The actual problem
After-hours enquiries are easy to miss when they arrive through different channels. The wrong fix is to install a chatbot that pretends to know prices, availability or your business rules. Start with a reliable capture-and-review process.
Build the minimum useful version
- Pick one intake channel. Start with your website form or shared business inbox. Don't connect everything at once.
- Ask only for necessary details. Name, work email, business and a short problem description often suffice. Explain what you'll use them for; don't silently add people to marketing lists.
- Record a lead ID and timestamp. Keep a status such as
new → reviewed → replied → qualified → closed. Record a reference so a network retry doesn't silently create multiple jobs. - Set an owner review reminder. A phone notification, a short daily inbox check, or an authorised workflow can work. Storage without an alert is not a sales system.
- Draft, don't blindly send. If AI drafts a reply, check the actual question, facts, availability and customer expectations before it goes out.
- Track results. Separately count submitted enquiries, human-reviewed enquiries, replies sent, booked conversations, signed work and cleared payments. A page visit isn't a lead.
Copyable 90-second review checklist
Where AI helps — and where it shouldn't decide
AI can summarise a long message, identify a likely category, draft a clear response and flag missing information. It shouldn't invent a quote, commit staff time, make an irreversible promise, or silently copy sensitive customer data into an unapproved provider.
Costs and limits to check first
- Software: free tiers can have request/storage ceilings, overage charges or limits that change; check the live vendor terms.
- Ownership: who's looking at leads each day? A form that quietly fills with data is not enough.
- Privacy: UK GDPR and PECR may apply, particularly to retention, unsolicited follow-ups and third-party services. Get suitable advice for sensitive sectors.
- Failure modes: broken inbox integration, repeated submissions, spam, expired API keys, unauthorised access and poor error messages all need testing.
How to tell if the pilot worked
Choose a short baseline period, record enquiry counts and time-to-first-reviewed-response, then retest after adding the workflow. Example: five enquiries a week and a one-hour saving per enquiry would be five hours a week if those assumptions were real. It is only arithmetic, not a verified saving or guarantee.
Want this built around your business?
You can follow the steps yourself. Or Newton Revenue can assess the workflow, quote a clearly defined pilot and keep human approval where it matters.
Describe your enquiry problem →Editorial disclosure: AI Profit Brief is produced by Newton Revenue. No paid links or third-party endorsements are included in this guide.