What AI Can Actually Do for a Ten-Person Business

If you run a small business, the AI advice you get comes in two flavours: breathless (“adopt AI or die”) and dismissive (“it’s all hype”). Both are useless, because neither tells you what to do on Tuesday.

Here is the practical version — what AI is genuinely good at in a business of roughly ten people, what it is bad at, and the one policy you should write before anyone on your team uses it with client information.

Where it actually saves hours

The honest pattern: AI is a strong first-draft machine and a patient summarizer, and most small businesses are full of first drafts and unread documents.

• Drafting that starts from something. Quotes built from your past quotes. Follow-up emails. Job postings. Policy documents you have been putting off. The machine writes the draft; you spend five minutes correcting instead of forty minutes staring at a blank page.
• Summarizing what you do not have time to read. A 30-page contract, a long email thread, meeting notes, a competitor’s website. Ask specific questions of a long document instead of reading all of it.
• Intake triage. Turning a rambling enquiry into a structured summary — what they want, when, budget signals — before a human reads it.
• Repetitive formatting. Invoices to a standard shape, data moved between formats, product descriptions from a spec sheet. Boring transformations, done reliably. (If you want named tools with dollar figures, we keep a separate tool guide for Atlantic Canada businesses.)
• Being findable by AI at all. A growing share of buyers now ask an AI assistant instead of searching. Whether your business appears in those answers depends on how legible your website is to those systems — that is a fixable property of your site, and most small-business sites fail it today. (You can check yours in about a minute, free.)

Where it fails

• Anything unsupervised and customer-facing. A chatbot that answers wrongly in your name is worse than no chatbot. If you would not let a first-week hire reply to clients unreviewed, do not let a model do it.
• Judgment. Pricing a tricky job, reading a difficult client, deciding whether to take the work. The model produces something confident either way; confidence is not correctness.
• Your books, unattended. Drafting and categorizing, fine — with review. Anything that files, remits, or commits money needs a human who is accountable for it.
• Knowing your business without being told. Generic tools give generic output. The useful results come when your real documents, prices, and voice are in front of the model — which is exactly the setup work most businesses skip.

The policy to write before any of this

One page, this week: what your team may and may not paste into AI tools.

Client names, health information, financials, anything covered by privacy law (in Canada, PIPEDA) — not into consumer chatbot accounts, ever. The free tier of a chatbot is not a place of business. If AI tools are worth using on client work, they are worth configuring properly: business accounts with training turned off, or systems that keep the data on infrastructure you control.

This is the likeliest gap in any small business today: not a missing tool, but staff already using AI informally, with no rules, on data that should never have left the building.

How to start without a leap of faith

Pick one repetitive workflow that eats hours weekly. Set it up properly once — the right tool, your real documents, a written guardrail — and measure whether it holds up for a month. Then decide about the second one.

If you would rather have that done for you, that is exactly the shape of our AI readiness sprint: an audit of where you stand (including how your site looks to AI assistants), two or three quick wins implemented rather than recommended, and a written roadmap in plain language. And if AI genuinely cannot help your business yet, the roadmap says so — a no is cheaper than a wrong yes.

Or start with the free version: run the scanner on your own website, and tell us what you are working on — we reply in writing within one business day.