Most pricing pages for AI chatbots are written by software companies that want you on a per-seat or per-conversation tier. That's fine for big companies running thousands of conversations a day. For a UK small business with a hundred enquiries a month, those models are a trap. You either underspend and end up with something that doesn't work, or you overspend on capacity you'll never use.
This guide gives you an honest breakdown of what an AI chatbot actually costs for a small UK business in 2026, the five main ways to buy one, and what to watch out for in the fine print.
Short answer: A focused, fully managed AI chatbot for a UK small business typically costs £300–£500 to set up and £200–£400 a month to run. A bespoke managed implementation with a large knowledge base, technical product routing, tailored reporting, and ongoing optimisation will more typically cost £3,000–£10,000 to set up and £500–£1,500 a month. A fully custom system with several integrations or business-critical workflows starts higher again. Beware anything that charges per conversation, as it can scale badly the moment the bot starts working.
The Five Ways to Buy an AI Chatbot
The market splits into five broad tiers. They cost wildly different amounts, and they are not really competing for the same buyer. Knowing which tier you are in saves a lot of pain.
1. DIY chatbot platforms
Cost: £20–£200/month, plus your time
You sign up, paste your website URL, the platform scrapes your content, and you get a working chatbot in an hour. Sounds great. The catch is everything after that hour. You're the one who decides what it says when it doesn't know. You're the one who updates the training when your prices change. You're the one watching for the conversations where it confidently makes things up. For a busy business owner, the monthly fee is the smallest cost. The real cost is the time it eats, or the leads it loses when nobody's maintaining it.
2. White-label resellers
Cost: £100–£400/month, often with a hidden setup fee
A growing category. These are usually solo operators reselling a generic SaaS platform under their own brand. The chatbot itself is decent. The risk is what happens when the operator gets busy or moves on. You're paying a managed-service price for a product they don't actually control, and if their account gets shut down, your chatbot goes with it. Check how long the reseller has been trading and whether they're a UK-registered company before signing anything.
3. Fully managed service
Cost: £300–£500 setup, then £200–£400/month, no per-conversation fees
Someone else builds it, trains it, maintains it, and runs it. You don't touch the platform. You get a chatbot on your site, lead summaries in your inbox, and a real person to email when something needs changing. This is the tier most small UK businesses should be in if they don't want chatbot maintenance to become a second job. The fixed monthly fee removes the worst part of the DIY model, which is per-conversation pricing that punishes you the moment the bot starts working.
4. Bespoke managed implementation
Cost: £3,000–£10,000 setup, then £500–£1,500/month for management, usage, reporting, and improvement
This uses an established AI platform but involves substantially more consultancy, content work, testing, and ongoing management than a standard small-business assistant. It suits organisations with a large or technical knowledge base, several product lines, detailed routing rules, tailored lead handoff, usage controls, and reporting requirements.
The initial fee pays for auditing and structuring the source material, separating current information from outdated or conflicting versions, mapping visitor needs to the right products, configuring boundaries, testing difficult questions, and managing client review before launch. The monthly fee is higher because monitoring, content refreshes, reporting, usage capacity, and continuous routing improvements are part of the service. This is often the right middle ground when the requirement is genuinely bespoke but does not justify building an entire software platform from scratch.
5. Fully custom engineered system
Cost: £10,000–£50,000+ upfront, then £1,000–£5,000+/month for support, monitoring, and infrastructure
A developer or specialist studio designs software around the way your business operates: connecting several CRMs, booking platforms, payment services, inboxes, or internal databases; writing information back; triggering workflows; and handling permissions, audit trails, testing, and failure cases. A tightly scoped workflow may sit near the bottom of the range. Multi-system, regulated, voice-enabled, or business-critical work can go considerably higher.
It is the right route when the system is expected to take meaningful operational actions, not just answer questions, recommend content, and capture an enquiry. For a business that needs a sophisticated assistant on top of an established platform, a bespoke managed implementation is normally the more proportionate option.
What Does £275 a Month Actually Get You?
This is roughly the going rate for a fully managed AI chatbot for a UK small business in 2026. For that, you should be getting:
- A branded chatbot in your colours, embedded on your website
- Training on your actual services, prices, hours, locations, and policies
- Lead capture that emails you the enquiry within seconds
- Stress-testing with realistic scenarios before it goes live
- Monthly retraining as your business changes
- A monthly report so you can see what it handled and what it caught
- A real person on the other end of an email when something needs adjusting
- Documented data handling, a clear processing location, and agreed retention terms
- A flat fee, regardless of how many conversations the bot has
If you're paying that monthly fee and you're still expected to log into a dashboard to train the bot yourself, you're paying for software, not a service. Worth knowing the difference.
Price Comparison: What Are You Actually Paying For?
| Option | Setup | Ongoing | Your time |
|---|---|---|---|
| DIY chatbot platform | £0 | £20–£200/mo + per-conv. fees | 2–5 hrs/mo |
| White-label reseller | £0–£500 | £100–£400/mo | Minimal |
| Fully managed service | £300–£500 | £200–£400/mo flat | None |
| Bespoke managed implementation | £3,000–£10,000 | £500–£1,500/mo | Discovery and review input |
| Fully custom engineered system | £10,000–£50,000+ | £1,000–£5,000+/mo | Ongoing operational input |
The Per-Conversation Trap
The single biggest thing to watch out for is per-conversation pricing. It sounds reasonable on the surface. You only pay for what you use. In practice, it's structured to punish you the moment the bot becomes useful.
SaaS chatbots typically charge per message, per conversation, or per resolved query, on top of a base fee. The exact rates vary widely between platforms, but the structural problem is the same: a "conversation" can be anything from a one-message bounce to a 40-message booking flow, and the meter is running the whole time. The month the bot finally starts working is the month the bill jumps, and there's no ceiling. Predictable monthly billing matters more than people realise.
If you're being quoted per-conversation pricing, ask what happens to the bill if the bot has 100 conversations a day. The answer is usually uncomfortable.
What Drives the Price Up or Down?
Integrations
There is a large difference between sending a lead summary by email and building a system that can safely read, update, and coordinate business data. A proven, one-way connector to a standard platform may be included in a managed service or quoted as a separate phase. A custom integration with an in-house system will commonly add £2,000–£8,000; several two-way integrations, complex permissions, or business-critical workflow logic will usually move the project into the £10,000–£50,000+ custom-engineering bracket.
Complexity of the conversation flow
A chatbot that answers FAQs and captures leads is relatively straightforward. A system that triages emergency calls, takes payments, remembers a customer across sessions, updates records, or decides which workflow to trigger is a different product. The extra cost is not for a longer prompt; it pays for discovery, engineering, safeguards, testing, monitoring, and a dependable route back to a person when something goes wrong.
Volume
For a fully managed flat-fee service, volume usually doesn't change the price unless you're in the top 1% of usage. For per-conversation models, volume is the entire pricing model, and it's the main reason flat-fee managed services exist.
How niche your business is
An estate agent or a dental practice has well-understood workflows. The build is fast. A specialist B2B business with unusual terminology needs more discovery and training time before the bot is useful. Expect a higher setup fee or a longer launch period.
Hidden Costs to Watch For
- Per-conversation fees on top of a monthly fee. If both exist, your bill is unpredictable.
- Token-based pricing. Some platforms charge by the AI token. Useful for engineers, awful for budgeting.
- "Free" tiers with hard caps. Once you hit 1,000 messages, the price often jumps by 4–10x.
- The wrong contract for the service. A standard managed assistant should be able to earn its keep on a short rolling agreement. A bespoke managed implementation may need an initial launch and stabilisation period. A fully custom system will normally have staged project payments and a separate support agreement because substantial discovery and engineering happen before launch.
- Setup fees that look like subscriptions. Some providers charge a "monthly setup" for the first three months. Read the small print.
- Extra charge for "premium models". A few platforms keep GPT-4-class models behind a higher tier and use weaker ones by default. You usually want the better model, especially for anything customer-facing.
Will an AI Chatbot Actually Make You Money?
Honestly, that depends on what kind of business you run. If most of your enquiries already come in by phone during office hours and you answer them within minutes, the value of a chatbot is mainly in saving time on repeat questions. Useful, but not transformative.
The businesses where an AI chatbot earns its keep many times over are the ones where enquiries come in outside business hours and the response time is the deciding factor. That's estate agents catching evening property enquiries (Rightmove's busiest hour is 8–9pm), trades getting emergency call-outs at midnight, restaurants taking bookings while the kitchen is closed, dental practices fielding implant questions on a Sunday morning. The underlying research is consistent: responding to a web lead within 5 minutes makes you 21 times more likely to qualify it than responding in 30 (Oldroyd et al., Harvard Business Review, March 2011). Lead capture out of hours is the killer use case. Everything else is a bonus.
A useful rough calculation: if a single captured enquiry per month would cover the monthly cost of the chatbot, the maths almost certainly works. For most service businesses, that bar is low.
Six Questions to Ask Before You Sign Anything
- Is the pricing flat or per-conversation? Flat is what you want unless you're absolutely certain the bot will never become busy.
- Who actually maintains it? Get a clear answer. If the answer is "you, through the dashboard", you're buying software, not a service.
- What happens when my services or prices change? Find out whether retraining is included or billed extra.
- Where is the data stored and processed? If any processing happens outside the UK, ask for the Data Processing Addendum, the international-transfer safeguards being used, the retention period, and the wording visitors will see.
- What am I committing to? For a managed assistant, ask whether you can cancel monthly. For a bespoke managed implementation, clarify the launch and stabilisation period, usage allowance, reporting, and ongoing management. For custom software, ask for milestones, acceptance criteria, ownership, support terms, and what happens if the scope changes.
- Can I see one working? Any decent provider has a live demo. Not a video, an actual bot you can talk to.
The Bottom Line
A focused, fully managed AI chatbot for a UK small business in 2026 should cost around £300–£500 to set up and £200–£400 a month to run. Flat fee. No per-conversation surprises. Clear data-processing and retention terms, with a human on the other end of an email when you need to change something.
If the brief involves a substantial technical knowledge base, product-routing logic, tailored reporting, usage controls, and ongoing optimisation, budget around £3,000–£10,000 to set up and £500–£1,500 a month for a bespoke managed implementation. If it must also connect several systems, update records, take actions, or become part of day-to-day operations, a custom-engineered build will more typically start at £10,000 and can run well beyond £50,000, with ongoing support on top.
The DIY tools are cheaper on paper but cost you time. A focused managed service remains the sensible answer for routine questions and lead capture. The bespoke managed tier is for organisations that need more depth, control, and continual improvement. Fully custom engineering is for businesses with valuable workflows to automate and a budget to match.
If you're a small business in Cheshire, the North West, or anywhere in the UK, and you want a straight conversation about what a managed AI chatbot would cost for your specific setup, have a look at how Delegait AI works or get in touch below. No pitch, no pressure, just a clear answer.