AI will not fix a broken process.

Most AI projects fail not because the technology is wrong. They fail because the process they were automating was already broken before AI showed up.

The promise every team is buying.

The promise of AI is that it makes things faster. Customer service faster. Decisions faster. Content generation faster. Most leadership teams are buying that promise right now. Software vendors are happy to sell it.

If your team is overwhelmed, behind, or under-resourced, the pitch lands. You install the AI tool. You run the pilot. The first month looks good.

The reality most leaders find.

The second month, the team is frustrated.

The third month, you are paying for tools you are not using.

The pilot becomes shelfware. The decision becomes a question: was this the wrong technology, or did we use it wrong?

Neither answer is correct.

The pilot did not fail because the technology was wrong. The pilot did not fail because your team used it wrong. The pilot failed because the process underneath was not ready to be automated.

What a broken process actually means.

A process is broken when any of these are true.

The decisions in it are unclear. The same request produces different outcomes depending on who is on shift, what day of the week it is, or how the customer phrased it.

The ownership is fragmented. Different parts get done by different people with nobody tying it together. Handoffs are informal. Reviews are inconsistent.

The exceptions outnumber the rules. There is a documented path most cases are supposed to follow, but most of the actual work is in edge cases. People are constantly improvising.

The success criteria are vague. You cannot tell at the end whether the process worked. You can only tell that it finished.

If any of those describe a process in your business, putting AI on it gives you the same broken result, faster, and at scale.

What this looks like in real life.

Last month I sat with the leadership team at an insurance brokerage that had put an AI chatbot on their site before they had decided what their customer service process actually was.

Three months in, the chatbot was handling customers. Some well. Some poorly. Nobody on the team could tell you which path a particular customer would take when they arrived on the site — billing question, claim, new application, or just trying to talk to a person.

The bill went up. The customer experience went down. The brokerage had taken the ambiguity that had always existed in their service process and given it a meter.

The team had been working around the ambiguity manually for years. They knew which customer requests went to which person. They handled the edge cases on the fly. The system worked because humans were absorbing the ambiguity.

When the chatbot arrived, the humans came out of the loop, and the ambiguity went to the bill.

The technology was fine. The expectations were correct. The process was not designed.

Before you automate, simplify.

This is the part most leaders skip.

Before you put AI on any process, you should be able to write down four things in plain language.

What comes in. Specifically. Not “customer inquiries” — what kind of inquiry, where it arrives from, what the customer is trying to get done.

What the correct result looks like. Specifically. Not “resolution” — what the customer walks away with, what gets recorded, what gets sent back.

Who reviews the result before it is final. By name or role. With authority to override.

What happens when the process produces an exception. Where it goes. Who catches it. How fast.

If you cannot answer those four questions without hedging, the process is not ready for AI. The simpler version of your process should run cleanly with people before you ask AI to run it at scale.

When AI actually pays off.

AI is the right answer when the process is clear and you are bottlenecked on speed or volume. You know exactly what should happen. You just need it to happen more times per day, or in less time per case.

AI is the wrong answer when the process is unclear and you are bottlenecked on decisions. You do not know what should happen. Adding AI just means the wrong thing happens faster.

Most growing businesses think they have a speed problem. Most growing businesses actually have a decision problem.

Root cause first. Prescription second.

What to do next.

If you are about to put AI on a process — or you have put it on a process and it is not landing — sixty minutes is enough to look at the process underneath and tell you whether AI is the right next step.

Before you automate, simplify.