How Much Should a Small Business Pay for AI Automation?

Nobody will give you a number. You ask what it costs and you get "it depends on your requirements", which is true and completely useless when you are trying to work out whether you can afford to ask.

So here are numbers.

What automation actually costs

It falls into three bands, not twenty.

A simple automated workflow: £500 to £2,000. Connecting two or three things you already use with some straightforward logic. A contact form that drops enquiries into your customer list. A follow up email that sends itself. An answer to the same three questions, with anything harder passed to you. There is little real AI in this band. It is just sensible automation, and it takes days rather than weeks.

A custom AI tool for one bottleneck: £2,500 to £8,000. This is where most small businesses should end up if they build anything at all. Something that actually reads and writes. It reads an enquiry and replies with a proper answer or a quote. It reads a supplier invoice and updates your accounts. It builds a quote for your standard jobs from the details you already collect. Two to four weeks to build properly.

A connected system across the business: £10,000 and up. Customer records, bookings, accounts, website and diary all working together. It is rare for a small business to need this first. If you are growing fast with a team of ten or more it might be right. Most should start in the band above and prove one thing works.

Running costs. For a simple workflow, usually tens of pounds a month. For something that uses AI to read and write, anywhere from tens of pounds to a few hundred a month depending on how much it handles. You should see that figure before you agree to anything.

If somebody quotes you fifteen thousand for one automated email, you are being charged for their office.

What makes the price move

How messy the starting point is. If your customer details live in three places and one of them is your head, most of the cost is untangling that, not building anything. This is the single biggest hidden factor.

Whether it has to talk to your existing software. Some systems open up easily. Others fight you. An accounting package with a proper connection is an afternoon. One without is a fortnight. Every extra system it has to connect to adds time, so the fewer the better for a first build.

How many paths it has to handle. One trigger and one action is quick. An email that has to be sorted into a quote request, a complaint or a booking, each handled differently and checked before it goes out, takes far longer to build and test properly.

How wrong it is allowed to be. Something drafting an internal note can be wrong occasionally. Something sending prices to customers cannot. Getting from mostly right to reliably right is where the real work sits, and it is worth paying for.

Whether you need it explained. Handing over something you can actually run yourself takes longer than handing over a black box. It is also the difference between owning a tool and renting a dependency.

The three ways small businesses get overcharged

The monthly fee that never ends. You pay a retainer, work happens for two months, then it quietly stops but the fee does not. I have seen this more times than I can count. A retainer is only fair when new work arrives every month, you can see what changed, and you can leave on short notice. If you cannot point at what it bought you last month, stop paying it. It is also why I think you should ask much harder questions before paying anybody.

The strategy phase. Six weeks and several thousand pounds to produce a document describing what could be built. For a business your size, the strategy is a conversation and a page of notes. Anyone selling you a discovery phase for a job this size is selling billable hours.

The platform you do not need. Being sold an enterprise system with a per user licence when the actual job is one process that runs eleven times a day. Big tools are priced for big companies. You are not one, and that is a good thing.

How to work out your own budget

Take the job you want fixed. Work out roughly how many hours a week it costs you. Multiply by what an hour of your time is genuinely worth, then by fifty.

That is what the problem costs you a year. If the fix costs less than a few months of that, it is worth doing. If it costs more than a year of it, it is not, whatever anyone tells you.

That maths is also how you spot the jobs worth doing first, which I went into properly in what to automate first in a small business.

What I would tell you if you rang me

Start with one thing. The smallest job that annoys you most. Get it working, live with it for a month, and see whether it actually gave you the time back.

If it did, do the next one. If it did not, you have learned something for a few hundred pounds rather than several thousand.

The businesses that get burned are almost always the ones who bought everything at once because someone convinced them they were behind. You are not behind. Most of your competitors are still doing it all by hand and complaining about it.

What I charge

Every price is fixed and on the page, because a range this wide is useless to you otherwise.

The AI Audit, £297. I look at how your business actually runs, find the bottleneck worth fixing first, and give you a written roadmap with three prioritised ideas and rough prices against each. If you do not get three ideas you could genuinely act on, you pay nothing. You decide afterwards whether to build anything, with me or anyone else.

The Build, £2,500 to £8,000. One tool that fixes the bottleneck, scoped in writing before any money moves. If the job turns out to need more than we agreed, I tell you before I do it and you decide.

The Scale Retainer, from £800 a month. Ongoing new builds as the business grows, looking after what is already built, and a monthly note of what changed. Thirty days notice, and I will tell you when you have stopped needing it.

Full details are on work with me.

When to keep your money

Do not spend anything on this if the business is in survival mode, if cashflow is the problem, or if the process you want to automate changes completely every month. Automating a broken process just gets you the wrong result faster. Fix the business first.

And expect the first version to need tuning. Something that handles most cases well and passes the rest to a person is a good result, not a failure.

If you want an honest figure for your situation rather than a range, get in touch and I will tell you what it would actually take.