AI agents for service businesses: what they can do today, and what they can't

Every week someone asks me whether they should be using AI agents. My answer is almost always yes, with a condition: AI is only as good as the system underneath it.

What AI agents do well today

• Reply to new enquiries instantly by email, text or chat, in a tone that sounds like you.
• Ask qualifying questions and record the answers against the contact.
• Offer times and book calls straight into your calendar.
• Follow up with people who went quiet, without sounding like a robot.
• Draft emails, posts and review replies for a person to approve.
• Answer the same twenty questions you answer every week.

Where you still need a person

• Pricing conversations that need judgement.
• Complaints, and anything emotional.
• Strategy: who to target, what to offer and what to say.
• Checking that what the AI says is still true when your business changes.

Why the foundations matter

An AI agent plugged into a messy system just makes mess faster. If your pipeline stages don't reflect reality, your services and prices aren't written down and nobody knows which leads are hot, the agent has nothing solid to work from.

So we build the foundations first: a clean pipeline, clear rules, your services and pricing written down, and follow up that works without AI. Then the agents go on top, with guardrails. They never offer discounts, they hand complaints to a person, and nothing is posted publicly without approval.

Beyond the CRM

The same thinking applies outside your CRM. We connect Afoofa to Claude through MCP, so you can ask plain English questions about your pipeline, draft and send follow ups and update contacts from a chat. We also build Claude workflows for reporting, research and content that sit outside the platform.

Used properly, AI doesn't replace your team. It takes the repetitive work off them, so they can spend their time on the conversations that need a human.