Forecasting and scenario planning with AI

Membership owners make big decisions with little to go on. Should you raise the price? Can you afford a part-time community manager? What happens if the launch brings in half the members you hoped for? Most owners answer these with a gut feeling and a single hopeful number, then feel blindsided when reality lands somewhere else.

Scenario planning is the calmer alternative. Instead of one forecast, you sketch a few plausible futures, see what each would mean for members and money, and decide in advance what you would do in each. AI makes this much easier. It can help you structure a simple model, question your assumptions and think through responses. It cannot see the future, and it will happily produce convincing numbers from nothing, so the approach below keeps your own data at the center.

A forecast is a set of assumptions

Every forecast rests on a few assumptions: how many people join each month, what share of members leave, what they pay and when. The number at the bottom is only as good as those inputs. The most useful thing AI can do is help you write the assumptions down clearly and ask whether each one is reasonable, based on your own history.

Start by pulling a year of your own figures: joins, cancellations and revenue by month. If you are new, use whatever months you have, and accept that your ranges will be wider. The non-AI side of turning your history into goals is covered in setting targets from your own numbers.

Build the model in a spreadsheet, not a chat

Ask AI to design the model, then build it in your spreadsheet so the arithmetic is visible and correct. A simple monthly membership model has a few rows: members at the start of the month, new joins, cancellations, members at the end, price and revenue. Each month's starting members are the previous month's ending members. The assistant can write the formulas and explain them, as described in AI for spreadsheets, but the spreadsheet should do every calculation. Chat answers that do long arithmetic in their heads are where errors creep in.

A prompt for setting up scenarios

Once the model works, use AI to shape the scenarios around a specific decision:

I run [name of your membership], a [one-line description]. I am deciding whether to [decision, for example: raise the monthly price from X to Y for new members]. Here are my last [number] months of figures: [paste monthly joins, cancellations and revenue as totals only]. Help me set up three scenarios for the next twelve months: cautious, expected and hopeful. For each, suggest values for new joins per month and monthly cancellation rate, and explain how each value relates to my own history. Do not use industry averages or outside figures. Then list the assumptions I am least sure of, and suggest an early warning sign that would tell me which scenario I am in.

A good answer ties every value to your history, for example: “Cautious assumes joins drop to your weakest month for the whole year.” If it quotes a typical industry churn rate or a benchmark it has not sourced, ignore it; figures like that are often invented.

Three scenarios, each with a plan

Anouk runs a ceramics membership and was considering a price rise. She plugged the three scenarios into her spreadsheet. In the expected case, revenue rose even with slightly fewer joins. In the cautious case, revenue fell for about four months before recovering. That told her the real question was whether she could live with four lean months, not whether the rise would “work”.

For each scenario, write down what you would do. Cautious might mean pausing a planned hire or running a win-back offer. Hopeful might mean opening a waiting list for a new cohort. Then name the early signal for each, such as joins in the first six weeks after the change. Deciding now saves you making a panicked call later.

Use AI as a sparring partner

Once your scenarios are in place, ask the assistant to attack them. Good questions include: what could make the cautious case worse, what am I forgetting and which assumption would change the answer most if it were wrong? It may raise things like seasonality, annual members renewing all in one month, or failed card payments that look like cancellations. Rafael, who manages an association for independent bookshops, had forgotten that most dues renew in a single month, so a bad renewal season would hit all at once. The model changed shape completely.

Keep the privacy line clear: totals by month are fine to share, but member-level records and financial statements should stay out of consumer AI tools. Use a paid business plan with training on your data turned off if you work with anything more detailed.

Check the forecast against reality

A forecast is only useful if you compare it with what actually happens. Each month, add the real figures beside the scenarios and see which line you are tracking. Summarizing your membership numbers with AI fits neatly here. When reality drops below your cautious line, or an early warning sign fires, treat it as your cue to act on the plan you wrote, and to revisit the assumptions that turned out wrong.

What to do next

• Choose one decision you are facing that depends on future numbers.
• Export a year of your monthly joins, cancellations and revenue.
• Build a simple monthly model in a spreadsheet with AI help on the formulas.
• Run the scenario prompt and enter all three cases.
• Write an action and an early signal for each scenario.
• Put a monthly reminder in your calendar to compare the forecast with reality.