Is your data ready for AI?

An AI tool is only as good as the information it works from. Point it at clean, consistent records and it can do genuinely useful work. Point it at a mess and it will produce a confident-sounding mess of its own.

The good news is that "ready for AI" does not mean perfect. Very few businesses have perfect data, and you do not need it. You need data that is good enough for the specific task you want to fix. This post gives you five questions to check that, and an honest look at what to do when the answer is no.

Start with the task, not the database

It is easy to ask "is our data ready?" as if it were one question about the whole business. It is not. It is a question about one task at a time.

Your customer records might be a tangle while your invoices are perfectly tidy. If the task you want to fix runs on invoices, the state of your customer records barely matters. So before you look at anything, pick the task. Then look only at the information that task depends on.

Question 1: Is it digital?

If the information lives on paper, in a binder or on a whiteboard, an AI tool cannot reach it. That does not end the conversation. Scanning and reading documents is one of the things AI does well. But it adds a step, and that step has a cost.

Check whether the information your task depends on is stored digitally. If most of it is, and only some arrives on paper, that is a common and workable starting point.

Question 2: Can you get it out?

Plenty of business software is happy to take your information in and much less helpful about letting it out. Before anything can be built, the information has to be reachable, either by an export or by a connection to the system it lives in.

A quick test: try to export a month of the relevant records to a spreadsheet. If you can do it in a few clicks, good. If nobody knows how, or the software does not allow it, that is worth knowing now. It does not always stop a project, but it shapes what the project costs.

Question 3: Is it reasonably consistent?

Open that export and scroll. Look for the things that trip up any automated process:

• The same customer entered three different ways.
• Dates written in several formats.
• Key fields left blank on some records and filled on others.
• Notes fields carrying information that should have its own column.

Some of this is normal. AI is often better than older software at coping with messy text. But the more inconsistent the records, the more checking a person has to do on the results, and the fewer hours the fix gives back.

Question 4: Is it in one place, or five?

Many tasks are slow precisely because the information is scattered. Part of it is in email, part in a shared drive, part in the accounting system and part in someone's head. Staff spend their time gathering it before they can use it.

That is both a warning and an opportunity. Scattered information makes a build harder. But pulling it together is often the very thing that saves the hours. Write down every place the task's information comes from. The length of that list tells you a lot about where the time goes.

Question 5: Does someone know who should see it?

Any fix that touches business information has to respect who inside the business is allowed to see what. If nobody can say who should have access to the records in question, sort that out first. It is a business decision, not a technical one, and a builder cannot make it for you.

What to do when the answer is no

A "no" to one of these questions is not a reason to give up on AI. It is a reason to adjust the plan.

Sometimes the fix is cleanup. A few hours spent standardizing names or filling in a missing field can make a later build cheaper and more reliable. That work is worth doing whether or not you ever build anything.

Sometimes the fix is a different task. If the information behind your first choice is in rough shape, the next task on your list may run on cleaner data and pay back sooner.

Sometimes the fix is the gathering itself. If the real problem is information spread across five places, the most valuable thing to build may be the step that brings it together.

The only real test

You can answer all five questions and still not be sure. The way to know for certain is to try the approach on a sample of your real information and see what comes back.

That is why our audit looks at your real data, not tidy samples. If your information is not ready, the report says so and lists what to tidy first. You learn that for the price of an audit, $2,500 over 5 business days, instead of halfway through a build.

Check yourself first

The free AI readiness quiz has a full section on data, alongside questions on your workflows, tools and payoff. Ten multiple-choice questions, under five minutes, scored out of 100. If you would rather talk it through, book thirty minutes with us.