The most common question we get isn't "what can AI do for my business?" It's "what does it actually cost?"

The answers online are useless. You'll find articles quoting £500 chatbots, £500,000 transformations, and everything in between — with no framework to work out which one applies to you.

Here's the honest breakdown.

AI automation has four cost layers

Any system you deploy will have some or all of these:

1. Model API costs. Every time your system calls an LLM, you pay per token. For most SME workflows, this is £10–£200 per month.

2. Platform + hosting. The infrastructure your automation runs on — n8n, Make, a custom Vercel app, Retell for voice, etc. Typically £50–£500 per month.

3. Build / implementation. The one-off cost of designing and shipping the system. This is where the range is widest: £2,000–£50,000+.

4. Ongoing maintenance. Monitoring, tuning, handling errors, updating prompts, retraining when models change. £500–£5,000 per month.

Total year-one spend for a typical SME project: £10,000–£60,000. That's the range covering most of what we ship.

What drives the range

Three variables move the cost needle more than anything else.

Complexity of the decision being made

A chatbot that books meetings is cheap. An AI system that reads contracts, flags risks, and routes to the right person is expensive. The more judgment the AI has to apply, the more effort goes into grounding it, testing edge cases, and building review loops.

Integration surface area

One-tool automation is cheap. Five-tool orchestration (CRM + calendar + email + accounting + phone) costs more — not because the AI is harder, but because connecting real-world systems always takes longer than you think.

Custom vs. off-the-shelf

Sometimes the right answer is a £30/month SaaS tool with a 20-minute setup. A good agency will tell you that. A bad one will build you a bespoke system that costs £20,000 and does 80% of what the SaaS already does.

The cheapest AI project is the one you don't build because the right tool already exists.

Rough ranges by project type

Real numbers from what we typically quote:

| What you want | Typical build | Typical ongoing | |---|---|---| | Single workflow automation (e.g. lead intake to CRM) | £2–5k | £300–800/mo | | Content system (brand-tuned, approval queue, multi-channel) | £6–12k | £500–1.5k/mo | | Voice AI for inbound calls + CRM sync | £8–15k | £800–2k/mo | | Custom AI chatbot grounded in your docs | £5–12k | £400–1k/mo | | Full Faith OS deployment (multi-agent, persistent memory) | £15–30k | £2–4k/mo | | Bespoke AI product build (members-only tool, internal platform) | £25–75k | £3–8k/mo |

These are ballpark. Every engagement scopes before it prices.

What you should budget

If you're starting from zero:

  • £3–10k gets you a first live win — one workflow automated, real hours saved.
  • £10–50k in year one buys a genuine operational shift — three to five systems running, measurable output increase, a team that knows how to ask for more.
  • £50k+ is warranted when AI becomes core infrastructure — your sales pipeline, your service delivery, your product itself depends on it.

Most SMEs who go in expecting £100k of value for £5k spend are disappointed. Most who go in expecting £5k of value for £5k spend are delighted.

The bottom line

AI doesn't have to be expensive to be transformative. But cheap AI shipped without a system around it is usually worse than no AI.

Get an honest quote. Ask what's included. Ask what happens at month six.

If you want ours, book an audit. We'll map your ops, flag the two or three highest-ROI automations, and tell you exactly what each would cost to build and run.