How do I know if my business needs AI?

Almost every page answering this question is written by somebody who sells AI, which means almost every page answers yes. Here is the version with the failing grades left in.

Four conditions, and all four have to hold

A process is a real candidate for AI when every one of these is true of it. Not most of them. All four.

1. It repeats. Somebody does it many times a week, in roughly the same shape each time. A task done twice a quarter cannot pay back the work of automating it, no matter how annoying it is.
2. It has a shape. There is an answer that a knowledgeable person would agree is correct, and you could explain to a new hire how to get there. If two experienced people on your team would reach different conclusions and both be defensible, you have a judgment call, not a process.
3. The volume is real. The hours add up to something you would notice on a payroll line. If eliminating it entirely would free up two hours a month, it is not the place to start.
4. The data is reachable. The information the process runs on already lives in a system, a file, a form, or an inbox. If the answer lives in one person's head, or in a paper folder, or in a system nobody can get an export out of, the first project is not AI. It is getting the data somewhere a system can read.

The fourth condition is the one that ends projects mid-build, and it is the one nobody checks before signing.

When the answer is no, or not yet

• The process is not written down and people disagree about what it is. Automating a process nobody has agreed on does not settle the disagreement. It buries it inside software, where it now runs faster and is far harder to inspect. Slow and visible is a better problem than fast and opaque. Write the process down first. Sometimes that alone is the fix, and you save the whole budget.
• The real bottleneck is a decision nobody has made. Quotes take three weeks because an approval sits with one person who is busy, not because anyone is typing slowly. No amount of automation upstream moves that. Find the queue before you buy anything.
• You are looking for AI because it is overdue, not because something hurts. This is the most common reason a project starts and the most reliable predictor that it will not be used after launch. If you cannot name the process, the people doing it, and roughly what it costs you, there is nothing to aim at yet. That is a fine place to be. It is not a place to spend from.

How to size it, using your numbers and not ours

Pick the single process you would most like to stop doing by hand. Then:

• Count the hours a week your team actually spends on it. Ask the people doing it, not the people who own it. The two answers are rarely the same, and the people doing it are right.
• Multiply by your own fully loaded hourly cost for those roles, and then by 52.
• Separately, write down what the errors cost when that process goes wrong: rework, a missed deadline, a customer who had to be made whole.

That annual figure is what you are buying against. If it is small, you have your answer, and it cost you an afternoon to get. If it is large, you now have the one thing that makes every subsequent vendor conversation useful: a specific process, a measured cost, and a number to compare a quote against.

What to do with the answer

If the four conditions hold and the number is meaningful, the next step is mapping the process properly before anything gets built, which is what a workflow audit is for. If they do not hold, the next step is the underlying process, and it is worth saying plainly that this is the cheaper outcome and often the better one.

One last test, and it is free. Ask any firm you talk to, including this one, to describe the case where they would tell you not to do this. A vendor who cannot answer that question has not evaluated your business. They have evaluated their pipeline.