Handling enquiries with AI without sounding like everyone else

When a small business first thinks about AI and customers, it reaches for a chatbot on the website. For most Midlands SMEs — particularly the ones selling considered, technical products — that is the least valuable option available and carries the most brand risk.

Why the chatbot is usually wrong first

Your website gets a modest number of visitors, and the ones who matter are often engineers checking whether you can make something. What they want is a specification, a capability list or a phone number — not a conversation with a bot that has read your marketing copy and cannot answer the one question they came with.

A bad chatbot is worse than none. It signals a firm that has bought technology it does not understand, which is a poor look when you are selling technical competence.

Where the real time goes

The expensive part of enquiry handling is almost never the first response. It is everything after:

• Working out what the customer actually needs from an ambiguous email
• Finding the last job you did like this
• Assembling a reply that covers lead time, tolerances, minimum quantities and terms
• Chasing the missing information you need before you can quote at all

Every one of those is internal work, invisible to the customer, and every one benefits from AI without any of it being customer-facing.

Three things worth doing

Triage on arrival. Classify each enquiry — genuine prospect, existing customer, supplier, recruitment spam — and route it. Most firms have someone doing this manually, badly, between other tasks, and real enquiries sit for a day because of it.

Draft the reply, do not send it. Prepare a response with the relevant technical detail and previous job references pulled in, ready for a human to adjust and send. This routinely takes first response from a day to under an hour, and responsiveness wins work.

Catch the missing information. An enquiry with no quantity, no material and no delivery date will cost two more emails before it can be priced. Flagging those gaps immediately, so the first reply asks for everything at once, removes days from the cycle.

If you do want something customer-facing

Make it narrow and honest. A tool that answers "can you machine this material to this tolerance" against your real capability list is useful. A general chatbot pretending to be a person is not.

Say plainly that it is automated — and note that this is no longer only a courtesy. Under the EU AI Act's transparency requirements, which took effect on 2 August 2026, people are entitled to know when they are interacting with an AI system rather than a person, and that reaches UK firms dealing with EU customers.

The thing worth protecting

In a market where your competitors are increasingly sending AI-drafted, identically-structured replies, a response that reads like a knowledgeable person who read the drawing is a genuine differentiator.

Use the technology to get to that reply faster. Do not use it to replace the part the customer values.