A household budget question usually arrives as a single sentence that hides a dozen unknowns: “What happens if I go down to four days a week?” “Can we manage if the rent goes up 150 a month?” “What if we have the baby next spring instead of saving another year?” The honest answer requires working through several versions of the same spreadsheet, changing one assumption at a time, and seeing where the numbers get uncomfortable. Most people never do this, because building even a simple scenario model by hand is tedious enough to skip.
This is exactly the kind of task a model is good at: taking numbers you already know to be true and doing transparent, checkable arithmetic across several versions of the future. It is not a task a model should turn into advice. The model does not know your risk tolerance, your family’s other resources, your country’s tax rules, or what a “reasonable” emergency fund looks like for your situation — and it has no business recommending an investment, a debt payoff order, an insurance product, or a pension contribution. Its job here stops at showing you what a change in one number does to the rest.
Do not ask a chatbot to recommend an investment, a debt strategy, an insurance product, a pension contribution, or how to allocate savings. A model can show you the arithmetic consequences of an assumption you supply, but it has no license, no view of your full financial picture, and no accountability for the outcome. Any decision beyond “what does this look like on paper” belongs with a qualified, licensed financial adviser in your country.
Step 1: Build a table of verified figures, not impressions
Before any scenario work, list your actual recurring numbers with a source and a date for each one — the current lease agreement, the last three payslips, the utility provider’s rate page, the childcare center’s fee schedule. Mark each line as fixed (the same every month, like rent) or variable (groceries, fuel, discretionary spending), because the two behave very differently in a what-if scenario.
housing: 1,050/month, fixed, source = lease dated 2026-01-15
income (combined): 3,400/month, fixed, source = last 3 payslips
groceries: avg 620/month, variable, source = 3-month bank export
transport: 210/month, variable, source = 3-month bank export
childcare: 480/month, fixed, source = center invoice dated 2026-06-01
subscriptions: 95/month, fixed, source = subscription-drift-review
(see /articles/subscription-drift-review)
This table, not the scenario output, is the part worth trusting completely — because you built it from documents, not from a model’s guess about “typical” household costs. Never let a model fill in a number you have not verified; if you do not know a figure, mark it “unknown” rather than accepting a plausible-sounding estimate.
Step 2: Turn one change into a scenario, in plain arithmetic
With the verified table in hand, ask for one change to be modeled at a time.
Here is our verified monthly household table: [paste table with
source and date columns]. Model one scenario: combined income drops
to 2,720/month (a reduction to four days a week for both earners),
starting next month. Keep every other figure exactly as given.
Show: new monthly total, new monthly surplus or shortfall, and — only
if the result is a shortfall — how many months our current savings of
[X] would cover it at this new rate. Do not suggest what to cut or
where the money should come from — just the arithmetic.
Illustrative shape:
New income: 2,720/month (was 3,400)
Fixed costs unchanged: 1,625/month (housing 1,050 + childcare 480 + subscriptions 95)
Variable costs unchanged: 830/month (groceries 620 + transport 210)
New monthly surplus: +265/month (was +945 surplus)
Monthly cushion shrinks by 680; under these assumptions the household is not drawing savings down
(If you need a shortfall example for planning, change only income again - for instance 2,200/month yields 2,200 - 1,625 - 830 = -255/month, and 4,200 of savings would cover that for roughly 16 months. Recalculate by hand either way.)
Run this once per real scenario you are actually facing — a reduced-hours option, a rent increase, an added expense — rather than asking for a menu of hypothetical futures. Three or four scenarios that map to real decisions in front of you are far more useful than a dozen generic ones.
A household budget table reveals income, address-linked rent, employer-visible pay changes, and sometimes childcare or medical costs. Redact account numbers and any identifying references before pasting figures into a chat tool, use a session with training turned off, and see privacy and data hygiene at work for what else to strip from exported statements before they leave your device.
Step 3: Stress-test the assumption that scares you most
Every household has one number that, if it moved, would matter more than the others — a variable rate, a single income the household depends on, a benefit that could change. Ask the model to show you the sensitivity of the whole scenario to that one figure, holding everything else fixed.
Using the same verified table [paste], show what our monthly total
looks like at three versions of rent only: current (1,050), +100,
and +250 — holding every other figure exactly as given. Do not
recommend whether to move, renegotiate, or take on debt; just the
three totals and the resulting surplus or shortfall at each.
Sensitivity checks like this turn a vague worry (“what if rent goes up”) into a specific number you can actually plan around, and they make it obvious which assumptions are load-bearing for the household’s plan and which barely move the outcome.
Step 4: Flag what needs a professional, not another scenario
Once you can see the arithmetic clearly, some items on the list will not be spending questions at all — they are decisions about debt order, tax treatment, insurance coverage, or retirement contributions. Ask the model to sort these out explicitly rather than let it drift into offering an opinion on them.
From this scenario output [paste], list anything that looks like a
tax question, a debt-repayment order decision, an insurance coverage
choice, or a retirement/pension contribution decision — rather than
ordinary what-if arithmetic. Label each "needs a qualified adviser,
not this workbook." Do not attempt to answer them.
A model can describe general mechanics (“some mortgages recalculate interest monthly, others daily”) without being the source of truth for your contract, your country’s rules, or your specific numbers. The workbook’s job is to make the flagged list short and specific enough that a real conversation with an adviser takes twenty minutes instead of two hours.
Common pitfalls
- Letting the model estimate a figure you have not verified. A plausible-sounding “typical grocery bill for a family of four” is not your grocery bill. Every input should trace back to a document or an export.
- Asking for a recommendation instead of arithmetic. “What should we do?” invites an opinion with no accountability behind it. “What does this look like if X happens?” keeps the model in its lane.
- Changing more than one variable at a time. A scenario that moves income, rent, and childcare costs simultaneously is impossible to learn from — you cannot tell which change drove the result.
- Treating a “surplus” scenario as safe. A model showing a positive monthly number after one assumption change says nothing about one-off costs, seasonal variation, or a second bad month landing on top of the first.
Validation and fallback
Recalculate at least one scenario by hand or in your own spreadsheet before trusting it for a real decision — a model can silently misadd a column or apply a percentage to the wrong base, especially across several chained what-ifs. If your manual total does not match, do not ask the same model to find its own error; start the calculation over with a clean, explicit prompt, or move it into a spreadsheet formula you can audit line by line.
Build your workbook this week
Pull your last three payslips, your lease, and your utility statements, fill in the verified-figures table, and run the scenario and sensitivity prompts above using the household scenario workbook. If your household is heading into a joint conversation about what the numbers mean, pair this with preparing for a shared finances meeting; if the workbook itself is the bottleneck, AI for spreadsheets covers formulas and pivot tables for the underlying file.
