SMB leaders often assume AI is enterprise-priced. That was true two years ago. It's not true today.
Direct costs are falling fast. A team of ten can get capable AI tools for $200-600/month. The cost of a given level of AI performance has been falling roughly tenfold a year (a16z, Epoch AI), and it continues to fall.
The real cost is implementation. Making AI useful requires workflow design, prompt engineering, system integration, and training. For most SMBs, the first-year cost of setup, integration and training is well above the software subscription itself. After that, costs drop dramatically because the workflows are built.
Token economics. A typical business email costs $0.01-0.05 to process. A long document analysis might cost $0.50-2.00. Where costs can spike is in agent workflows that make 20-50 sequential API calls. Even then, you're usually talking dollars, not hundreds.
The cost of waiting. If AI saves each employee two hours per week across twenty employees at $40/hour blended cost, that's $83,000 per year in recaptured productivity. Against a first-year investment of $15,000-30,000, the math isn't close.