AI and customer service: what is actually possible for small businesses right now

AI in customer service gets talked about in one of two ways: breathless hype about robots replacing call centres, or dismissive scepticism from people who have been on hold for 45 minutes talking to an automated phone tree. Neither is helpful.

Here is what is actually happening, and what is genuinely available to a small professional services firm right now.

The spectrum: from basic to advanced

At the basic end, AI customer service has existed for years. FAQ chatbots. Pre-written canned responses triggered by keywords. Automated out-of-office emails. These are useful but limited. They answer the same question the same way, regardless of context, and fall over the moment a query goes slightly off-script.

The middle ground is where most small firms are starting to operate. Tools like Tidio, Intercom, and Freshdesk now use genuine language models underneath them. They can understand a question even when it is not phrased exactly right, and they can pull from a knowledge base to give a contextually relevant answer. Not perfect, but meaningfully better than keyword matching.

At the advanced end, things are moving fast. Meta announced in early 2026 that its AI handles support queries on Facebook and Instagram in under five seconds, across more than a billion interactions. That is not a chatbot reading from a script. That is an AI agent understanding context, accessing account information, and resolving issues autonomously.

That level of capability is not available as an off-the-shelf product for small firms today. But the direction is clear, and the tools are getting there.

What is realistic for a small professional services firm right now

Let's be specific about what you can actually implement.

A Tidio chatbot answering common questions. Tidio (and similar tools like Tawk.to or Crisp) lets you build a chatbot that sits on your website and handles the questions you get asked repeatedly. Opening hours. What you charge. How to book. How to get started. A well-configured chatbot handles around 40 to 60 per cent of incoming queries without any human involvement. Setup takes a few hours. Cost: free to low.

Claude-drafted response templates. If you handle client queries by email, you probably find yourself writing versions of the same email repeatedly. Use Claude or ChatGPT to write a library of response templates. "Requesting more information from a new enquiry." "Explaining your onboarding process." "Following up on an unpaid invoice." Humans still send the emails, but the drafting time drops to almost nothing.

Automated follow-up sequences via Zapier. Connect your contact form to Zapier and build a sequence: immediate acknowledgement email when someone submits an enquiry, a follow-up 24 hours later if no response, a reminder after 48 hours. No AI required for the logic, but you can use ChatGPT to write the email copy and Zapier to send it. The result is a professional, consistent follow-up process that runs without you.

AI-assisted triage. Some CRM tools now let you use AI to categorise incoming messages, flag urgency, and route to the right person. Useful if you have a team. HubSpot and Zoho both have features in this area.

What requires caution

Two things to take seriously.

Regulated advice. If your firm offers legal, financial, tax, or medical advice, AI cannot give that advice to clients. Full stop. It does not matter how capable the AI is. Regulated advice requires a qualified person to be accountable for it. An AI chatbot that answers questions like "should I set up as a limited company" on a solicitor's website is a risk, not a feature. Use AI to handle logistics and process. Keep the advice with humans.

Transparency. In many regulated contexts, and increasingly under general consumer expectations, a client-facing AI must be clearly identified as AI. Do not try to make your chatbot seem like a human. It erodes trust when people realise, and in some sectors it may breach FCA or SRA rules.

The principle that holds everything together

AI handles volume and routine. Humans handle relationships and judgement.

The questions that arrive at 2am, the angry client who needs to feel heard, the unusual query that sits outside your FAQ, the moment where someone needs reassurance rather than information: those stay with you.

The 40 enquiries a month asking the same three questions, the follow-up email you always forget to send, the client portal FAQ that should be answering basic questions before they reach you: those are exactly what AI is for.

Start there, do it properly, and build from it.

---

AdaHQ covers AI tools for UK professional services businesses. Browse the AI tools directory for more.