What an AI Employee needs to know about your business

You ask AI to review your website. It gives you a polished list of suggestions: sharpen the headline, add proof, make the next step clearer.

Some suggestions may be useful. But which ones address the questions your customers actually ask? Which reflect your offer? Which deserve your team's attention first?

An AI Employee needs enough business context to answer those questions. Its instructions should connect the assignment to your customers, your standards, and the decisions the work is meant to support.

Explain the business in plain language

Begin with what you sell, who buys it, and what they are trying to accomplish. Describe what is included, what is outside the offer, and what a customer needs to know before taking the next step.

Be concrete. “We help companies transform” gives an AI role little to work with. A description of the customer, the service, and the result gives it a basis for evaluating whether your website explains the business clearly.

You do not need a perfect brand document before you begin. You need an accurate account of the business as it operates today.

Supply evidence from customers

Your description of the business is one input. Customer evidence is another.

Reviews, sales-call notes, recurring questions, and objections can show what buyers notice, value, or struggle to understand. Use relevant material you have permission to share, with unnecessary personal or confidential information removed.

Keep the source attached to the observation. A customer asking about implementation time is evidence of that customer's concern. It is not proof that every buyer has the same concern.

Ask the role to distinguish direct evidence from interpretation and to tell you when the information is too thin for a confident conclusion. That makes its findings easier to inspect and act on.

Show what a useful result looks like

If you want website findings, define the output. For example, you might ask each finding to include:

• The page or passage that needs attention.
• The customer question or business issue involved.
• The evidence behind the observation.
• A specific change to consider.
• Any uncertainty a person needs to resolve.

This is an illustrative output structure. The important part is giving the role a standard that serves the next person in the process.

“Improve our website” is difficult to review. “Identify the three most important gaps between customer questions and what our offer page explains” creates a focused assignment.

Set the limits of the role

An AI Employee reviewing your site should not invent testimonials, rewrite your commercial terms, or treat an unverified inference as a customer fact.

Tell it which sources govern the work, what it can recommend, and which decisions need approval. If two documents disagree about the offer, it should surface the conflict rather than quietly choose one.

Keep business information current. An outdated price or retired service can undermine an otherwise useful result.

Make the handoff explicit

In The AI Employee Lab, the first supplied Employee analyzes customer evidence. You then use those findings with the second Employee to assess your website and offer.

That connection matters: the website assessment has customer information to work from. You can trace a recommendation back to a question, objection, or need, then decide what to change.

For your own first role, assemble a small starting package: an accurate business description, relevant source material, an example of a useful result, and clear approval boundaries. Test it on one assignment and see what information is missing.

You are giving the role what it needs to do a particular job well. Add context because it improves that job, not simply because another document is available.

Build with your own business information

Inside The AI Employee Lab, I’ll guide you through activating two supplied AI Employees in Claude. Use your customer evidence to generate signals, assess your website, and identify three practical actions.