How a two-person business can use AI without hiring a consultant
Small businesses don't need an "AI strategy." They need Monday mornings that take two hours instead of five. Here are six workflows we've seen actually stick in teams of one to five — each with the specific tool, the setup cost, and the honest failure mode to watch for. Total budget: roughly $40–70/month, no developer required.
1. Support email triage and drafting
The workflow: every incoming support email gets classified (question / complaint / refund / spam) and a draft answer written in your voice from your own docs. You edit and send; you never start from blank.
The setup: Zapier AI watching your inbox → Claude or ChatGPT drafting with your FAQ pasted as context → draft lands in Gmail as a reply-not-sent. One afternoon to build.
Watch for: refunds and angry customers always go to a human first. Automate the "what are your hours" tier, not the "this went wrong" tier.
2. Proposals and quotes in 20 minutes
The workflow: paste a client enquiry into your assistant with your past winning proposal and price list; get a tailored draft; drop it into a Gamma or Canva template. The tool doesn't win the deal — responding same-day does.
Watch for: always re-check quoted numbers against your actual price list. Models are unreliable at arithmetic with money; treat pricing as copy-paste, never generated.
3. One content engine, five outputs
The workflow: record one 30-minute voice memo about something you know (walk, drive, dishes). Transcribe with Otter.ai → rewrite into a newsletter issue, a LinkedIn post, three short-video scripts and a help-center article, each using the example-based prompting habit. One input becomes a week's content because expertise is the bottleneck, not typing.
Watch for: publish with your voice edited back in. Readers can smell unedited AI prose, and "sounds automated" costs trust.
4. Meetings that produce work, not notes
The workflow: record client calls (with consent) → Claude gets the transcript plus the standing instruction "list decisions, owners, deadlines, contradictions" → action list pastes into your task tracker. The killer feature isn't summarising; it's catching the contradiction between what two people said.
5. Bookkeeping prep, not bookkeeping
The workflow: export bank CSV → assistant classifies transactions against your chart of accounts and flags only the ambiguous ones → your accountant gets a clean file. This saves ~4 hours of your month and, because it's prep, your accountant stays the reviewer. AI drafts; humans stay liable.
Watch for: don't upload documents with bank credentials or client-confidential data to consumer AI accounts; check your provider's training opt-out first.
6. Weekly competitor and price check
The workflow: a recurring research agent (ChatGPT Tasks or Gemini, both included in paid plans) visits five competitor pages every Monday, diffs pricing and offers against last week's table, and emails you the change list. Thirty seconds of reading replaces an hour you always skipped.
Watch for: agents can misread a page. Anything that triggers a decision gets a manual click-through of the source.
The order to build these in
- Start with whichever workflow eats most of your week (for most teams: support email or content).
- Run it AI-drafts/human-sends for two weeks and fix the prompts, not the process.
- Automate one more step only when the drafts need under five minutes of editing.
- Never automate payments, legal answers, tax filings or anything you can't undo.
Tool details for every step are in the AI tool directory, and the free-tier versions are in 15 genuinely free AI tools.