Real UK pricing for workflow automation in 2026, from £500 quick builds to £12,000+ custom projects, and why most SMEs still see no return.
Published 23 August 2026. Updated 23 August 2026.
A basic workflow automation build in the UK costs somewhere between £500 and £1,500 in 2026. A proper multi-system build with an AI agent doing some of the thinking runs £5,000 to £30,000. Ongoing platform and maintenance costs then add £500 to £3,000 a month on top. Those are the real numbers. What most SMEs don’t budget for is the 20 to 40% that gets added afterwards in data cleaning, extra integrations and scope creep.
We get asked “how much does this actually cost” more than almost any other question at Nuvaleo AI, usually by a business owner in Salford or Trafford who’s just been quoted three wildly different prices by three different agencies. One quote says £800. Another says £15,000 for what sounds like the same job. Neither is necessarily wrong, they’re just pricing different things. So this is the honest version, built from published 2026 UK pricing data rather than a sales page, with the caveats most sales pages leave out.
Two things tend to go wrong when SMEs shop for this. The first is comparing quotes without comparing scope, so a £900 Zapier setup gets weighed against a £12,000 custom build as if they solve the same problem. The second is trusting the ROI promises without checking who funded the research behind them. Most of the eye-catching “AI pays for itself in six weeks” claims trace back to vendors selling the tool in question. That doesn’t make them false. It does mean they’re not neutral.
What workflow automation actually costs in 2026
A simple workflow automation build costs £500 to £1,500 in the UK, a mid-complexity project with several connected systems runs £1,500 to £6,000, and a full custom AI agent build typically lands between £5,000 and £30,000. Hourly rates for the people doing the work sit at £50 to £95, with a median around £65 to £70.
Pricing scales with complexity, not with how impressive the sales pitch sounds. Here’s the breakdown, based on 2026 UK market data from AutomationHire and ExpertSure:
| Complexity | Budget | What you get | Typical timeline |
|---|---|---|---|
| Simple | £500–£900 | 1–2 basic workflows | 2–5 days |
| Standard | £900–£1,500 | 3–5 connected workflows | 1–2 weeks |
| Intermediate | £1,500–£3,000 | Multi-tool integration | 2–4 weeks |
| Complex | £3,000–£6,000 | Multi-workflow builds tied to core operations | 4–8 weeks |
| Custom AI agent build | £5,000–£30,000 | Full AI-driven automation, agent development | 4–12+ weeks |
| Enterprise programme | £6,000–£12,000+ | Multi-department rollout | 2–4 months |
Most Manchester and Salford SMEs we talk to land in the £1,500 to £6,000 bracket for their first project. That covers a genuine multi-step workflow, not a single Zapier trigger someone could have built themselves on a Tuesday afternoon.
Off-the-shelf tools vs custom builds: the real cost difference
Off-the-shelf platforms (Zapier, Make, n8n) cost £8 to £480 a month and suit rule-based, high-volume tasks. Custom AI agent builds cost more upfront but handle judgement calls off-the-shelf tools can’t, and they don’t fall over the moment your process changes slightly.
This isn’t a case of one being universally better. It’s a case of matching the tool to the task, and most agencies quoting you have a reason to push you towards whichever one they sell.
| Off-the-shelf (Zapier/Make/n8n) | Custom AI agent build | |
|---|---|---|
| Monthly cost | £8–£480 | Often £0 subscription, but £70–£95/hr development |
| Setup cost | Usually £0–£1,500 | £5,000–£30,000 |
| Handles unstructured input (emails, calls, free text)? | Poorly | Well |
| Breaks when a form field changes? | Often | Rarely |
| Best for | High-volume, identical, rules-based tasks | Judgement calls, conversation, varied input |
n8n developers charge £60 to £95 an hour, Make.com specialists £55 to £90, and AI agent developers £70 to £95 and up, per current UK freelance and agency rate data. If your process is genuinely just “when X happens in system A, do Y in system B”, paying for a custom AI build is a waste of money. If the process involves reading something and deciding what it means, a rules-based tool will keep breaking and you’ll end up paying for the fix anyway.
Take a missed-call problem, which is one of the most common jobs we’re asked to fix. A rules-based tool can text a caller back automatically. It cannot tell the difference between a genuine new enquiry, an existing customer chasing an invoice, and a wrong number, because that requires reading the context of the call and deciding what matters. An AI agent can. Whether that distinction is worth £5,000 rather than £50 a month depends entirely on how many calls you’re missing and what a missed enquiry is actually worth to your business. For a plumber turning away three or four jobs a week, it usually is. For a business fielding the occasional wrong number, it usually isn’t.
The hidden costs nobody quotes you
Most quotes only cover the build. The extras are where budgets actually go over, and they rarely make it onto the first invoice.
- Data cleaning and preparation: often 20 to 40% of total project budget, according to ExpertSure’s 2026 cost analysis. If your CRM data is a mess, expect this to bite.
- Per-integration costs: roughly £3,000 to £12,000 each time you connect an additional system. Three integrations is not three times the price of one, but it’s not far off either.
- Annual maintenance: budget 15 to 20% of the initial build cost every year. Automation isn’t a one-off purchase, whatever the sales deck implies.
- LLM/AI token usage: £15 to £60 a month for AI-driven workflows, small on its own but easy to forget when you’re comparing quotes.
- Internal team time: reviewing outputs, handling exceptions, sitting in scoping calls. AutomationHire estimates this at £160 to £480 of staff time on a typical small project, and almost nobody puts it in the budget.
Get three quotes before you commit to anything. ExpertSure’s guidance on this is blunt, and it’s right: pricing variance between UK consultancies for comparable work is wide enough that the middle quote is rarely the safe choice. Ask each agency to itemise data prep, integrations and maintenance separately rather than folding them into one headline number. If they won’t, that’s the answer to whether they’re being straight with you.
Why the scary AI failure headlines don’t quite apply to your project
RAND Corporation’s research states that “by some estimates, more than 80 percent of AI projects fail, twice the failure rate of information technology projects that do not involve AI.” Separately, MIT’s Project NANDA found 95% of organisations investing in generative AI are seeing zero measurable return. Both come from large enterprise AI programmes costing millions, not a £3,000 Make.com workflow for a 12-person business, so the numbers don’t transfer directly. But the underlying reason for failure does.
Here’s where those figures actually come from, because they get thrown around constantly and rarely with the context attached.
RAND Corporation’s 2024 report, “The Root Causes of Failure for Artificial Intelligence Projects and How They Can Succeed,” is based on 65 interviews with AI practitioners and academics. It’s the source of the widely repeated “80% fail” line, though RAND itself is careful to frame it as an outside estimate rather than a number it calculated first-hand. MIT Project NANDA’s “State of AI in Business 2025” report is more directly sourced: interviews with 52 organisations, survey responses from 153 senior leaders, and analysis of over 300 publicly disclosed AI initiatives, all pointing to that 95% zero-return figure. Separately, UK consultancy Emergn calculated that large UK firms outside financial services write off roughly £67bn a year on transformation and AI work that fails to deliver, based on a survey of 700 senior leaders. That figure applies to big firms running an average of 6.4 concurrent AI initiatives at once. It has nothing to do with your business booking a chatbot build.
The part that does transfer is more interesting than the headline number. MIT NANDA found that AI tools built through an external partner reached deployment around 67% of the time, against roughly 33% for tools built entirely in-house. Whatever you think of that as a plug for hiring someone like us, it lines up with the same pattern the Federation of Small Businesses found closer to home: 55% of UK small firms now use AI, up from 20% in 2023, but only 24% report a revenue increase and only 22% credit it with actual business growth. Adoption went up. Results mostly didn’t follow, and the gap tends to sit with how the project was scoped and delivered, not with whether AI was the right idea in the first place.
The failure mode at SME scale isn’t a £7 million enterprise write-off. It’s simpler and duller: buying a tool, bolting it onto a messy process, and never measuring whether it actually saved anyone any time. We wrote about this in more detail in our honest guide to AI consultancy in Manchester and Salford, and the pattern hasn’t changed since we wrote it.
A realistic ROI timeline for a UK SME
Most UK SMEs recover the cost of a workflow automation project within three to six months through time saved, assuming the process was mapped properly before anything was built. That’s slower than most sales pages promise and faster than most sceptics expect.
AutomationHire cites one real example: a £1,400 build that saved nine hours a week reached break-even in about 1.3 months, on the basis that nine hours of admin time is worth roughly the cost of the build within five to six weeks. That’s a genuinely fast payback, and it’s also a fairly small project. Bigger builds take longer to earn out, not because they’re less effective, but because there’s more money to recoup.
There’s a trade-off worth being honest about. A £900 Zapier setup will often pay for itself faster in pure pound terms than a £15,000 AI agent build, simply because there’s less to recoup. That doesn’t mean the smaller build is the better decision. If the cheap setup only saves two hours a week and keeps breaking every time a supplier changes their order form, the “faster ROI” is an illusion built on a low starting number. Payback speed only means something alongside what the automation is actually worth to the business once it’s running properly.
How to not be one of the ones who gets nothing back
- Map the process before you price anything. Write down every manual step. If you can’t describe the current process in plain English, no developer can automate it properly, whatever they tell you in the sales call.
- Set a success metric before you start. Hours saved, leads recovered, error rate reduced. Pick one number and write it down. MIT’s data says this roughly quadruples your odds of the project actually working.
- Get three quotes, not one. UK pricing for comparable automation work varies enormously between agencies. A single quote tells you nothing about whether it’s fair.
- Start with the highest-volume, most annoying task, not the most technically interesting one. Missed calls and manual data entry are boring. They’re also usually where the money is.
- Budget the hidden 20 to 40%. Data cleaning, integrations and internal review time are real costs. Put them in the spreadsheet before you sign anything, not after.
- Review it after three months. Not five years. Three months. If it isn’t saving the time you expected by then, fix it or stop, rather than letting it quietly become another line in someone’s £67bn write-off.
The bottom line
The honest number for most UK SMEs starting their first automation project is somewhere between £1,500 and £6,000, with a further 20 to 40% in costs that rarely make it onto the first quote. That’s not a small sum for a small business, but it’s a fraction of what the scary enterprise failure headlines are describing, and the projects that actually pay off share one thing in common. Somebody measured them.
10. Author bio
Reggi, Founder & CEO, Nuvaleo AI
Reggi is the founder of Nuvaleo AI, a Salford-based automation and AI agency building custom chatbots, workflow automation and CRM systems for small and medium-sized businesses across Greater Manchester. Nuvaleo AI’s client work includes freeing more than 30 hours a week of manual admin for operations client PYT. Follow Nuvaleo AI on LinkedIn or X.
11. FAQ
How much does workflow automation cost for a small business in the UK? Most UK SMEs pay £500 to £1,500 for a simple build, £1,500 to £6,000 for a mid-complexity project connecting several systems, and £5,000 to £30,000 for a custom AI agent build. Add 20 to 40% for data cleaning, extra integrations and ongoing maintenance that rarely appear on the first quote.
Is Zapier or Make cheaper than a custom automation build? Yes, upfront. Zapier and Make cost £8 to £480 a month against £5,000 or more for a custom build. But they only handle structured, rules-based tasks well. If your process involves judgement calls or varied input, a custom build often costs less over 12 months once you count the fixes.
Do most AI automation projects actually fail? Large-scale figures like RAND’s “more than 80% fail” estimate and MIT’s 95% zero-return finding describe enterprise AI programmes costing millions, not small workflow builds. At SME scale, the real risk isn’t the technology failing. It’s buying a tool without mapping the underlying process or setting a measurable target first.
How long does it take to see ROI from workflow automation? Most UK SMEs break even within three to six months, based on time saved. Smaller, well-scoped projects can break even in under two months. Larger builds take longer to pay back, simply because there’s more cost to recover.
What’s the biggest hidden cost in a workflow automation project? Data cleaning and preparation, which can eat 20 to 40% of the total budget if your existing systems (CRM, spreadsheets, inboxes) are disorganised. Per-integration costs and ongoing maintenance are the next two most commonly missed.
Should I automate with off-the-shelf tools or hire an agency? Depends on the task. Simple, high-volume, rules-based work suits Zapier, Make or n8n and you can often build it yourself. Anything involving unstructured input, conversation, or a process that changes regularly usually needs a custom build, because rules-based tools break every time the input varies.
12. Related articles
- Half of UK SMEs now use AI. Most get nothing back: an honest 2026 guide to AI consultancy in Manchester and Salford, anchor text: “our honest guide to AI consultancy in Manchester and Salford”
- Business Automation: The Control Layer SMEs Miss, anchor text: “the control layer most SMEs miss”
- AI Business Process Automation: The Complete Guide for UK SMBs, anchor text: “our complete guide to AI business process automation”
- AI for Salford Businesses: 2026 Sector-by-Sector Guide, anchor text: “our sector-by-sector guide for Salford businesses”
- Nuvaleo AI Case Studies, anchor text: “see how Nuvaleo AI clients like PYT freed 30+ hours a week”