The most common thing I hear from small business owners when AI comes up: "That's great, but I can't afford it." That assumption is almost always wrong. Here's the real math.
Where the misconception comes from
AI gets covered in the news as a technology for big companies — Microsoft, Google, Goldman Sachs. The price tags in those stories are real: enterprise AI contracts can run hundreds of thousands of dollars per year. That's for platforms, licenses, dedicated compute, and teams of engineers to maintain it.
That is not what I build, and it's not what a small business needs.
What a small business needs is a specific automation that solves a specific problem. And those cost a fraction of what the headlines suggest.
What AI actually costs to run
Most of the automations I build use one or more of the following:
| Tool | Monthly Cost | Notes |
|---|---|---|
| Claude API (Anthropic) | $5–$50/month | Depends on usage volume. Most small businesses land well under $20/month. |
| n8n (self-hosted) | $0/month | Workflow automation tool. Runs on your own machine. No subscription. |
| Ollama (local AI) | $0/month | Open-source AI models running on your own hardware. Zero API cost. |
| Google Workspace APIs | $0/month | Included with your existing Google Workspace subscription. |
A typical automation built for a small business has ongoing costs of $0–$50/month depending on usage. That's it. There's no enterprise contract, no vendor lock-in, no per-seat pricing.
The ROI math
Let's take a simple example. Say you're spending 8 hours a week on tasks that could be automated — data entry, email follow-ups, report generation. That's realistic for most small business owners.
8 hours/week saved × 50 weeks/year = 400 hours/year
× $75/hour (conservative value of owner time) = $30,000/year in recovered time
Cost to build the automation: $997–$2,497 (one-time)
Monthly ongoing cost: $20–$50
The automation pays for itself in the first two to three weeks. Every week after that is pure gain — either in time you can spend on revenue-generating work, or in capacity to take on more clients without adding headcount.
The option most businesses don't consider: run it locally
For businesses with privacy concerns or tight budgets, there's another option: local AI. Tools like Ollama let you run powerful open-source AI models directly on a Mac mini or any decent computer you already own — with zero API costs and no data leaving your building.
I run a local AI server for one of my clients this way. It handles internal Q&A, document lookups, and workflow routing with zero monthly cost beyond the electricity to keep the machine on.
It's not the right fit for every use case — cloud APIs like Claude are faster and more capable for complex tasks — but for many small business applications, local AI works perfectly and costs nothing ongoing.
What you actually need to get started
You don't need a data team. You don't need a custom platform. You don't need to learn to code. You need:
- A clear picture of what's eating your time
- Someone who can build the automation
- A willingness to spend 30 minutes on a discovery call
The budget question almost always resolves itself once we map out what you're actually trying to solve and what it's costing you to not solve it.