AI for business leaders · Lesson 2 of 5

Choosing a first use case

The best first project is narrow, frequent, and measurable, not the most ambitious one.

Marksai · Free course · About 10 minutes

The idea

The first AI project sets expectations for every one after it. A first project that is narrow and succeeds earns the next one. A first project that is ambitious and stalls makes the whole subject harder to raise again.

So the goal of the first project is not the largest possible benefit. It is a real benefit you can demonstrate.

What a good first use case looks like

  • Narrow. One task, for one group of people. You can describe it in a sentence.
  • Frequent. It happens every day or every week, so the benefit adds up and you get feedback quickly.
  • Measurable. You can say what "good enough" means before you start.
  • Internal first. Mistakes in front of your own staff are cheap. Mistakes in front of customers are not.

Four questions

  1. Which task costs your team the most time each week?
  2. Do you have the data for it, and can you reach it?
  3. How would you know the result was good enough?
  4. Who would own the system once it runs?

If you can answer all four, you have a use case worth scoping. If you cannot answer the fourth, stop there: a system nobody owns will not survive launch.

Common mistakes

  • Choosing the idea that is most exciting in a meeting instead of the one that costs the most time.
  • Starting with something customers will see.
  • Starting without a measure, so that six months later nobody can say whether it worked.

A made-up example

An engineering firm is weighing two ideas. One is an assistant that writes client proposals. The other finds relevant documents from past projects. The first is more impressive. The second is narrower, used daily, easy to measure by time saved, and internal. The second is the better first project, and what it teaches the firm makes the first idea easier later.

Exercise · about 15 minutes

Take the task from lesson 1 that you sorted as the best fit. Answer the four questions for it in writing. Note which one was hardest to answer.

One thing to take to your team

Before discussing how to build it, agree on how you will know it worked.

For a shorter version of this lesson, see our checklist for a first AI project.