What AI Automation Really Costs a Small Business

Here is the honest answer to "what does AI automation for small business cost": you can start free, most owners running it themselves land around £30 to £60 a month all in, and paying someone to build and run it for you starts around £400 to £1,200 a month plus a setup fee. The tool is cheap. The thing that costs is either your time or someone else's.

Below are the real, current numbers so you can budget properly and spot where the money quietly leaks.

The three things you actually pay for

Every AI automation setup has three cost buckets. Miss one and your budget is wrong.

• The automation platform: the tool that connects your apps and runs the workflow (Zapier, Make, or n8n).
• The AI assistant: the brain that writes, replies, or classifies (ChatGPT, Claude, or Gemini).
• The labour: either your hours doing it yourself, or a fee to someone who builds and maintains it.

Most "AI automation costs a fortune" horror stories come from ignoring bucket three, then being surprised when the tool that cost £15 needed forty hours of fiddling.

The automation platform: free to about £40 a month

This is the engine. All three main options have a free tier and cheap entry plans, but they count usage very differently, which is where bills go wrong.

• Zapier is the easiest to learn. Its free plan includes two-step Zaps and 100 tasks each month, which works for simple workflows and light usage. Paid plans start higher. Starter and Professional both include multi-step Zaps, the main differences are task limits (750 vs 2,000) and access to premium apps. Expect roughly $19.99 to $49 a month once you are past the free tier.
• Make is the value pick. Free is $0 with 1,000 credits a month and 2 active scenarios; Core is around $12 a month, Pro around $21, Teams around $38, all annual pricing.
• n8n is the power option, cheapest if you are technical. The paid cloud tiers are Starter at $20/mo and Pro at $50/mo, there is a free self-hosted Community Edition, and Enterprise is custom-priced.

The billing difference matters more than the sticker price. The expense comes from action steps, anything that actually writes to, reads from, or pings a third-party app. On Zapier, a 5-step Zap burns 5 tasks per run. n8n is the opposite: it charges per execution, and one execution is a single run of your entire workflow no matter how many nodes it contains. A two-step workflow and a fifty-step workflow each burn exactly one execution per run.

Plain version: for simple, low-volume jobs, start on a free tier. For anything with many steps running many times a day, Make or n8n will cost you a fraction of what Zapier does.

The AI assistant: about £16 to £20 a month

If your automation needs to write, summarise, or reply like a human, you are paying for an AI model too. The good news is the market has settled on one clear price. All three companies now converge around the same price points, roughly $20/month for a solid individual plan, and $100 to $200/month for power users.

For a UK owner that is ChatGPT Plus, Claude Pro, or Google AI Pro, all within a pound of each other. There are cheaper new tiers if £20 stings: Google AI Plus at $7.99/month is currently the cheapest paid AI subscription among the three major providers, and ChatGPT Go at $8/month is nearly identical in price.

One rule: pick one and learn it. You do not need all three.

When an automation calls the AI in the background (not you sitting in the chat window), you may pay a tiny per-use API fee instead. It is genuinely small. AI API costs run approximately $0.003 per average email processed and scale with usage.

What a real small business actually pays a month

Stop adding up sticker prices and look at a worked example.

A plumber wants missed enquiries chased automatically. A web form or missed call triggers a workflow that logs the lead, sends a text and email, and drafts a follow-up if there is no reply in 24 hours.

• Automation platform: Make Core, around $12 a month.
• AI assistant: ChatGPT Plus at $20 a month, which he uses for quotes and content anyway.
• Background AI calls inside the automation: a few pounds at most.

All in, that is roughly £30 a month, and it stops a single missed job, which is worth far more than the tool. A salon automating booking reminders and review requests lands in the same range. A small ecommerce shop syncing orders and firing off dispatch emails might push into Make Pro or a bigger Zapier tier if volume is high, so budget £40 to £70.

The pattern holds: for a solo or small team doing a handful of automations, £30 to £60 a month covers the software. The cost that dwarfs this is your time to build it.

Where the hidden costs actually hide

This is the part the pricing pages do not show you.

• Runaway usage. A badly built loop can chew through your allowance overnight. Run a five-action workflow 100 times a day over a month and that is 15,000 tasks, which exhausts a 750-task Professional plan in less than two days and puts you into a tier costing $600+/month. Set a usage cap and test on low volume first.
• Hitting a wall with no warning. n8n Cloud stops all your workflows the moment you exceed the monthly execution cap, no warning, no grace period, no overage option. Everything halts until the next billing cycle resets.
• Annual billing traps. Monthly is convenient but pricier. On Zapier, all headline prices assume annual billing, and monthly billing costs roughly 30 to 50% more. Only commit annually once a workflow has proven itself.
• Premium app gates. Some connectors only work on paid tiers, so check yours before you build. If your workflows depend on specific premium integrations, verify which apps require paid tiers during the free trial, as some may only be available on paid plans, affecting your total cost.
• Self-hosting is not free. "Free" software still needs to run somewhere. A self-hosted n8n runs a few pounds a month for a server, but you own the updates, backups, and fixing it at 9pm.

Doing it yourself versus paying someone

Here is the real fork.

Doing it yourself: your only cash cost is the software, so £30 to £60 a month. The price is learning time. A first useful automation is genuinely doable in an afternoon on Make or Zapier, and the free tiers exist so you can prove it works before paying a penny. This is the right route for one or two clear, simple automations.

Paying someone: you buy speed and someone else owning the headache. The numbers are much bigger and vary wildly. At the affordable end, most single-system engagements land between $500 and $1,500 per month. Setup is usually separate: most agencies charge a build fee for initial workflow development, separate from the ongoing retainer, and a typical setup fee runs $2,000 to $12,000 depending on scope.

Two things protect you when paying someone. First, insist you own it: some agencies build on proprietary platforms and charge platform fees, and when you exit your automations stop working because they only run inside the agency's tool, so ask directly whether you can take the automations with you. Second, ignore false urgency: "this rate is only valid for 48 hours" has no legitimate basis in professional services, and reputable agencies do not manufacture urgency.

Simple rule: if it is one or two clear workflows, build it yourself. If it is call handling, several connected systems, or something that must never break, pay a professional for the build and keep it on tools you own.

What to do next

1. Pick your one automation this week. Write down a single repetitive job that costs you time or loses you money (missed enquiries, review chasing, order emails). One only. 2. Open a free account on Make and build that one workflow on the free tier. Connect your form or inbox, add the steps, and run it on a handful of real cases before you pay anything. 3. Set a spend limit and test on low volume before you switch anything to run automatically. Watch your task or credit count for a week so a runaway loop cannot surprise you.

If you would rather work through your first automation alongside other owners doing the same thing, that is exactly what we do inside the AIBL Club. Either way, start with one workflow, prove it saves you hours, then build the next.

Keep reading