How Can AI Help Startups? Runway and Hiring Scenarios in One Sheet

AI can draft the structure and formulas of a startup sheet that shows monthly burn, three hiring scenarios and runway in months, from a plain-English specification. It cannot know your costs or growth. You supply every assumption, and you check every number, because a runway model is only as good as the guesses typed into it.

The model is small. A cash figure, a monthly cost list, a revenue line, a table of hires and one formula for months until cash runs out. The skill is in setting it up so the assumptions are visible and the scenarios can be compared at a glance, and AI is good at that scaffolding.

Follow me on Instagram@sagnikteaches

One sheet, four blocks

Keep everything on a single tab if you can. A founder who has to open three tabs to see what changed will stop updating it. The layout I would use:

Connect on LinkedInSagnik Bhattacharya

1. Assumptions block (columns A to D): one row per assumption, with the value, the unit, where it came from and a confidence flag (known, estimated, guess).
2. Hires table: one row per possible hire with role, start month, monthly loaded cost and one-off cost, and a switch for each scenario.
3. Monthly engine: months across the top (1 to 24), then rows for revenue, fixed costs, hire costs, net burn and closing cash.
4. Results block: runway in months for each scenario, side by side, and the month cash hits zero.

The assumptions column matters more than the formulas. It is the document you show an adviser, an investor or a co-founder. Every number that is a guess should say so on the sheet, in words, in a cell next to the number.

Subscribe on YouTube@sagnikteaches

A small software startup with $420,000 in the bank

Illustration, with invented numbers throughout. A small software startup has $420,000 in the bank. Its monthly revenue is $22,000. Its monthly costs are:

Net burn is costs minus revenue: $53,500 minus $22,000 is $31,500 a month. The simplest runway formula is cash divided by net burn: $420,000 divided by $31,500 is 13.3 months. That one-line answer treats everything as flat, which no startup ever is. The scenarios below show how far it can mislead.

Three scenarios, all with revenue growing from the same base:

• A. No hires. Revenue grows 3% a month.
• B. One engineer from month 3, $9,500 a month loaded plus a $6,000 one-off recruitment cost. The founders assume the extra product speed lifts growth to 4% a month from month 6.
• C. The engineer plus a sales hire from month 6, $8,000 a month plus a $5,000 one-off cost, with growth at 6% a month from month 9.

Running the monthly engine for each case gives these results, which I computed by simulating each month in turn:

Look at the gap between 13.3 months (the flat formula) and 16.4 (no hires, with 3% growth). Growth, even at a modest rate, buys three months. And each hire takes months away that the assumed extra growth does not win back within the runway. If you strip the extra growth out, so every case grows at 3%, the engineer case lasts about 11.7 months and the engineer-plus-sales case about 10.2. In other words, the 4% and 6% growth assumptions barely change the runway inside the first year; they matter for what the company looks like after it. That is a finding you would struggle to see in your head, and the sheet shows it in a minute.

Now the decision. Hiring the engineer costs about 4.5 months of runway in this model (16.4 minus 11.9). Whether that is wise depends on whether you can raise or reach break-even inside the remaining time, which is a judgement about your market, not a number in a cell. What the sheet does is make you state it.

Step 1: Write the specification before you open an AI tool

AI builds what you describe. A vague request ("make me a runway model") returns a generic grid. A good specification is half a page. The prompt for the example above:

Build a monthly cash runway model for 24 months in a single worksheet.

Assumptions block at the top, one row each, with columns: Assumption, Value, Unit, Source, Confidence (known/estimated/guess):
opening cash 420000; starting monthly revenue 22000; monthly revenue growth 3%; monthly fixed costs 53500.

Hires table below: Role, Start month, Monthly loaded cost, One-off cost, In scenario A (Y/N), In scenario B (Y/N), In scenario C (Y/N).
Rows: Engineer (3, 9500, 6000, N, Y, Y); Sales hire (6, 8000, 5000, N, N, Y).

Monthly engine: for each scenario, rows for revenue, fixed costs, hire costs, one-off costs, net burn, closing cash.
Revenue growth in scenarios B and C is 4% from month 6; in C it is 6% from month 9.

Results: for each scenario, the month when closing cash first goes below zero and runway in months with one decimal place.
Use cell references, not typed-in numbers, in every formula. Do not use merged cells. Explain each formula in a note column.

Every phrase there removes a way for the result to go wrong. "Use cell references, not typed-in numbers" stops hidden constants. "No merged cells" keeps the sheet sortable and easy to copy. "Explain each formula" gives you something to check against.

Step 2: Choose where the model gets built

The tools differ in how they fit into a spreadsheet, and they change month to month, so check the current state of whichever you use.

• Chat assistant with a file (ChatGPT or Claude): describe the sheet, get a file or a block of formulas to paste. ChatGPT Plus is $20 a month and Claude Pro $20 a month ($17 a month on annual billing).
• Claude for Excel: an add-in that works inside the workbook. Anthropic's pages say inputs and outputs are deleted on its side within 30 days; read the current terms.
• Copilot in Excel: a side pane that edits, plans and chats. The separate =COPILOT() worksheet function was retired on 14 September 2026; cells that still contain it return #NAME? when they recalculate, so older guides that use it no longer work.
• Gemini in Google Sheets: works from a side panel to create tables, formulas, charts and pivot tables on an eligible Google Workspace or Google AI plan. Usage limits differ by plan, so check Google's current help page.

The choice matters less than the checking. Whatever you use, ask it to write formulas you can read, and avoid anything that hides logic inside a one-off script you cannot see. The older tutorial on Claude for Excel covers that add-in in detail.

Step 3: The formulas worth understanding

You do not need to write them, but you should be able to read them. Four matter.

Revenue by month

Month 1 is the starting revenue. Each later month is the previous month multiplied by one plus the growth rate. If growth changes in month 6, the growth cell for month 6 onward points at the new rate. In a sheet, that is a row of growth rates by month, and revenue is the previous cell times (1 + that month's growth).

Hire cost by month

For each hire and each month: the monthly cost applies if the month is the start month or later, and the one-off cost applies in the start month only. In spreadsheet terms, something like =IF(month>=start, monthly, 0) + IF(month=start, oneoff, 0), multiplied by the scenario switch (1 or 0). Getting the greater-than-or-equal sign wrong delays or doubles a cost; it is the commonest mistake.

Closing cash

Opening cash minus net burn, month by month, where net burn is fixed costs plus hire costs plus one-off costs minus revenue. Closing cash of one month is opening cash of the next.

Runway in months

The first month closing cash goes below zero, less the fraction of that month you were still funded. If cash at the end of month 16 is $7,451 and the burn in month 17 is $18,196 in net terms, runway is 16 plus 7,451 divided by 18,196, about 16.4 months. A formula that finds the first negative month and interpolates handles this.

Step 4: Founder checks, every one of them

The model is arithmetic, and arithmetic can be checked by a person with a calculator. These seven checks catch nearly every error I would expect from an AI-built sheet:

1. Zero-hire test. Switch all scenario switches off and set growth to zero. Runway should equal opening cash divided by net burn, 13.3 months in the example. If not, something else is wrong.
2. Month-1 check. Add up the cost lines by hand and compare with the engine's month-1 cost, then do the same for revenue.
3. Start-month check. For one hire, look at the months either side of the start month. The one-off cost must appear once and the monthly cost from the start month on.
4. Hard-code hunt. Click into ten formulas at random and check none contain a typed number that should be a cell reference.
5. Sign check. Costs reduce cash, revenue adds to it. Look for a minus sign in the wrong place.
6. Unit check. Percentages as 3% or 0.03, never 3; monthly versus annual figures; thousands versus units.
7. One external check. Compare month-1 cash movement with last month's bank statement. If the model says you burn $31,500 and the bank shows $38,000, find out why before using it for anything.

Ask the assistant to find its own errors as well, but as a second reader rather than the first. A prompt that works: "Read these formulas and list any that would behave differently if the start month were 1, or if a scenario switch were 0." Then check what it lists.

The assumptions column, filled in

This is the table you will actually show people. Each row says what the number is, where it came from and how sure you are.

Notice the last row. A simplification is an assumption too, and writing it down means someone can challenge it. In the model, hosting costs might rise with growth; a founder who sees "none" in the column knows to look.

Turning vague plans into numbers

Founders describe hires loosely: "we'll bring someone in around Q2". AI is useful for forcing a plan into cells. Paste your notes and ask for a table.

Before (founder notes): "Probably want a designer part-time from around summer, maybe an ops person if the pilot goes well,
contractors for the launch."

After (assistant's table, to be corrected by the founder):
Role | Start month | Monthly cost | One-off | Open question
Designer, part-time | 5 | ? | ? | Hours per week and day rate?
Operations hire | 8 | ? | ? | What triggers the hire: pilot result or revenue?
Launch contractors | 4 to 6 | ? | ? | Total budget, or per month?

The question marks are the useful bit. The assistant has shown you which numbers you have not decided, and until you fill them in the model cannot answer. Convert each into a number, even a rough one, and mark it as a guess.

Sensitivity: change one assumption at a time

After the base cases, test how fragile they are. Change one assumption by a fixed amount and record the runway. For the example, with flat growth of 0%, 3% and the scenario growth ladders:

The point of such a table is to see which assumption you can least afford to be wrong about. Here, whether revenue grows at all moves the answer by about three months in the no-hire case; whether you hire moves it by four to six months. The extra growth after hiring changes the runway by only a fraction of a month inside the window. It would matter for later months, but a hire justified purely by that extra growth is a bet beyond the runway. Ask the assistant to build the sensitivity table, then check two cells of it by hand using the zero-hire test.

A mistake the model made and how it showed up

A common failure is the one-off cost counted every month. In a first draft, an assistant wrote the hire-cost formula as monthly plus one-off for every month from the start month, rather than one-off in the start month only. The symptom was immediate: runway for scenario B came out at 10.5 months instead of 11.9, with the $6,000 recruitment cost appearing in every column from month 3 onwards. The seven checks above caught it at the third: the start-month check showed $6,000 in month 4 as well as month 3.

The correction was one formula, but the lesson is that a plausible-looking result is not evidence. The only way to know is to test a case whose answer you can work out by hand.

Keeping it alive: a monthly update routine

A runway model goes stale in a month. Make the update a ten-minute habit on a fixed day after the books close.

1. Replace opening cash with the real bank balance.
2. Replace the last month's forecast with actual revenue and costs, and note the gap between forecast and actual in a column.
3. Re-estimate the growth rate from the last three to six months of actuals.
4. Move any hire whose start date has slipped, and add any new commitments.
5. Read the results block and write one sentence: "Runway is now X to Y months; the next decision is Z by this date."

If your books are in accounting software, the monthly close does most of this; AI accounting for startups: monthly close, burn and runway shows how to get the actuals out cleanly. For a weekly view of cash when things are tight, see building a 13-week cash flow forecast with AI help.

What to do with the answer

A runway figure is an input to decisions, not a decision. Three uses:

• Hiring. Compare scenarios, but also ask whether software or a contractor could do the job for less. AI or a new hire: how to decide before you recruit gives a way to compare.
• Fundraising. Work back from the cash-out month to the date you need to be in conversations, and allow for how long raising takes in your experience. AI agents for startup fundraising covers the workflow.
• Cutting or delaying. Try removing $6,000 of marketing from scenario A: on these invented numbers, revenue growing 3% a month overtakes the lower costs in month 28, cash bottoms out at about $33,000 in month 27 and never runs out. That result rests entirely on growth holding for over two years, which is exactly the sort of assumption to doubt, so test it with growth at 1% before believing it. Decide before you are forced to.

The sheet does not make the decision for you, and I would not let it. What it does is stop you deciding on a feeling. If you want a second pair of eyes on the structure, or help setting up the monthly routine so it survives busy weeks, that is the kind of job my AI implementation consultation covers, and you can reach it through the contact page. For the wider picture of what AI can do for a young company, an AI playbook for a small startup with a 90-day plan is the place to continue.

Runway and hiring models: questions founders ask

Which AI tool builds the sheet best?

Any of the main assistants can draft a model from a clear specification: ChatGPT with an uploaded file, Claude for Excel, Copilot's side pane in Excel, or Gemini in Sheets. Judge them on whether the formulas they write are readable and survive your checks, not on how fast they produce a grid.

Can I paste my real bank figures into a chatbot?

Use rounded or placeholder numbers while building the structure, then type your real figures into the finished sheet yourself. If you do upload real data, check the plan's data terms first, and keep investor details and personal salary data out.

How accurate is a runway figure?

It is only as good as the assumptions. A runway based on flat burn is a floor or a ceiling depending on growth, so model at least three cases and treat the answer as a range of months rather than a single date.

Should I model hires by month or by revenue trigger?

Both. Month-based is simple and shows what a hire costs. Trigger-based (hire when monthly revenue reaches a figure) is closer to how you will actually decide. Start with months, then add a trigger column once the basics work.

What should I do when the model shows under six months?

Take it as a decision point, not a verdict. List what you could cut or delay, when a raise or revenue milestone could land, and the date by which you must choose. Talk it through with your accountant or an adviser before acting.

Further reads

• AI Automation for Startups: Seven Workflows, Ranked by Set-Up Effort — Pick the workflows that free founder hours first.
• Startup AI Tools for Growth: 8 That Earn Their Subscription — Find which subscriptions earn their cost.
• AI HR Software for Startups: Which Tool Fits Which Job — Compare HR tools once the hiring plan is set.
• AI Cash Flow Forecasting Tools for Small Businesses Compared — See dedicated forecasting tools against a spreadsheet.
• Claude vs ChatGPT for Startups: Which Plan for a Team of 3 to 10 — Choose a plan for a team of three to ten.
• AI Tools for Startup Ideation and Validation: Test an Idea First — Test demand before you commit hiring money.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: Microsoft Support's COPILOT function page and Microsoft 365 message MC1454373 on its retirement, Anthropic's Claude for Excel documentation, Google's help page on Gemini in Google Sheets, and the ChatGPT and Claude pricing pages. All model figures are invented for illustration; the scenario results were recomputed month by month.