured data is where conventional software already worked. If somebody tells you to clean everything first, be sceptical.

Old systems. Most of what matters can be read out of almost anything, including email and documents. A modern platform is convenient rather than necessary.

No technical staff. The decisions that determine success are about process and ownership. You need somebody with access, not somebody who can build.

Being small. The opposite. One owner, one process, one decision, and the distance between something working and it being how you operate is a conversation.

The seventh, which is not a checklist item

Somebody in the business has to want it to work.

Not enthusiasm about the technology, which is common and worth very little. Somebody with standing who is prepared to make the decisions the six require: to settle the process disagreement, to name the owner, to switch the old route off when people are nervous.

Projects with all six conditions and nobody behind them still fail, quietly, because every one of those decisions can be deferred and each deferral is individually reasonable. Projects missing one or two but with a determined owner usually get there.

If nobody in your business will take that role, the honest answer is to wait until somebody will.

Readiness expires

One last point, because the six conditions are not permanent once met.

The process everybody agreed on drifts as the business changes. The named owner leaves and nobody reassigns it. The baseline number stops being tracked because the person who cared about it moved on. The access credentials belong to somebody who has gone.

That is why a business can run one successful project and then find the second one much harder eighteen months later, and conclude the first was luck. It was not. The conditions decayed and nobody was watching them.

Check the six again before each new project. It takes an afternoon and it is the difference between a business that automates several things over three years and one that automated a single thing once.

The conversation to have first

Before any of this, one question, put to the person who does the work: what part of your week would you hand over tomorrow if you could.

The answer is almost always specific, almost always something repetitive that waits on them being available, and almost always fits one of the patterns that actually pays. It also tells you whether the team is worried about being replaced, which is worth knowing before rather than after.

Businesses that start from that question tend to pick correctly. Businesses that start from what AI could we use tend to buy something and look for a use for it.

If you fail two or more

Do not start yet, and do not conclude this is not for you.

Every one of the six is fixable in weeks rather than months, and all six are worth doing whether or not you automate anything. A process everybody agrees on, a number you track, a named owner and documented knowledge are simply how a business is run well.

Fix the two that are failing, then pick one daily process and build the narrow version. That is a quarter of work, and it is the sequence the 5% followed.

Sources

• Aditya Challapally, Chris Pease, Ramesh Raskar and Pradyumna Chari, The GenAI Divide: State of AI in Business 2025, MIT Project NANDA, July 2025.

AI Optimize runs the audit before the build, so you find out which of the six you are missing while it is still cheap to fix. Then we build the narrow version of one daily process, prove it holds, and widen it from there. That work sits under Workflow Automation.