Using AI to prepare enquiries and quotes in a UK business

In short

A contractor uses AI to organise an incomplete request and expose missing details. A person chooses the follow-up questions, calculates the price and checks every commitment before sending. AI remains a drafting aid rather than the authority behind a quotation.

Practical AI for enquiries, quote preparation and admin

Imagine ending a busy day with three enquiries, scattered notes and a quote still to prepare. AI can help create a first organised draft, but the business owner still checks facts, makes choices and sends the final message.

Imagine a contractor reading every enquiry by hand. A safe first use is a draft containing the request, location, missing information and suggested follow-up questions; it is not an automatically sent quotation.

The contractor chooses one real type of enquiry as the starting point. It contains an address, a short description and two photographs, but no dimensions, material choice or preferred work period. That is not enough for a reliable price. The first AI task is therefore not “write a quotation”; it is “organise what is known and mark what the customer still needs to confirm”.

Start with the missing information

Pick one recurring task: summarising an enquiry, preparing intake questions, drafting a quote outline, or turning notes into a task list. Provide only information needed for that task and remove sensitive detail where possible.

Make a list of facts a person always confirms: prices, timing, terms, personal data and commitments. That list belongs in the workflow, not in a footnote.

The input uses consistent fields: type of work, location, access, available images, desired timing and special circumstances. A missing field stays visibly missing. The draft may suggest follow-up questions, but it must not invent a measurement, material, price or date. This gives the owner a readable preparation document without treating fluent prose as a complete project brief.

Turn a request into useful questions

Use a fixed input and a clear output. For example: the request, requested date, open questions and next step. Ask AI to produce a draft that flags missing information. That is more useful than a broad instruction to automate everything.

Confirm the cost of each explicit AI action before selecting a workflow. A useful trial describes the input, draft, review and stop point.

The owner compares the draft with the original message. Names and addresses are corrected, irrelevant personal details are removed, and only useful follow-up questions survive. A question about foundations might sound professional but be irrelevant to this job. Human trade knowledge is needed not just to polish the sentences, but to decide whether the conversation is moving towards the right piece of work.

Facts do not come from the draft

Review every draft for prices, promises, names, dates and consent. AI may add an incorrect detail or miss context. It is not an automatic authority to send messages, quotations or administrative decisions.

A quotation outline can follow only after the customer replies. It remains a draft. Prices come from the contractor’s calculation, terms from approved documents and availability from the actual schedule; AI is not the source of those facts. Before sending, an authorised person checks scope, quantities, tax treatment, validity, exclusions and any wording that a customer could reasonably read as a commitment.

Trial one narrow workflow

A suitable AI implementation is assessed and scoped separately; it is not an automatic website-plan entitlement. Trial one narrow workflow first, then decide whether review, data handling and integrations are adequately controlled.

During a limited trial, the contractor observes where the draft genuinely helps. Are missing fields spotted sooner? Do the follow-up questions produce useful answers? Is staff re-entering the same information? Sensitive data is not shared more widely than the task requires. Automatic sending stays off, and mailbox, CRM or estimating integrations wait until access, data use and failure handling have been assessed separately.

A person decides what gets sent

The useful outcome is deliberately modest: AI can turn a messy request into an organised draft. The contractor remains responsible for understanding the job, setting the price and making every promise. This workflow is separately assessed work, not an automatic website-plan feature. The narrow trial comes first; only a successful, controlled process becomes a candidate for a defined implementation.

For one enquiry, the owner decides not to use the draft at all. The photographs conflict and the location needs an in-person safety assessment. The system can flag uncertainty, but a person chooses the next move: call first and perhaps arrange a visit. That choice matters. Polished language is never a reason to skip missing trade knowledge, customer confirmation or an inspection that the work genuinely requires.