Every SME owner has now seen an AI chatbot demo. Fewer have seen one that actually helps a small business without embarrassing it. We've deployed these for clinics, restaurants and logistics firms — here's the honest playbook: where AI assistants genuinely earn their keep, where they don't, and how to deploy one safely.
The problem AI actually solves
It isn't 'replacing staff'. It's the 9pm problem: most enquiries arrive when nobody is working. A customer messages at 9:14pm asking if you deliver to Lekki. By morning, they've ordered from whoever answered — and somebody always answers. Speed-to-response is the single most underrated growth lever in small business, and it's exactly what machines are good at.
Where assistants genuinely earn their keep
- Repetitive questions — opening hours, prices, delivery zones, insurance accepted. Typically 60–80% of all enquiries, answerable instantly from your approved content.
- Lead qualification — greeting every enquiry, asking the three questions your sales process needs, and handing your team a summary instead of a cold name.
- Appointment booking — checking real availability and confirming, without the four-message WhatsApp dance.
- After-hours coverage — the assistant takes the 9pm enquiry, answers what it can, and books the callback for 9am.
Where they don't (and shouldn't)
- Complaints — an upset customer needs a human, fast. A good assistant recognises frustration and escalates immediately; a bad one loops them through menus.
- Anything medical, legal or financial beyond published facts — the assistant should state your published information and offer the professional, never improvise advice.
- Negotiation — pricing exceptions and custom deals are human work. The assistant's job is to hand them over warm.
How to deploy one without embarrassing your brand
The failures you've seen — bots inventing prices, promising refunds that don't exist — come from one mistake: letting the model improvise. A properly built assistant is trained only on your approved content, constrained to your topics, and given one graceful exit: 'Let me connect you with the team.' You review its behaviour before launch, and its conversations after, the way you'd supervise a new hire.
Start narrow. Launch with your top ten questions and booking, measure for a month, then widen. An assistant that does five things reliably beats one that does fifty things badly — and customers genuinely don't mind talking to a bot when it's fast, accurate and honest about being one.