Putting AI to work in your business costs you three things: a one-off build, a monthly running cost, and your team’s time. For a single automation, the build is usually the big number, often somewhere between £750 and £5,000, depending on who builds it and how much it connects to. The running cost is usually small: often under £50 a month for one workflow.
The real question isn’t “how much does AI cost?” It’s “how quickly does this particular job pay for itself?” This guide gives you the numbers to answer that for your own business.
Full disclosure: AI Business Systems builds AI systems. Our own prices are shown alongside the others so you can compare. Tool and AI prices change often: the figures here were correct at the time of writing (October 2026), and we review this guide regularly. Check each provider’s page for today’s prices.
Is AI too expensive for your business?
Probably not, but you’re not alone in worrying about it. Among UK businesses not using AI, 23% said “the cost is too high” (DSIT, AI Adoption Research).
Actual spending is lower than most people expect. Businesses using or planning to use AI reported a median spend of £2,000 in 2024 (DSIT). Half spent less than that.
The three costs of AI implementation
- The build. A one-off cost to design, connect, test and hand over the system.
- Running costs. Monthly subscriptions for the tools it runs on, plus AI usage.
- Your time. Explaining how the work happens now, testing on real examples, and learning the new way of working.
Most quotes only show the first. Ask for all three.
What does it cost to build?
Most UK AI firms don’t publish build prices. Of those that do, in October 2026 we found simple workflow automations starting from about £1,500, AI chatbots from about £5,000, and fuller first implementations quoted at £4,000 to £12,000.
| What you’re buying | What it does | Published UK starting prices we found | AI Business Systems |
|---|---|---|---|
| Workflow automation | Connects your tools and handles repeat steps, such as enquiries, reminders and data entry | From about £1,500 | From £750 |
| AI chatbot or enquiry assistant | Answers customers from your own information and hands over to you when unsure | From about £5,000 | From £1,500 |
| Custom AI tool or agent | Checks documents, researches leads, builds reports, using your rules | £4,000–£12,000 for a first implementation | From £3,000 |
What moves the build price:
- How many tools it connects. Each connection needs setting up and testing.
- How much judgement it needs. Moving data is cheap. Reading, deciding and writing take more building and more testing.
- How it’s checked. Anything customer-facing needs a person-in-the-loop step and more testing.
- How messy the current process is. If nobody agrees how the job is done, that gets sorted first.
Always get a fixed quote. If the price is a day rate with no cap, you’re carrying the risk.
What does it cost to run each month?
Running costs come in three parts. For one workflow, the total is often under £50 a month.
1. The automation platform. This is the software the workflow runs on. Entry prices at the time of writing (October 2026):
- Make: from $9 a month (Make pricing)
- Zapier: from $19.99 a month, billed annually (Zapier pricing)
- n8n Cloud: from €20 a month, billed annually. The self-hosted version is free, but you pay for the server it runs on (n8n pricing).
2. AI usage. When an automation uses AI to read or write, you pay per use, and it’s usually pennies. AI is priced in “tokens” (small chunks of text), and token prices change often, usually downwards. At the time of writing (October 2026), Anthropic’s Claude Haiku 4.5 costs $1 per million input tokens and $5 per million output tokens (Claude pricing). So if your automation reads 1,000 enquiries a month (about 1,500 tokens each) and drafts a 500-token reply to each, that’s about $4 a month:
- reading: 1.5 million tokens × $1 = $1.50
- writing: 0.5 million tokens × $5 = $2.50
3. AI assistant seats for your team (optional). If you want your team using an AI assistant day to day, that’s a per-person subscription. Prices at the time of writing (October 2026):
- Microsoft 365 Copilot Business: £16.10 per user a month on annual billing (£13.80 on a promotion to 31 December 2026), excluding VAT, on top of your Microsoft 365 plan (Microsoft UK)
- ChatGPT Business: $20 per user a month billed annually, or $25 monthly, with a minimum of two seats (OpenAI Help Centre)
- Claude Team: $20 per user a month billed annually, or $25 monthly (Claude pricing)
Support. Tools change, models get retired and connections break. Budget for someone to keep things running. One published UK range we found was £300 to £900 a month. Our care plan starts from £149 a month.
The costs people forget
- Your team’s time. Workshops, testing and learning the new process. Plan for a few hours per person.
- Data protection work. If AI touches personal data, the ICO says you’ll usually need a data protection impact assessment before you start (ICO).
- VAT and exchange rates. Many UK prices are quoted before VAT, and many tools bill in dollars or euros, so your monthly cost moves with the exchange rate.
- Seat minimums. Some team plans need at least two paid seats.
- Upkeep. AI models and app connections change. Something will need adjusting within the first year.
How to work out payback
Use this sum:
Payback (months) = build cost ÷ (monthly value of time saved − monthly running cost)
Example: a £1,500 automation saves 5 hours a week, and that time is worth £25 an hour to you.
- Time saved: 5 hours × 52 weeks ÷ 12 months ≈ 21.7 hours a month, worth about £542
- Minus £20 a month running costs = £522 a month
- £1,500 ÷ £522 ≈ just under 3 months to pay back
If your payback comes out at more than 12 months, question whether that job is worth automating yet. Most businesses that use AI report it pays back in time: 75% said it improved their workforce’s productivity (DSIT).
How to keep costs down
- Use what you already pay for. Check your Microsoft 365, Google Workspace and CRM for AI features before buying anything new.
- Start with one high-volume job. The more often a task happens, the faster automating it pays back.
- Use the cheapest AI model that does the job well. Prices vary several times over between models. A good build tests a cheaper model first.
- Insist on a fixed quote, with running costs listed in writing.
- Keep the accounts in your name, so you’re never paying someone else’s markup on your subscriptions.
Next steps
See every AI Business Systems price in one place. If you want each opportunity in your business costed, with build cost, running cost and payback, that’s what the AI Opportunity Audit does. Or start with one problem in the £197 AI Quick Start.