1. One clear task
Describe the job in one sentence. For example:
- Answer customer questions about our services and opening hours.
- Help staff find answers in our HR and operations policies.
- Extract supplier, date and amount from incoming invoices.
2. Real examples
Collect a few dozen real questions, messages or documents that represent the task, including difficult and unusual ones. They show what the AI must handle and become the test set before launch.
3. The information it may use
List the approved sources: FAQs, policies, product data, price lists or internal guides. Note which are current, who updates them, and anything the AI must never use or reveal.
4. Systems and access
Note where it will run (website, mobile app or internal tools) and which systems it needs to read from or write to, such as a booking system or CRM. Check whether those systems offer an API or another way to connect.
5. Where people stay involved
Decide which answers or actions need a person to review them, how a conversation is handed over to your team, and who is responsible for following up.
6. How you will judge it
- The share of test questions it answers correctly.
- How often it hands conversations over, and whether it does so at the right time.
- Whether your team receives fewer repetitive questions or manual tasks.
- Feedback from the customers or staff who use it.
7. Data and privacy
Agree which data the AI may access, where it is processed, who can read conversations and how long they are kept, especially when personal or customer data is involved.
Next step
With the task and a few examples ready, you can discuss the project with Elite Solutions through the inquiry form. We review the task, the sources and the systems involved, and propose a first version you can test.