A useful first step with AI is choosing one task you can describe clearly. Drafting a reply to a common customer question is one example. A small experiment lets you check quality before changing the way the whole team works.
Find a recurring task
During a working day, note what you repeat: rewriting emails, organizing notes, or preparing a service description. Pick a task with understandable inputs, an easily checked result, and errors you can fix before a customer sees them.
For your first experiment, a draft in a separate document is enough. Keep sending messages and changing important records under human control.
Describe a good result
Write a short brief in plain language. State the outcome, audience, format, and facts the tool can use. Include an example of a reply you consider useful.
| What to define | Example for an email |
|---|---|
| Goal | Explain the customer’s next step |
| Input | An anonymized request and approved terms |
| Format | A short draft in English |
| Review | Accurate terms, tone, and contact details |
Compare two ways of working
Try the same task manually and with AI assistance. Record preparation, review, and correction time separately. A quick first draft is only part of the work; the ready-to-use reply is what matters.
Use several different examples, including an unusual request with missing information. If corrections take as much attention as writing manually, change the task or the approach. Keep the examples and your notes so a colleague can understand what you learned.
Keep the decision with a person
Before using the result, check facts, names, numbers, and promises. Decide who approves it and when the team returns to the manual process.
NIST offers a systematic approach to managing AI risk. Our practical inference for a first experiment is to define the task’s boundaries and responsibility for its outcome in advance.
Next step: choose one task and write down three signs that its result is ready to use.
