Running Three Businesses With AI: An Operator's Playbook
How do solo operators use AI to run multiple businesses?
The operator playbook is to keep showing up for high-leverage moments (sales calls, strategy) and let AI plus assistants handle everything in between.
- Meeting capture: AI notes from Zoom calls are routed directly to assistants, who then build estimates from those notes — no follow-up email needed from the founder.
- Daily operations: AI handles sending emails and internal communication alongside accounting tasks.
- Founder role shrinks to the call: "I'm just showing up for all the calls, but everything else is done by my assistant."
The same operator runs an outsourced bookkeeping firm for U.S. SMEs, a franchisee network, and an education arm teaching aspiring entrepreneurs to become franchisees — a workload only possible because AI absorbs the connective tissue between meetings and execution.
What AI workflows help with territory and market expansion?
Territory and market research used to be a manual slog. A franchise candidate asking about buying Atlanta meant pulling zip codes, checking population data, and stitching it together by hand.
With ChatGPT, that workflow collapses into a back-and-forth conversation that returns:
- All zip codes in the requested territory
- Territory divisions and their names
- Exact population figures
- Advantages of each territory
The output goes straight to the candidate, who picks which territory to buy. As the operator put it: "Even a year ago, the technology wasn't available to do it." The lesson for expansion-minded operators: research that previously gated a deal can now be produced in a single session.
“I'm just showing up for all the calls, but everything else is done by my assistant”
Where should small business owners NOT use AI?
Not every workflow is a fit. The same bookkeeping operator hired an AI company to build a robot on top of QuickBooks to automate processes — and it failed for their client base.
- The robot worked for bigger clients but hit capacity limits with small ones.
- Each additional robot meant another per-use fee, breaking the economics.
- Humans were more productive and cheaper than scaling more robots.
The broader principle from operators in hardware and finance: AI being right 86% of the time is fine for some workflows and lethal for others. For accounting, taxes, and financial modeling, "you can't really do mistakes... it has to be the truth." For more on where to draw that line, see why tax strategy is the last thing you should hand to AI.
“Even a year ago, the technology wasn't available to do it.”
How should operators pick their first AI workflow?
The Google GTM playbook: start with low-hanging fruit that isn't expensive. Define the use case, then immediately ask how you'd evaluate the output at scale.
- If evaluation today depends on one domain expert eyeballing answers, that won't scale — your project is now "build the AI system plus replicate your evaluation person."
- Start by playing with an off-the-shelf model before committing to custom builds or fine-tuning.
- Almost anything you do on a computer could be automated — the decision is whether it's worth automating.
Small teams can punch above their weight: even operators with weak dev backgrounds report running roughly 25 workflows simultaneously across ventures, calling in a technical partner only when something breaks.
“I had bigger hopes for AI, but you know, took me time and took the team a lot of experimenting to realize that AI is not messiah that can help you with everything”
How should small business owners think about research with AI?
AI shifts research from a gated, expert-driven task to a conversational one — but the critical-thinking layer stays human.
Operators who came up doing school presentations with Wikipedia already learned to question sources. With AI, that discipline matters more, not less: "Question everything, do the research" — those critical skills stay, just at a much larger scale.
Practically, this means treating AI as a first-draft research partner: get the territory analysis, the market breakdown, the cultural context in minutes — then verify before acting. The same operator who automated territory research still hands the final pick to the franchise candidate.
“I'm able to run sort of 25 workflows simultaneously at the moment that work quite well across our different ventures”
Frequently asked questions.
- How does a solo operator use AI across multiple companies?
- By automating the connective tissue between meetings and execution. AI notes from Zoom calls flow to assistants who build estimates and follow-ups. Day-to-day communication and email drafting are AI-assisted. The founder's role shrinks to showing up for calls while the system handles the rest — enabling one person to run a bookkeeping firm, a franchise network, and an education arm simultaneously.
- Can AI replace bookkeepers for small businesses?
- Not reliably today. One operator hired an AI firm to build a QuickBooks automation robot; it worked for bigger clients but hit capacity limits on smaller ones, and per-use fees for additional robots broke the economics. Humans turned out to be more productive and cheaper. Meanwhile, QuickBooks itself — which holds 90%+ of the U.S. small business bookkeeping market — is adding AI directly into its product.
- How fast can AI compile market or territory research?
- What used to take hours of manual zip-code lookups and population checks now happens in a single ChatGPT session. The operator describes getting all zip codes, territory divisions, names, exact populations, and a list of advantages per territory in one back-and-forth — then handing the output to the franchise candidate to choose. The same workflow was not technically possible a year prior.
- Where should operators NOT use AI?
- In domains where being right 86% of the time can kill you — hardware tolerances, financial modeling, accounting, and taxes. As one CEO put it, the output has to be the truth, and the 0.1% error must be caught before it reaches the customer. Operators should run AI on stochastic-tolerant workflows first and keep humans in the loop for anything where mistakes compound.
- How do non-technical founders pick their first AI workflow?
- Start with low-hanging fruit that isn't expensive. Define the use case, then ask how you'd evaluate outputs at scale — if evaluation depends on one expert eyeballing results, you also have to replicate that evaluator. Play with an off-the-shelf model before committing to custom builds. Almost any computer task can be automated; the real decision is whether it's worth the effort.
- How many AI workflows can one operator realistically manage?
- More than most expect. One GTM advisor with a self-described weak dev background reports running around 25 workflows simultaneously across different ventures, pulling in a technical partner only when something breaks. The pattern only works because modern automation tooling no longer requires deep engineering skill, and because the operator focuses on strategy while builders handle observability and maintenance.