Answering Customer Enquiries Faster With AI Drafts

Speed wins enquiries more often than polish

A homeowner fills in contact forms for three kitchen fitters on a Sunday evening. The first to reply with a sensible answer and a proposed survey date usually gets the job, and the one who replies on Wednesday with a beautifully written email is replying to someone who has already booked.

Most small businesses know this and still answer slowly, because replies need information from somewhere: the price list, the diary, the order system, the notes from last time. Hunting for it is what takes the time, not the typing. That is precisely the gap AI drafts close.

Some illustrative arithmetic. A trades business receiving 25 enquiries a week, taking 15 minutes each to research and answer, spends over six hours a week on first replies, much of it in the evening. If drafts cut that to three or four minutes of checking, the owner gets most of an evening back and customers hear back the same day instead of the next.

How AI-drafted replies work

The pattern is the same whether enquiries arrive by email, web form, WhatsApp Business or a marketplace inbox.

1. A new enquiry arrives and is read by the model, which identifies the type (quote request, availability, order status, complaint, something else) and extracts details such as location, product and dates.
2. Code looks up the facts: the relevant price band, next available slots, the customer's order history, the right help article.
3. The model writes a reply using only those facts, your approved wording and your tone.
4. A person sees the enquiry and the draft side by side, edits if needed, and sends.
5. Anything the model is unsure about, or that matches a sensitive category, is flagged rather than drafted.

Step two is what separates useful drafts from generic ones. A reply that says 'we'd be happy to help, what date suits you?' saves little. One that says 'we can survey in Leeds on Tuesday or Thursday afternoon; kitchens of that size usually fall in our mid range' saves the hunt.

Which enquiries to draft, and which to leave alone

Start with the two or three categories that make up most of your volume. For many service businesses that is quote requests and availability, which together might be 60 or 70 per cent of messages in a typical week.

Drafts, chatbots or full automation

The three approaches suit different volumes and risks.

• Drafts for approval suit most small businesses. Fast, safe, and staff stay in control of every word sent.
• A website chatbot suits businesses with many repetitive questions arriving out of hours. It answers instantly but needs careful limits on what it can say. Our post on choosing a chatbot for a small business covers that decision.
• Fully automatic replies suit only narrow cases such as acknowledgements and order status lookups, ideally after months of drafts showing the answers rarely need editing.

We usually recommend starting with drafts even when the goal is automation. The edit rate tells you, with real evidence, which categories are safe to automate later.

What goes wrong

• Invented prices or promises. A model without a price list will estimate one. Prices must come from your data or not appear.
• Rubber-stamping. After a few weeks of good drafts, people stop reading them. Keep sensitive categories flagged and sample sent replies each week.
• Stale information. Drafts are only as current as the price list, service areas and diary they draw on.
• Same reply, different customer. Repeat customers notice template-sounding replies. Include order history so drafts acknowledge the relationship.
• Personal data in the wrong tool. Enquiries contain names, addresses and sometimes health or financial details. Use tools with proper data processing terms.

The draft should make the right reply the easy one to send. It should never make a wrong reply the easy one.

Measuring whether it helps

Track three numbers for a month before and after: median time to first reply, conversion from enquiry to booking or quote, and the proportion of drafts sent without significant edits. If response time drops from hours to minutes and conversion holds or rises, it is working. If edits stay high for a category, the draft is not helping there and staff are better off with a template.

At SpiderHunts, when we build enquiry drafting for a business, we connect it to the systems that hold the answers first, then add the model. Our AI chatbot and assistant development covers both staff-facing drafts and customer-facing chat, and we will tell you plainly if a set of good templates would get you most of the way for a fraction of the cost.

Frequently asked questions