Strategy & Planning

How to run annual planning with AI

Strategy & Planning3 AI tools7 steps6 friction points

Want an agent to handle this on your live data? See the Starch version →

Annual planning is the process of setting company goals, allocating budget, and mapping out hiring and priorities for the coming year. For most operators, it means pulling together financial actuals, reviewing what worked and what didn't, pressure-testing assumptions, and landing on a plan the whole team can execute against. It's one of the highest-stakes documents a company produces each year — and one of the most time-consuming to actually build.

People reach for AI because annual planning is fundamentally a synthesis problem. You're pulling disparate inputs — revenue trends, headcount costs, competitive landscape, strategic bets — and turning them into a coherent narrative with numbers attached. LLMs are good at structure, summarization, and drafting, which makes them feel like a natural fit for the parts of planning that involve translating a pile of data and opinions into a readable document.

ChatGPT, Claude, and Gemini can realistically help with framing your planning structure, drafting narrative sections, stress-testing assumptions through Q&A, and generating slide or document outlines. Where they run into limits is the data layer: they can't pull your actuals from QuickBooks, don't know what your burn rate was last quarter, and have no memory of the plan you built last year. You're the integration layer — everything flows through copy-paste.

Skip the manual work → See the Starch version

Strategy & Planning3 AI tools7 steps6 friction points

AI walkthrough

How to do it with AI today

A practical walkthrough using ChatGPT, Claude, and other off-the-shelf LLMs — what they're good at, what you'll have to do by hand.

Tools that work for this

ChatGPTClaudeGemini

Step-by-step

1 Export last year's financial actuals from QuickBooks or your accounting system as a CSV, and copy the key numbers — revenue, burn by category, headcount costs — into a Claude or ChatGPT conversation.

2 Paste in your goals or OKRs from last year (copy from Notion, Google Docs, or wherever they live) and ask the model to summarize what you hit, what you missed, and what the pattern suggests about your planning assumptions.

3 Ask the model to generate an annual planning template — sections for financial targets, headcount plan, key initiatives, and risks — customized to your company stage and business model. Use this as your working document skeleton.

4 For each major initiative or budget line, paste in the relevant context and ask the model to help you build the assumption set — what growth rate are you modeling, what does that imply for spend, what are the key dependencies.

5 Use the model as a devil's advocate: paste in your draft plan and ask it to identify the three most likely ways this plan fails, then iterate on the assumptions it flags.

6 Draft the narrative sections — the 'where we're going and why' text that will accompany the numbers — by describing your strategy in the chat and asking the model to turn it into polished prose suitable for an all-hands or board deck.

7 Manually assemble outputs from each conversation into your final document, Google Slides, or planning template. This part is fully manual — the model can't write directly to your tools.

Prompts you can copy

Here are my actual revenue and spend numbers by category for the past 12 months: [paste data]. Summarize the key trends and flag anything that looks like an outlier or planning risk for next year.

I'm building an annual plan for a 15-person B2B SaaS company doing $2.4M ARR. Generate a complete planning document structure with sections, subsections, and a brief description of what belongs in each.

Here is our draft annual plan: [paste plan]. Act as a skeptical CFO and identify the 3-5 assumptions most likely to be wrong, and what the downstream impact would be if they are.

We're planning to grow headcount from 15 to 22 people next year. Given a $180K average loaded cost per hire and our current burn of $140K/month, model what our monthly burn looks like at each quarter boundary.

Turn the following bullet-point strategy summary into a 3-paragraph executive narrative suitable for opening our all-hands annual planning presentation: [paste bullets].

Reality check

Where this gets hard

The walkthrough above works — until your numbers change, the LLM hallucinates, or you have to re-paste everything next month.

No live connection to your actual financial data — you copy-paste last quarter's numbers manually, which means the plan is already stale before you finish it.

Each planning conversation is isolated; the model has no memory of the assumptions you agreed on in session one when you're refining headcount in session three.

Output structure drifts between runs — the table format or section logic you carefully prompted in October won't reproduce reliably in November when you pick the work back up.

Scenario modeling requires re-running the entire prompt chain from scratch each time you change an assumption; there's no 'update the model and see the impact' interaction.

Nothing persists into next year — the plan lives in a Google Doc and the prompts are gone, so the annual planning process starts at zero again in 12 months.

The model can't write outputs back to your tools — slides, Notion pages, and budget spreadsheets all require manual copy-paste after each generation step.

Tired of the friction?

Starch runs the whole workflow on live data — no copy-paste, no hallucinated numbers, no re-prompting next month.

See the Starch version →

Starch alternative

The same workflow on Starch

Starch is an agentic operating system — instead of one-off prompts you run manually, an agent builds persistent apps connected to your live business data that run the annual planning workflow on an ongoing basis.

Connect Plaid and Stripe once — Starch syncs your actual transactions and revenue on a schedule, so when you open the Budgeting app (coming soon), the baseline reflects real numbers, not a CSV from last month.

Use the Scenario Analysis app to compare multiple versions of your annual plan side-by-side — model the impact of delayed hiring, slower growth, or a price increase against your actual Stripe revenue and Plaid burn data without rebuilding a spreadsheet.

Describe the planning dashboard you want in plain English — 'show me headcount cost by quarter against our hiring plan, with variance from budget' — and Starch builds it as a persistent app that updates as new data comes in.

Use the Knowledge Management app to store last year's plan, assumptions, and retrospective notes in a searchable wiki, so next year's planning starts from a structured baseline instead of a blank doc.

Ask Starch to build an automation that pulls your monthly actuals from QuickBooks on a schedule and updates your annual plan variance tracker — no manual exports, no re-running prompts.

The Presentation Agent (coming soon) will let you describe your annual planning narrative and generate a polished board-ready deck from the same live data your Scenario Analysis app already uses.

Get closed-beta access →

Toolkit