"Why would I pay you ten grand when there is an AI chatbot for £29 a month?" It is one of the most reasonable questions a small business owner can ask, and one of the most badly answered.
Most AI agencies wave their hands about quality and customisation. Most SaaS vendors quietly charge ten times their advertised price by year three. Here is a real, three-year cost comparison that gives you the actual answer.
The two options, properly described
For a UK SME thinking about an AI agent for, say, customer support, the realistic options are:
- Off-the-shelf AI SaaS. A subscription product like Intercom Fin, Tidio AI, ChatBot, or one of the dozens of newer entrants. You sign up, paste in some FAQs, plug it into your site.
- Custom AI agent. A bespoke build that talks to your real systems, knows your real content, and is shaped for your business specifically.
Both are valid. They are very different products, sold to very different stages of business.
Year one cost: SaaS
The trial price is usually £20 to £60 a month. Real working tier for an SME with meaningful volume is £150 to £500 a month. Add usage charges (per conversation, per AI message, per resolution) which often add another £100 to £400 a month.
A typical SaaS year one for a UK SME with real traffic: £3,500 to £10,000. Plus a few weeks of internal time getting it set up.
You also spend time editing the chatbot's FAQ document every week, trying to fix the latest gap. That hidden cost rarely makes it into the budget but is always present.
Year one cost: custom build
Build cost for a focused first project is typically £5,000 to £10,000. Running cost is £150 to £400 a month. Some light maintenance, optional, another £200 to £500 a month.
A typical custom year one: £8,000 to £15,000. Higher than SaaS at the start.
Year two and three: where it flips
SaaS pricing usually moves up. Either through advertised price rises (most vendors raise 10 to 20% a year) or through tier creep (you outgrow Pro, you go to Business). Realistic year two for the same SaaS setup: £4,500 to £14,000. Year three, similar or higher.
Three-year total for SaaS: £12,000 to £36,000.
Custom builds, by contrast, are mostly paid for. Year two and three are running costs and small maintenance. Realistic year two: £3,000 to £6,000. Year three, similar.
Three-year total for custom: £14,000 to £27,000.
The cost comparison most agencies do not show you
Three-year total cost of ownership, all in:
- SaaS: £12,000 to £36,000
- Custom: £14,000 to £27,000
By year three, the ranges overlap. By year five, custom is reliably cheaper for any business with real volume.
And that is before factoring in the bit that does not show up on an invoice, which is whether the thing actually does the job.
The cheap option is rarely the cheap option for long. It is just the option that delays paying the real cost until later, when it has cost you customers as well.
Where SaaS is genuinely the right answer
To be fair, off-the-shelf is the right call in some cases.
- You have very low volume and want to test whether anyone uses chat.
- Your enquiries are all answerable from a static FAQ. No system lookups, no actions.
- You want to capture leads or emails, not actually answer questions.
- You want something live this afternoon and can replace it later.
If any of those describe you, ignore everything else and start with a SaaS chatbot. Just be honest with yourself about what it is.
Where custom is the right answer
- You need the agent to talk to your real systems (orders, stock, calendars, CRM).
- Your enquiries are too varied for a static FAQ to cover.
- You want the agent to take real actions, not just chat.
- Brand voice and customer experience matter to you.
- You expect this thing to be running for years, not months.
The other thing nobody talks about
SaaS chatbots fail more often than they succeed for SMEs with real traffic. The reasons are well-rehearsed: generic, no real business knowledge, no integrations, no graceful escalation. We wrote about this at length in the hidden cost of cheap AI chatbots.
The point is not that SaaS is bad. It is that the failure case is much more common than the marketing pages suggest, and the failure case has its own costs.
How to decide for your business
If you can answer "no" to all of these, start with SaaS:
- Does the agent need to look up live data?
- Does it need to take actions on the customer's behalf?
- Will more than 20% of your enquiries need anything beyond a static FAQ?
- Is your brand voice a meaningful part of the customer experience?
If you said "yes" to two or more, you are looking at custom. The maths will work out if the use case is real.
If you would like a hand sizing it for your specific business, our strategy audit includes a real cost comparison against your situation. Or, if you have a clear picture, tell us about it and we will be straight about whether custom is worth it.