The cost of AI automation depends on the scope of what you're building, so I quote every project on its full scope before any work starts. There are two separate costs to think about: building the automation, which you pay once, and running it, which you pay every month. For a lot of small-business automations the running cost is small, and sometimes it's close to nothing.
I don't publish prices, and this post won't give you a number. What it will do is explain what makes a project bigger or smaller, how long it takes and what you'll pay after launch, using a real client project as the example. It's part of my guide to AI automation for small business.
Why I don't publish a price list
Almost everything I build is custom software for one business, made to fit the way that business already works. Two projects that both get called "an AI automation" can be completely different jobs. One might be a small script that moves data between two tools, and the other an AI agent connected to a dozen systems with a dashboard for the team.
That's why the price ranges you find when you search this question are so wide. Any number I put on this page would be wrong for most people reading it.
How I quote a project
It starts with a conversation, on whatever you already use, whether that's a call, WhatsApp, Discord or email. You tell me what you want automated. I explain how it could work technically and what the options are. As we talk it through, the project gets narrower and clearer, and by the end I have a good idea of what it'll take.
Then I quote the whole project, with a price and a timeline, and we agree on both before I start. You know the full cost upfront, so there are no surprises halfway through.
If you're not sure yet what to automate, we can work that out first. I'll have your team track their work for a week so we can see where the time actually goes. I explain that process in what to automate first.
What makes a project bigger or smaller
A few things decide most of the scope.
The first is how many other tools and systems the automation has to connect to. An automation that reads from one spreadsheet is a much smaller job than a dashboard pulling from a dozen different services.
The second is how much AI is involved. A workflow that's mostly plain code is quicker to build and cheaper to run than an AI agent that juggles many tools or answers questions from a knowledge base. I explain the difference in AI agents vs workflows.
The third is whether your team needs screens to work with, like a dashboard or an internal app. A background automation that nobody sees is less to build than one people log into every day.
And the last is how clear the process is. If nobody has written down how the task is done, part of the project is figuring that out.
How long it takes
Most of my projects are finished in a couple of weeks. Bigger builds take longer, and we agree on the timeline together with the price.
Waiting on the client is rarely what slows things down. When I need answers or access to an account, my clients usually get it to me the same day or the next.
Running costs after launch
Once the automation is live, there are a few ongoing costs to expect.
If the automation uses AI, you pay for model usage, which is billed by how much text the model reads and writes. For most small-business tasks you don't need the most expensive models. Cheap open-source models handle things like sorting messages, pulling details out of documents, writing short summaries and answering simple questions, and millions of tokens (the small chunks of text models are billed by) can cost a few dollars. Several providers even offer free tiers of these models, and a small company with a few automations may never need more than the free tier.
When an automation does need a mix of easy and hard requests, I route them: a small, cheap model decides how hard each request is and sends it to a cheap model or a stronger one. By my own estimate, that can cut AI model costs by up to 80% compared with sending everything to a top model. I explain how in cut AI agent costs by 80%.
You also pay for hosting. Internal tools that only your team uses get very little traffic, and they can often run on a free hosting tier.
On top of that come subscriptions to any third-party tools the automation relies on, and a retainer if you want one. The retainer is optional. It covers fixes and new features after launch, and most of my clients skip it.
Why plain code costs less to run than an agent
An AI agent pays for AI model usage on every step of every run. A workflow built in plain code pays almost nothing per run, however many times it runs.
I still use AI heavily when I build. An AI coding tool writes the automation's code once, and that's a one-time cost. After that the code runs thousands of times without spending anything on AI, and AI is only called for the few steps that need real judgment. By my estimate, building this way cuts AI model spending by about 90% compared with having an agent do the whole job. I go through the reasoning in AI agents are overrated.
It's also cheaper to maintain. Plain code doesn't drift, so there are fewer surprises to pay someone to fix.
A worked example: Mastersealer
Mastersealer is a landscaping business in Vaughan, Canada. Johnny, who runs it, was writing every quotation by hand and emailing each one as a PDF himself, which was eating up a big part of his week.
I built them a quotation app. Johnny fills in a short form, and the app turns it into a branded PDF and sends it to the customer once he's checked it. It took me somewhere between 10 and 20 hours to build.
Here's what it costs them to run:
| Cost | What Mastersealer pays |
|---|---|
| AI model usage | Nothing. The app is plain code with no AI in it. |
| Hosting | Nothing. It runs on a free hosting tier. |
| Maintenance | A minor fix about once every six months. |
The hosting line used to be a real cost. Their internal app had been running on a server that was switched on all year, even though the business is seasonal and the app is only used at certain times of day. Moving it to a free tier saved them $500 a year.
What they got back for it: more than 10 hours a week saved on quotes and invoicing. The full story is in the Mastersealer case study.
How to keep the cost down
Start with one task, the one that eats the most time, and get that working before you add anything else. You don't need to automate all of it either. Automating 80% of a 10-hour-a-week task gives you 8 hours back, and the last 20% is often the part that needs a person anyway.
Write the process down before the build starts, so less of the project is spent working out what you actually do. Prefer plain code over agents wherever the steps are known. And keep internal tools on free tiers where the traffic allows it. I try to do all of this for every client, because running costs are something you pay every month for as long as the automation lives.
If you want to know what your project would cost, the only way to get a real number is to talk it through. Here's how I work, and if reliability is your worry, read how to make AI automations reliable.
