An AI readiness audit checks whether your business can get real value from AI before you build anything. It looks at six things: where time goes, written knowledge, system connections, human decisions, data residency and ownership after launch. If knowledge and ownership are both missing, the honest answer is to wait.
Most companies we speak to have already tried an AI pilot. Most of those pilots stalled. Not because the technology failed, but because nobody checked whether the business was ready for it.
A pilot can look great in a demo and still go nowhere. The demo runs on tidy sample data, with a friendly audience and no real customers. Then it meets the actual business: policies stored in someone's head, an order system with no API, and nobody whose job it is to keep the thing improving.
That is what an AI readiness audit is for. It is two weeks of finding out where AI will pay off in your business, and where it will not, before you spend real money. Here is exactly what we look at.
The six questions.
- 01Where does time go? We count hours per task, not opinions.
- 02Is the knowledge written down? Policies, FAQs, price lists.
- 03Can systems talk? APIs for orders, CRM and bookings, or a plan to add them.
- 04Where must a person decide? Refunds, complaints, anything regulated.
- 05What does data residency require? Rules differ by country.
- 06Who owns it after launch? No owner, no improvement.
Where does time go?
Everyone has a theory about what eats their team's week. Theories are usually wrong, or at least incomplete. We talk to four to six people across service, marketing and operations, then look at the real numbers: ticket volumes, handling times, the reports someone builds by hand every Monday.
The goal is a short list of tasks that are frequent, repetitive and rule-based. Those are where AI earns its keep. A task that takes two hours once a quarter is rarely worth automating, however annoying it is.
Is the knowledge written down?
An AI system can only be as good as what it can read. If your returns policy lives in a PDF from three years ago, and the real rules live in your best agent's memory, the system will give confident answers that are out of date.
You do not need perfect data. Many quick wins need only your FAQs, policies and product catalogue. But the core knowledge has to exist in writing, and someone has to agree that it is correct.
Can systems talk?
An assistant that can only talk is a nice FAQ page. An assistant that can check an order, start a return or book a slot takes real work off your team. That difference comes down to integrations.
We map which systems hold what, whether they have usable APIs, and what is missing. If a key system is closed, that does not end the project. It changes the order you do things in, and it goes into the roadmap as a cost you can plan for.
Where must a person decide?
Every business has decisions that should stay with a person: refunds above a certain amount, complaints, anything regulated, any customer you cannot afford to lose. We write those down as explicit handoff rules before anything is built. It is easier to widen an AI system's limits over time than to win back trust after it oversteps. We go deeper on this in where we always keep a person.
What does data residency require?
Rules on where customer data can be stored and processed differ by country and by sector. A bank in the Gulf and a retailer in the Netherlands will get very different answers. We check what applies to you early, because it can decide which cloud region and which models you are able to use.
Who owns it after launch?
This is the question people skip, and it is the one that kills the most projects. An AI system is not finished at launch. Policies change, products change, and new edge cases appear every week. If nobody owns that work, quality slips and people stop trusting the system. We explain how that plays out in why most AI chatbots stall in week three.
How we score each idea.
By the end of the first week there is usually a long list of ideas. We score each one on value, effort and risk, so the decision does not depend on who argued loudest.
If question two and question six both come back empty, we tell you to wait. Building now would waste your budget.
Waiting is a real result. If the knowledge is not written down and nobody will own the system, the most useful thing we can give you is a list of what to fix first. You keep the report either way.
What you get at the end.
A 20-page report, a scored list of opportunities, and three quick wins, each with a clear scope. Most clients ship the first one within a month.
Common questions.
How long does an AI readiness audit take?
Ours takes two weeks from kick-off to report. Interviews happen in the first two days, the process and data review runs through day seven, and scoring finishes by day ten.
Do we need clean data before an audit?
No. Part of the audit is finding out which data is good enough today and what to fix first. Many quick wins need only your FAQs, policies and product catalogue.
What if the audit says we are not ready?
Then you get a clear list of what to fix and in what order, and you keep the report. Being told to wait is a useful result, because it saves the cost of a project that would have stalled.
You're on the list.
Your email app has opened. Press send to subscribe.