I keep asking this question to a lot of folks and I rarely get a straight answer. Sometimes folks even pull up their old slides to show the projection numbers on it.
Here is the pattern that I see a lot.
You show something with AI and the demo works and leadership is excited. But nobody wrote down what number you were trying to move. Six months later when someone asks if it worked, you have no answer because there was nothing set up to measure.
Let me explain that with something I have learnt over the years.
Last year I worked on a medical equipment marketplace where the job was connecting hospitals with suppliers, and every match was taking four to eight months. They wanted to use AI to speed it up.
Before writing any code, I spent a week watching how the team worked and talking to their users. In those conversations, someone will start explaining how things work and then they tell you what it costs them. A deal that did not close because they could not find a supplier fast enough. A supplier who went with a different company because the response took too long. A quarter that ended up slower than expected.
The problem kept coming back to matching. A buyer would send in what they needed and someone would go through the supplier list by hand, check location, check if they had the right licenses, check if they were available, look at past orders, then put together a shortlist. Same steps, done from scratch every time.
We built something to handle that matching, and what used to take weeks now takes a couple of days.
Before we wrote any code, we sat down and picked the numbers we were trying to move. Time to first match. Deals moving forward within the first month. Money closing in the same quarter instead of getting pushed out. We wrote them down. What they were today. What they should be after. Who was in charge of tracking each one.
A few months after we shipped, time to first match went from three weeks to under two days. Deals moving forward in the first month went from 30 percent to 60 percent. Money that used to slip to the next quarter started closing in the same quarter. We had those numbers because we decided what to track before we started.
If you want to know whether your AI project worked, you need to decide what number you are trying to move before you start building. Not what problem you are solving or what the AI does, but what number changes, what it looks like today, and what it should look like after.
Most teams skip this part completely. Not because they do not care, but because they do not know where to start. Or something more urgent comes up and they move on. Six months later when leadership asks if it worked, there is no answer.
Always ask these questions before you start.
What number are you trying to move? Do you know what it is today? Have you written down what it should be after? Who is tracking it?
Top comments (0)