How to pick your first job for AI

A good first job for AI happens often, uses facts you can trust, has a person check the result, and has one result you can count.

Start with work your team does a lot

Pick one task that happens often. Give it a clear start and end. This makes it easier to test if AI helps.

Do not start with a goal like “Use AI in the whole company.” Start with one task your team can show and tell.

  1. What task happens a lot?
  2. Where does it start and end?
  3. What should get better?

Check the facts and who may use them

List the facts the task needs and where they come from. The right people must be able to use the facts. The facts must be good enough for the job. You must follow your privacy and security rules.

Fix missing, old, or hard-to-find facts before you add AI.

  1. Can the team get the facts?
  2. Are the facts full and up to date?
  3. What facts must stay private or have limits?

Have a person check the work

Name the person who checks the result. A person must be in charge when the work can affect a customer, a promise, a payment, a rule, or a big business choice.

Plan what to do when facts are missing or the AI is not sure. A safe first use needs a clear path for these cases.

Pick one result you can count

Pick one change you can see. You may reply faster, type less, send work sooner, see where work stands, or spend less time getting facts ready.

Write down what happens now. Then check the first test. Keep the test small until you can see that it helps.