A founder asks three AI automation agencies to quote the same job: automate lead intake and follow-up. The quotes come back at £4,000, £28,000 and £190,000. Same brief. Same outcome described. He assumes two of them are trying it on. In fact, all three might be reasonable – they are pricing different amounts of governance, integration, ownership and risk, and none of them explained which. Pricing opacity is the norm in this market. Most agencies defer pricing until deep into the sales process, after you’ve sunk hours into discovery.
Illustrative scenario: The opening scene is a composite used to explain the operational risk. It is not a client case study or claimed result.
This page fixes that. Here are real 2026 ranges, what drives them, and the pricing patterns that should make you walk away.
Short answer: In the UK, a defined AI automation build typically costs £5,000–£30,000 for most SME projects, with monthly retainers of £3,000–£15,000 for ongoing multi-workflow work; smaller productised support starts around £300–£1,500 a month. In USD, project fees commonly run $5,000–$75,000 and retainers $3,000–$20,000 a month. Where you land depends on scope, data quality, integration depth and how much risk the system carries.
The Four Pricing Models
| Model | Typical UK range | Typical USD range | Best for |
|---|---|---|---|
| Project / build fee | £5,000–£30,000 (SME); £25,000–£250,000+ (enterprise) | $5,000–$75,000; $250,000+ enterprise | A clear, one-time build |
| Monthly retainer | £3,000–£15,000/mo; £300–£1,500/mo productised | $3,000–$20,000/mo | Ongoing build, monitoring, iteration |
| Per-workflow | £4,000–£12,000 per connected set | $2,000–$12,000 each | A menu of standard, repeatable builds |
| Day rate / hourly | £400–£2,500/day; £150–£400/hr | $150–$350/hr | Discovery, audits, undefined work |
A few honest notes on these numbers. The average AI-related retainer sits nearer the lower-middle of the USD range than the top – one industry benchmark (Digital Agency Network) puts the average AI SEO/services retainer around $3,200 a month. UK day rates tier cleanly: roughly £400–£800 for a freelance practitioner, £900–£1,600 for a mid-tier consultancy, £1,200–£2,500 for a boutique specialist, and £1,500–£3,000-plus for Big Four and enterprise firms. Outcome-based pricing – a fee tied to measured impact – is growing, but it only works when both sides share honest metrics.
What Actually Drives the Cost
A quote is not arbitrary. The same lead-routing or invoice-processing workflow can justify a £3,000 fee in one business and a £15,000 fee in another, because five things move the number:
- Scope and complexity. One automation is cheap. A connected set of agents that plan and execute multi-step tasks is not. Cost climbs steeply as you move from rule-based, to AI-powered, to fully agentic.
- Integration depth. Plugging into Salesforce, HubSpot or Sage typically adds £3,000–£12,000 per integration. Legacy systems cost more.
- Data quality. Clean, well-structured data is cheap to work with. Messy data is where budgets quietly double, and it is the single biggest cause of failed projects.
- Risk and domain. A system that touches clinical, financial or legal claims needs red-tier governance – mandatory human sign-off, audit trails, evidence ledgers. That is real work, and it is the work that keeps you out of trouble.
- The value of the workflow. A workflow worth £400,000 a year in saved time justifies a bigger fee than one worth £4,000. Good agencies price partly on value, which is fair only when the value is measured.
The Hidden Costs Nobody Quotes
The sticker price is not the total. Budget for these or be surprised by them:
- Model and API spend. This is usage-based and it scales with volume, not with the build. A platform-plus-model bill for a small business might be £40–£150 a month; a high-volume system is a different story. One widely reported cautionary tale, cited by IBM-study coverage in 2026, involved an enterprise burning through roughly $500m in AI token costs in a single month after failing to set spend limits – an extreme case, but the mechanism (no hard cap) is common. Always set hard limits in your model provider’s console.
- Integration maintenance. Third-party APIs change. Budget several developer hours per integration per year just to keep things connected.
- Ongoing maintenance. Models drift, prompts degrade. Budget annual maintenance at roughly 15–20% of the initial build cost. A system with nobody maintaining it usually breaks within months.
- Governance overhead. The interlock map, the risk tiering, the evidence ledger and the human review time. This is not a cost to minimise. It is the cost that prevents the 2am clinical-claim rewrite. Underbudgeting it is why so many projects fail after launch, not before.
How to Read the Model / API Line
Here is the mechanism, because it trips up buyers. You pay for build once. You pay for tokens forever, and in proportion to use. An automation that handles thirty enquiries a month and one that handles three thousand cost roughly the same to build and wildly different amounts to run. If an agency bundles model costs into a flat fee with no visibility, ask what happens when volume spikes. Pass-through with a modest markup and a hard cap is more honest than an opaque bundle.
Red-Flag Pricing
Walk away, or ask hard questions, when you see:
- Hourly-only pricing on a complex, definable build. Open-ended hours shift all budget risk to you. For anything you can define, insist on a fixed fee or a capped retainer.
- A price with no governance line. If the quote has no line for review, monitoring or oversight, the agency is either hiding the cost or not doing the work. Both are bad.
- A demo-ware discount. A suspiciously cheap quote often means a templated chatbot dressed as a bespoke system. It will look great in the demo and fail in production.
- Proprietary-framework lock-in. If they insist on their closed framework and won’t discuss code ownership or handover, you are locked in forever.
- Outcome pricing with no shared metric. “We only charge on results” sounds fair until you realise nobody agreed how results are measured. Without an evidence ledger, it’s theatre.
For the full checklist on separating a real agency from three freelancers and a subscription, see how to choose an AI automation agency.
Frequently Asked Questions
How much does an AI automation agency cost in the UK? Most SME builds run £5,000–£30,000 as a one-off; ongoing retainers run £3,000–£15,000 a month for multi-workflow work, with productised support from around £300–£1,500 a month. Enterprise programmes run £25,000–£250,000 and up.
How much does it cost in USD? Project fees commonly run $5,000–$75,000, retainers $3,000–$20,000 a month, per-workflow builds $2,000–$12,000, and senior hourly rates $150–$350. Enterprise agentic systems can exceed $250,000.
Why won’t agencies publish their prices? Because scope, data quality, integration depth and risk vary so much that a single number would mislead. That’s a fair reason to defer a precise figure – it is not a fair reason to refuse a written range and honest scoping before you commit.
What’s the most underbudgeted cost? Governance and ongoing maintenance. Models drift and API costs scale with use. Budget 15–20% of the build annually for maintenance, and treat the governance line as non-negotiable, not as padding.
Is outcome-based pricing a good idea? Only with a shared, honest metric. It aligns incentives well when both sides agree how impact is measured and log it – which is exactly what an evidence ledger is for. Without one, it’s a slogan.
Next step
send us a workflow and we’ll scope it with written deliverables and an honest range before you commit anything – including the governance line most quotes leave off.*
Continue the buyer’s guide
What an AI Automation Agency Actually Does (2026)AI Automation Agency vs In-House: Honest GuideHow to Choose an AI Automation Agency (2026)AI Automation Agencies UK: 2026 Buyer’s GuideStart with the interlock map
We map what an agent may read, write and release before we build the production system around it.
Book a strategy call