Summarizing inquiries is one way to test AI in a sales process. The task is small enough to check: read what a customer submitted, pull out the details and prepare questions for anything missing. A salesperson reviews it before contacting the customer.
1. Describe the task
Suppose a showroom receives long form submissions about kitchen projects. Ask the AI to extract the project type, location, timing and stated budget, then draft two questions about missing details.
Write one example yourself first. Ask the salesperson whether those details would help them prepare for a call. Measure how long the task takes by hand so you have something to compare.
2. Set the limits
Use only the information supplied for the task. Mark missing details as unknown. A vague phrase such as “mid-range kitchen” is not permission to invent a budget.
Keep the first version limited to preparing text. It should not send messages, promise prices, change customer records or book appointments. Name the person who will check each result.
- Use the customer’s stated details.
- Flag missing or contradictory information.
- Keep questions separate from the summary.
- Wait for a person to approve any reply.
3. Try incomplete requests
Test requests with no budget, conflicting dates, vague descriptions and locations outside your service area. Remove personal details that are unnecessary for the test. Decide what a correct response should contain before checking the AI’s answer.
Record omissions and invented details. Count the time spent reviewing and correcting the result, too.
4. Check whether it saves work
Run a short trial with one salesperson. Compare the total time with doing it manually. If summaries help but the suggested questions need constant rewriting, keep the summaries and drop the questions. Stop the trial if errors are reaching customers.
Try this
Choose five past inquiries, remove unnecessary personal details and write the summary you would want for each. Use those to check the first version.