Key Takeaways
- A useful AI automation project can cost from hundreds of pounds for one tightly defined workflow to several thousand pounds for connected, business-critical processes.
- The build fee is only part of the cost. Allow for software, AI usage, testing, monitoring, training and changes after launch.
- At MattDarm, our current starting points are £499 for one workflow, £1,500 for a connected growth setup and £5,000+ for complex automation architecture. Prices exclude VAT and third-party fees.
- Use normal automation for fixed rules and AI only where the work involves language, documents, classification or judgement support.
- The right question is not whether automation looks clever. It is whether it improves a real process enough to justify its full cost.
For a UK small business, AI automation will usually cost somewhere between a focused £499 setup and a £5,000+ implementation, depending on the number of workflows, systems and risks involved. That is the honest short answer based on our current AI automation setup packages. Software subscriptions, model usage and ongoing support sit on top.
That range is wide because “AI automation” can describe very different work. Automatically moving a website enquiry into a spreadsheet is not the same as reading documents, checking customer records, updating a CRM, drafting a reply and alerting a manager when something looks wrong.
The cheapest build is not always the best value. Equally, a small business should not pay for an elaborate AI system when a straightforward rule would solve the problem. A sensible quote starts with the process, shows where the money goes and makes it clear how the workflow will be checked after launch.
Our Current AI Automation Prices
As of 18 July 2026, MattDarm's public starting prices are:
- Starter: £499 one-off, excluding VAT. One automation workflow, up to three app integrations, basic error handling, setup, testing, a team walkthrough and 30 days of support.
- Growth: £1,500 one-off, excluding VAT. Up to five connected workflows, broader integrations, AI decision support, monitoring, CRM or email automation, team training and 60 days of support.
- Enterprise: £5,000+ one-off, excluding VAT. A larger architecture with custom APIs, multiple departments, more advanced monitoring, data classification and a fuller training programme.
- Ongoing monitoring: from £49 per month, excluding VAT. The appropriate level depends on how frequently the workflow runs and how serious a failure would be.
These are our prices, not a claim about a universal UK market average. Another provider may package discovery, software and support differently. When comparing quotes, check the scope rather than comparing the headline number alone.
What You Are Actually Paying For
A reliable automation is more than a few boxes connected on a screen. There are six cost areas worth understanding.
1. Process Mapping
Before anything is built, somebody needs to understand the current job. What starts it? Which systems are involved? What information is required? What counts as a successful result? Which exceptions need a person?
This stage can expose a problem that technology will not fix. If different team members follow different rules, customer records are incomplete or nobody owns the final decision, automating the process will move the confusion faster.
A small project may need one focused workshop. A more involved project may need interviews, sample documents and a written workflow map. That discovery time belongs in the budget because it prevents expensive rebuilding later.
2. Building the Workflow and Integrations
The number of steps matters, but the difficulty of the connections matters more. Joining a common form, Gmail and Google Sheets may be relatively straightforward. Connecting an older CRM, an internal database or software with a limited API can take longer even if the finished workflow looks simple.
Tools such as n8n and Make can reduce development time, but they do not remove the need to design, test and maintain the workflow. Their charging models also differ. Make generally charges by credits consumed by actions, while n8n's cloud plans count complete workflow executions. Estimate usage from the real process before choosing a platform.
3. AI Model and Software Usage
AI steps can create variable costs. Reading a short enquiry is inexpensive compared with processing large documents or running several model calls for every customer. The workflow may also rely on a CRM, document store, email platform, transcription service or paid data source.
Ask for a simple monthly estimate using realistic volumes. How many enquiries, documents or records will pass through the system? What happens in a busy month? Is there a spending cap? A good setup makes usage visible instead of waiting for a surprising bill.
4. Testing and Human Approval
A demo proves that the happy path can work. Testing proves whether the workflow can be trusted when the data is incomplete, duplicated, unusually formatted or simply wrong.
Testing should cover ordinary examples, awkward examples and failure cases. Important customer messages, prices, regulated decisions and permanent record changes should normally keep a human approval point. That review time is part of the operating cost, but it is far cheaper than correcting a confident automated mistake after it reaches a customer.
5. Security and Data Protection
If the workflow handles personal data, UK data-protection responsibilities do not disappear because an AI supplier is involved. The ICO's AI guidance points organisations back to the principles of UK GDPR, including purpose, fairness, transparency, minimisation and security.
The UK government's AI Cyber Security Code of Practice recommends documenting the business need, assessing risks, limiting permissions, testing the system and monitoring its behaviour. For a small business, that translates into practical questions: what data is sent, where is it stored, who can see it, what is logged and how can the workflow be stopped?
Security work adds cost, but skipping it creates a liability rather than a saving.
6. Monitoring, Training and Change
Business systems change. A form field is renamed, an API is updated or the team starts handling a new type of enquiry. A workflow that has no owner will eventually fail quietly.
Budget for alerts, a basic operating guide and somebody who knows how to respond. The team should understand what the automation does, what it does not do and when they must take over. Our guide to AI training for employees explains why adoption and clear rules matter as much as the tool.
Three Realistic Budget Levels
A Focused First Workflow
This is a good starting point for a repeated, low-risk task: route enquiries, create draft CRM records, prepare an internal summary or build a weekly report from agreed data.
A planning budget of around £499 to £1,000 plus VAT and software is reasonable within our current offer when the systems are common, the rules are clear and a person can review the result. It should have one owner and one measurable outcome.
A Connected Small-Business System
This may join forms, inboxes, a CRM, documents and follow-up tasks across several workflows. It needs stronger error handling, permissions, training and reporting.
Our current Growth package starts at £1,500 plus VAT, but the final quote depends on the integrations and data. This level suits a business that has already proved one use case and wants to remove several hand-offs without rebuilding its whole operation.
A Custom or Business-Critical Build
When automation touches multiple departments, unusual software, sensitive information or high volumes, the project starts to resemble a software implementation. Discovery, test environments, custom APIs, audit logs and formal support all become more important.
Our Enterprise automation work starts at £5,000 plus VAT. A genuinely bespoke AI product may be better scoped through custom AI solutions, where discovery and ownership need a separate discussion.
Work Out the Value Before Approving the Cost
Start with the current process. Record how often it happens, how long it takes, who does it and what goes wrong. Then use a conservative formula:
Annual time value = hours saved each week × realistic employment cost per hour × working weeks.
For example, suppose a workflow could save six hours a week and the fully loaded cost of that time is £25 an hour. Across 48 working weeks, the theoretical time value is £7,200. Do not assume all of that becomes cash in the bank. Some saved time will be absorbed by other work. You may decide to count only half, giving a cautious value of £3,600.
Now compare that with the setup, software, review time and support costs. Add other measurable benefits, such as responding to leads sooner or reducing missing CRM records, but do not invent a revenue number unless you can track it.
Our guide to measuring AI return on investment provides a fuller framework. The point is to make a decision with assumptions you can revisit, not to manufacture a dramatic percentage.
Costs That Are Easy to Miss
- Cleaning the input data. AI cannot make inconsistent records reliable by itself.
- Changing the process. Staff still need to agree who owns approvals and exceptions.
- Usage growth. A workflow may cost more as volume and document size increase.
- Duplicate software. New automation subscriptions can overlap with features you already pay for.
- Supplier changes. APIs, prices and product limits will change.
- No monitoring. A cheap build becomes expensive when nobody notices it has stopped.
- Automation of the wrong task. Saving seconds on a low-value job will not justify weeks of work.
What to Automate First
Choose something frequent, clear and easy to check. Good first candidates include enquiry triage, internal summaries, document field extraction, CRM housekeeping and regular reporting. Avoid final pricing decisions, sensitive complaints and high-stakes customer communication as a first project.
Our guide to what UK small businesses should automate first can help you shortlist the right process. If the choice is still unclear, an AI strategy session is cheaper than building the wrong thing.
Frequently Asked Questions
How much does a simple AI automation cost in the UK?
A tightly defined workflow can start from £499 plus VAT with MattDarm, before any third-party software or AI usage charges. The final figure depends on the systems being connected, the quality of the existing process, testing requirements and the support needed after launch.
What ongoing costs should I expect?
Allow for automation-platform fees, AI model or API usage, any connected software subscriptions, monitoring and maintenance. Our monitoring plans currently start from £49 per month plus VAT, but a busy or business-critical workflow may need a larger support allowance.
Is AI automation cheaper than ordinary automation?
Not automatically. If a task follows fixed rules, ordinary automation may be cheaper and more reliable. AI is useful when the workflow needs to read, classify, summarise or draft from less structured information. A good build uses AI only where it adds value.
How do I calculate whether automation is worth it?
Measure the time and errors in the current process, apply a realistic hourly cost, then compare the annual value with setup and ongoing costs. Include human review time and use conservative assumptions rather than treating every automated minute as cash saved.
What should a quote for AI automation include?
It should define the workflow, integrations, assumptions, human approval points, testing, security, monitoring, training, ownership, third-party fees and post-launch support. If the quote only describes a clever demo, the real operating cost is still unknown.
The Bottom Line
AI automation can be affordable for a small business when the first problem is narrow and measurable. It becomes expensive when the scope is vague, the data is untidy or several systems must behave reliably together.
Start with one repeated task. Price the whole operating model, not only the build. Keep people responsible for important decisions and measure the result after launch.
If you want an honest scope before committing, get in touch. We can map the process, show where ordinary automation is enough and quote an AI-assisted build only where it earns its place.
