How to Use AI to Decide on a Second Location
Make it argue against the opening. You already have the case for it. What you need is the assumptions your plan depends on, ranked by what they cost.
Use AI to decide on a second location by making it argue against the opening. You already have the case for it, which is why you are researching this. What you are missing is a disciplined version of the case against, built from your own numbers rather than your optimism, and that is a job a model does well because it has no stake in the answer.
The output you want is not a recommendation. It is a list of the assumptions your plan depends on, ranked by how badly you are hurt if each one is wrong.
First, separate the three questions
"Should I open a second location" is three decisions wearing one coat, and people usually answer the easiest one and assume the rest.
Question | What it actually tests | Failure mode if skipped |
|---|
|---|---|---|
Can the first site run without me? | Operational readiness | You spend a year commuting between two struggling sites |
|---|---|---|
Is there demand at the new site? | Market evidence | You replicate a business that only worked because of its street |
Can I fund eighteen months of it? | Cash | You close a profitable first location to save a second |
Answer them in that order. A no on the first makes the other two irrelevant, and it is the one owners most often skip because it is about themselves.
Build the model's input before you ask it anything
Everything useful here depends on your own data. A model with no numbers will produce a confident, generic business plan, and that is worth nothing.
Assemble, in a single document: twelve months of revenue by month, your gross margin, your fixed costs at the current site, current staffing and what each role costs, your own hours and what you personally do that nobody else does, and the opening costs you have been quoted for the new site.
Then the honest column most people leave out: what portion of current revenue comes from repeat local customers who will not travel. That number decides whether a second site is expansion or dilution.
One fixed cost worth pinning down before you model anything is business rates on the new premises, which are set by rateable value rather than by your turnover and are easy to underestimate. The government's introduction to business rates explains how the bill is calculated and which reliefs a second property may lose you.
The prompts that actually produce something
Three passes, in order.
Pass one, assumption extraction:
Here are my current numbers and my plan for a second site. List every assumption the plan depends on. For each, state what I would need to observe to know it is wrong, and how much it costs me if it is. Do not evaluate whether the plan is good. Only extract assumptions.
Pass two, the adversarial case:
Argue against opening this second location using only the numbers I gave you. Make the strongest case you can. Where the data does not support an objection, say so rather than inventing one.
Pass three, the break-even:
Using my gross margin and the fixed cost estimate for the new site, calculate the monthly revenue the second site needs to cover its own costs. Then show what happens if it reaches 60%, 80% and 100% of that in month twelve, and what each scenario does to my total cash position.
Check the arithmetic in pass three yourself. Models are unreliable at multi-step arithmetic in a way that is easy to miss when the output is formatted confidently, and this particular calculation is the one you will act on.
Where AI is genuinely useful when you decide on a second location
Demand estimation at the new site is the part that is hard to do alone, and demand forecasting with AI gives you the mechanics. The specific second-site version is narrower:
Catchment comparison. Describe both areas in detail, residential density, footfall pattern, nearby anchors, parking, and ask what differences would most affect a business like yours. It will not know your streets, so you are supplying the observations and asking for the analysis.
Cannibalisation estimate. If the sites are close, some of the new revenue is old revenue moving. Ask for a way to estimate the overlap from your own customer postcode data, then run it.
Staffing scenarios. Two sites need a management layer you currently are. Ask it to cost three structures: you at the new site, you at the old site with a manager at the new, and managers at both.
Sensitivity, not projection. A single forecast is a fantasy with decimal places. A range with named drivers is a plan.
Treat the whole exercise the way you would estimating ROI before starting any project: the value is in the assumptions becoming explicit, not in the final number.
The site-one test, which most people fail
Before any of the above matters, run this. Stay out of the first location entirely for two consecutive weeks. No calls, no dropping in, no approving things from your phone.
Then look at what happened to revenue, complaints, stock, and staff turnover conversations. If the answer is "it was fine," you have a business rather than a job, and a second site is a real option. If the answer is "several things only I handle piled up," the constraint is not capital or demand, it is that the operating knowledge lives in your head.
That constraint has a cheaper fix than a lease. Writing the procedures down is the fix, and building SOPs with AI is the fastest route through it. Do that first and re-run the two-week test in a quarter.
The three numbers to take to your accountant
Whatever the analysis produces, the conversation with your accountant goes better if you arrive with three specific figures rather than a plan.
Peak cash deficit. Not total cost, but the largest gap between money out and money in at any single point, which is usually two or three months after opening. This is the number that determines whether you need financing and how much, and it is the one most owners have never calculated.
Months to break-even under your worst modelled scenario, with the assumption that drives it stated out loud. "Fourteen months if the new site reaches 60% of the first site's revenue" is a sentence an accountant can interrogate. "About a year" is not.
Effect on the first site. Both the revenue you expect to move between sites, and the cost of your attention leaving. If you currently work forty hours in the first location and will work twenty after opening, something has to cover the other twenty, and that something has a salary.
Ask the model to produce these three from your inputs, then check each by hand. They are simple enough to verify in a spreadsheet and important enough that you should not take them on trust.
What it cannot do
It cannot see the site. It does not know the landlord, the road works planned for next spring, or that the unit floods. It has no access to current local rents, and if you ask for them it may produce a figure that sounds right and is not. It cannot judge whether you personally want to run two businesses, which is a real question and not a soft one.
Use it for structure, arithmetic, and argument. Use your own eyes, a local agent, and your accountant for facts about the actual place. The broader map of what these tools do and do not do for a small operation is in where AI actually helps a small business.
Frequently asked questions
What numbers do I need before this is worth doing?
Twelve months of revenue, gross margin, fixed costs, and a real quote for fit-out and rent at the new site. Without those, any output is a template rather than an analysis.
How do I estimate cannibalisation between two locations?
If you have customer postcodes or delivery addresses, map the overlap between the two catchments and treat the shared portion as at-risk revenue rather than new revenue. Without that data, a conservative assumption is better than an optimistic one, and the model can help you build a sensitivity range rather than a single guess.
Should I trust an AI market analysis of a neighbourhood?
Only as a framework for what to go and check. Descriptions of specific local conditions may be outdated or simply wrong, and there is no reliable way to tell from the output which is which.
Is a second location always better than growing the first?
No, and it is frequently worse. Extending hours, raising prices, adding a delivery channel, or improving throughput at the existing site all carry far less fixed cost and no new lease. Test those first, because a second site is the least reversible option available.
How much cash should I hold before signing a lease?
Enough to cover the new site's full running cost for the period it takes to reach break-even in your worst modelled scenario, not your expected one. If that number is uncomfortable, that discomfort is the analysis working.
How did this land?
About the author
Senior Editor, AI & Product
Cecilia leads the Swarmz editorial desk. She has spent a decade turning complex AI and product topics into writing people actually finish, and she owns the blog's quality bar.
