Find your first useful AI project.

Name the work that gets stuck

Ask your team where time goes missing. It might be turning meeting notes into actions, finding information or drafting a customer response.

Write down what starts the task, what information it needs and where the finished work goes. Keep the first test narrow: drafting replies to one common enquiry is easier to assess than automating all customer service.

Check whether you can judge the result

Collect a few examples and agree on what a good result looks like. Include an awkward case as well as an easy one.

Who can spot a missing fact, a wrong assumption or an unsuitable tone? If nobody can reliably check the work, choose a more manageable starting point.

Count the whole effort

A fast draft is only part of the picture. Add up the time spent preparing information, creating the draft, checking it and fixing mistakes. Compare that with how the task works today.

Check quality as well as time. A small saving may not be worthwhile if checking becomes harder or mistakes cost more.

Use the simplest test that answers the question

Try an approved tool with appropriate sample data before commissioning a larger system. Anthropic’s guidance on building agents also recommends starting with a simple solution and increasing complexity when it improves the outcome. For your first test, the useful question is whether this specific piece of work gets better.

Give the trial an owner and a review date. At that review, make an explicit choice: continue, change the approach, or stop. Learning that a task is a poor fit is a useful result too.

Further reading: Anthropic: Building effective agents ↗