Everyone's selling you an AI assistant right now. And if you run a small business with more revenue than time, you've probably gotten at least three vendor pitches this quarter for some variant of "AI for [your industry]."
Most of the pitches are wrong. Not because AI is fake — it's genuinely useful — but because the AI you're being sold is being positioned as a replacement for operational discipline, when what you actually need is more operational discipline.
The 80/20 rule still applies
Walk into most small businesses and you'll find the same three problems, in this order:
- Two tools that should talk to each other, don't. Shopify and QuickBooks. HubSpot and Gmail. Trello and your billing system.
- A process that exists in someone's head and breaks whenever that person is out sick or on vacation.
- A spreadsheet that's become load-bearing — the business can't function without it, but it's held together by one person's memory.
None of these are AI problems. They're plumbing problems. The leverage isn't in adding intelligence — it's in connecting the systems you already pay for so that the data flows correctly, the team isn't carrying invisible work in their heads, and the business runs whether or not anyone is watching it. That's where the time and money savings actually live.
What AI is actually good at (and not)
AI is good at:
- Summarizing unstructured text (support tickets, meeting transcripts, customer reviews)
- Drafting first-pass content (email responses, job descriptions, product copy)
- Classifying messy inputs (which of 40 categories does this support ticket belong to?)
AI is not reliably good at:
- Making a decision that commits money, inventory, or contracts
- Working with numbers that need to be exactly right
- Anything where you'd want a human to double-check the output anyway
Notice the pattern: AI is great when "approximately right, quickly" beats "exactly right, slowly." That's a narrow slice of most small businesses.
When AI does make sense for a small business
You should consider an AI investment when all three of these are true:
- You have a specific, repetitive task that consumes >5 hours of a human's time per week.
- That task is mostly about processing unstructured text (not numbers).
- "90% accuracy with human review" is acceptable for the output.
If that's you — great, it might be a 4-week project and a real win. If you're not sure — skip the AI pitch and spend those dollars on fixing your plumbing first.
What we'd recommend instead
Before hiring anyone (us or otherwise) to build AI for your business, ask yourself:
- What's my "master" spreadsheet? Replace it with a real workflow and you'll save more time than any AI could.
- What are my tool subscriptions? If two of them should be talking and aren't, connect them. That's usually a $2K–$5K project that pays back in months.
- What's the one process that breaks when Alex is on vacation? Document it. Automate what can be automated. Hire someone for the rest.
Every time we talk to a small-business owner excited about AI, we ask these three questions first. Most of the time, the honest answer is: we haven't done the boring work yet.
The boring work is where the money is.