Should Your Small Business Add an AI Chatbot? What It Actually Costs, and When It's Worth It

A few clients have asked us the same question this year, almost word for word: "should we get one of those chat things on our website?" Usually it's prompted by a competitor's site having one, or a supplier pitching a chatbot add-on, rather than a clear problem they're trying to solve. That's the wrong starting point, so this post works backwards from it: what an AI chatbot actually is now, what it costs, and the specific situations where it earns its place on a small business website rather than just sitting there as a gimmick.

What's changed since the old "chatbot" you're thinking of

If your last experience of a website chatbot was a rigid box that only understood three pre-written questions and sent you round in circles, that's worth setting aside. Those scripted, decision-tree bots still exist, and they're fine for very narrow jobs like booking a table or checking an order status. But the ones getting talked about now are built on the same large language model technology behind tools like ChatGPT, trained or fed on your own website content, pricing, and FAQs, so they can answer questions in natural language rather than matching a fixed script.

That's a meaningful difference. A model-based chatbot can handle "do you deliver to Inverness and how long does that take" even if you never wrote that exact question down, as long as the answer exists somewhere in the content it's working from. A scripted bot can only answer what someone explicitly anticipated and wrote a button for.

What it actually costs

This is where the two types diverge sharply, and it's worth knowing before anyone quotes you a number.

A ready-made widget from a provider like Intercom, Tidio, or Crisp typically runs somewhere between £25 and £300 a month depending on how many conversations it handles and which AI features are switched on, with most small businesses landing in the £50 to £150 range. You're paying a subscription, dropping in a snippet of code, and feeding it your existing help pages. It's live within a day or two.

A custom-built chatbot, trained specifically on your product catalogue, pricing structure, and policies rather than a generic knowledge base, is a different proposition. Expect a build cost somewhere from £1,500 for something fairly simple up to £8,000-£10,000+ for something that needs to pull live data (stock levels, order status, booking availability) from your other systems, plus an ongoing monthly cost for the AI model usage itself, which scales with how many conversations it handles. For most small businesses, that ongoing model cost is genuinely modest, often £20 to £100 a month, but it's a separate line item people sometimes forget to budget for.

The honest advice: most small businesses should start with an off-the-shelf widget. The custom route only pays for itself once your off-the-shelf bot is clearly working and you've hit a specific limitation it can't get past, like needing real-time stock or booking data.

When it's actually worth having one

A chatbot earns its place when a few things are true at once, not just one of them.

You get the same questions, repeatedly, outside office hours. If your analytics or your inbox show a steady trickle of "are you open Saturdays," "do you ship to Ireland," or "what's your returns policy" arriving at 9pm and 7am, a chatbot answers those the moment they're asked instead of the next morning. That's the single clearest case for one.

Your FAQ page is long and visitors don't read it. Most people won't scroll through fifteen questions to find the one that matters to them. A chatbot that can search that same content and surface the right answer in one line removes that friction.

You genuinely can't staff live chat. A one or two person business can't realistically monitor a chat window all day. An AI chatbot that's upfront about being automated, and hands off cleanly to an email or callback request when it can't help, covers that gap without pretending to be a person.

It's generally not worth it when your site gets low traffic, when most enquiries are genuinely complex and need a real conversation anyway (a bespoke quote, a technical scoping call), or when your business sells on relationship and trust more than information, in which case an impersonal bot in the corner can do more harm than good to that first impression.

The trust problem nobody skips past

AI adoption among UK small businesses has climbed fast. The British Chambers of Commerce's 2026 "Future of Work" report, run with Atos and the University of Essex, found 54% of UK SMEs are now actively using AI, up from 35% in 2025 and just 23% in 2023. Separately, the Federation of Small Businesses found a similar adoption rate (55%) among the small businesses it surveyed, with 59% of those using AI reporting better productivity and an average 3% revenue increase among adopters.

But the same FSB research found 92% of small business owners now have concerns about AI, up from 73% in 2023, and the top worry is inaccurate responses, followed by security risks and unclear accountability when something goes wrong. That concern applies directly to chatbots: an AI model that's confidently wrong about your return policy or your pricing in writing, on your own website, is a worse outcome than no chatbot at all, because the customer reasonably assumes it's correct.

The fix isn't complicated, but it's not optional either. Whatever chatbot you use needs to be restricted to answering from your actual, current content (most good providers call this "grounding" or give you a way to upload a specific knowledge base), not left to answer freely from general training data where it might guess. And it needs an easy, obvious way for a visitor to say "that didn't answer my question" and get to a real person or a contact form, rather than being stuck arguing with a bot.

If you're going to try one

Start with the off-the-shelf route, feed it your actual FAQ and policy pages rather than a vague summary, and test it yourself with the ten questions you most dread getting wrong before a single customer touches it. Check the transcripts weekly for the first month. You'll usually find the bot handles 70-80% of enquiries cleanly and routes the rest to a human exactly as intended, which is a good result and worth keeping. If it's confidently making things up or sending people round in circles, it needs retraining or switching off, not persevering with out of sunk cost.

If you're weighing up whether a chatbot, a better FAQ page, or just a clearer website would actually solve the problem you're trying to fix, that's exactly the kind of conversation worth having before you buy anything. Book a free 15-minute consultation with Digital Hand and we'll help you work out which one, if any, is right for your business.