Decide what a worthwhile improvement would look like
What you'll learn
Choose an outcome and a simple measure before testing whether AI helps with the task.
Define the improvement before choosing a tool
AI can write a natural-sounding email, summarize a long document, or suggest many ideas in seconds. Those things can be useful. A business sees value only when the work is genuinely better after you count the time needed to prepare, check, correct, and handle exceptions.
For a customer-support team, that might mean faster replies that still follow policy. For a project manager, it might mean updates that miss less important information. For a sales team, it might mean better preparation so people can spend more time with customers.
The result is not always "faster." It may be fewer errors, less waiting, a smoother way for a new colleague to do the work, or a decision made with more complete information.
Choose one thing to improve
Use the task and current problem you wrote down in the previous lesson. Before choosing how AI might help, decide what a better result would mean to the person who uses it:
| Question | Example |
|---|---|
| What is not working well today? | The weekly report often misses the latest information. |
| What takes time or causes rework? | Each week, someone opens several places to gather the same information. |
| Who benefits if this improves? | The manager receives a clearer update and the team spends less time answering follow-up questions. |
| How will we know the trial helped? | Compare preparation time, corrections, or missing information. |
A useful measure does not need to be complicated. It needs to reflect what the person using the result actually cares about.
What I want to improve:
Who benefits:
Before/after measure and source:
Total preparation, review, correction and exceptions:
Condition for changing or stopping:
- Why is answer-generation speed not the only measure of value?
- For your task, what would show that the trial is good enough to continue?