About a year ago, I ran a session called “AI-Powered Budgeting.” At that time, the examples felt more like prototypes than real, practical tools.
Fast-forward to today: the progress in large language models is so impressive that what was experimental last year is now part of my daily routine — from building models in Excel to summarizing meetings and testing assumptions.
And it couldn’t come at a better time. Budgeting season is traditionally busy, stressful, and full of headaches. Having technology that actually reduces pressure — rather than adds to it — feels almost like cheating. But it’s here, and it works.
Big news: we are launching a paid tier next month
It’s hard to believe this newsletter is almost a year old. What started as an experiment has become part of my weekly routine. I may use AI to draft ideas, but I spend hours refining each edition so it delivers practical, CFO-level insights.
Your feedback has been incredible — many of you have said this helps you cut through the hype and actually apply AI. That’s why I’m taking the next step.
Starting next month, I’m launching a paid subscription. The free version stays, but the paid tier will add:
Board-ready update packs
Step-by-step playbooks with prompts and templates
Compliance and governance alerts
Think of it this way: the free newsletter is your overview. The paid tier is your action plan.
👉 Support the newsletter and unlock subscriber-only content — subscribe now.
This week’s theme: Budgeting season
Here are five budgeting problems every CFO knows too well — and how an LLM can actually help.
1. Setting up the budget workbook
Problem: Building a clean budget file with cost centers, rollups, and dashboards takes hours. Every department insists on its own template.
Solution: Use Claude’s Excel creation feature to generate a structured workbook: department tabs, a consolidated sheet, and even a simple dashboard. It won’t be perfect, but it gets you to 80% in minutes instead of days.
How to do it:
Outline your cost centers and budget parameters in a quick list.
Paste this into Claude:
“Create an Excel budget workbook with tabs for Assumptions, each department (Ops, Sales, Marketing, IT, HR), a Consolidated sheet, and a Dashboard. Departments should include rows for headcount, salaries, benefits, travel, software, marketing, and other costs, with Jan–Dec 2026 columns. Consolidated should use SUMIFS to roll up all departments. The Dashboard should display total revenue, opex, EBITDA, and cash burn with simple charts.”
Tips:
Claude’s file creation feature was launched very recently, and the output sometimes contains broken formulas—check every tab carefully. Additionally, it is currently available only to Max, Team, and Enterprise plan users.
Ask Claude to color-code input cells and formula cells.
If you have a decent current-year budget file, upload it as well and ask Claude to reference it — this helps preserve formatting and assumptions.
2. Spotting sales trends to pressure-test assumptions
Problem: When revenue assumptions are based on gut feel, you spend the rest of the year explaining variances. Sales data exists, but it takes hours to slice across products, markets, and time periods.
Solution: LLMs can analyze large datasets quickly — finding trends, seasonality, anomalies, and then drafting starter scenarios. It’s like having a junior analyst work overnight.
How to do it:
Export a CSV with your sales data
Paste this into Claude or ChatGPT with:
“Analyze this dataset. First, audit for missing values, outliers, or inconsistencies. Then summarize trends by market × product group (YoY growth, momentum, seasonality). Highlight the top 5 drivers of growth or decline. Build three scenarios (best/base/worst) with clear levers: volume, price, discount rate. Return a summary table and a short narrative I can use in the budget deck.”
Tips:
Don't run the full prompt at once; break it down into steps and check the output for each step.
Ask the model to list the assumptions it made explicitly.
Save that assumption list next to your data — it serves as both documentation and a reference.
3. Reviewing department submissions for inconsistencies
Problem: Department budgets come in all shapes and sizes. You’ll see duplicate expenses, unexplained growth, or headcount assumptions that don’t tie to hiring plans. Reviewing it all manually takes forever.
Solution: Use ChatGPT (or any other LLM) as a first-pass reviewer. It won’t replace your judgment, but it will flag obvious red flags and save hours of scanning spreadsheets.
How to do it:
Combine department submissions into one file or upload the budgets as separate files.
Ask:
“Review these department budgets. Flag YoY changes over 15% without explanation. Identify duplicates (same vendor, GL, and month across cost centers). Check if revenue growth aligns with headcount growth. Identify mis-categorized expenses (e.g., training vs marketing). Produce a log of issues by department with severity (high/medium/low) and a suggested owner to fix.”
Tips:
Make “Assumptions” a required column in department templates. It helps separate justified changes from mistakes.
Treat the model’s feedback as suggestions, not verdicts.
4. Capturing meeting decisions
Problem: Budget meetings generate long discussions, side agreements, and to-do items that never get tracked. By the next meeting, nobody remembers what was agreed.
Solution: Record or copy notes into an LLM and have it create a structured decision log and to-do list. It keeps everyone aligned and accountable.
How to do it:
After the meeting, paste in the transcript.
Ask:
“From these notes, list all decisions with topic, owner, and effective date. Extract all requested model changes with tab, cell, or logic, due date. Create a to-do list with owner, action, and deadline. Draft a summary email for the team.”
Tips:
Do not use the automated meeting summary feature or app; instead, take a meeting transcript and paste it into an LLM with your specific instructions.
Always get consent before recording/transcribing meetings.
Assign a unique ID to each decision and reference it in your budget model’s Change Log.
5. Market expansion research
Problem: Budgeting sometimes involves debating entry into new markets. Obtaining reliable information on taxes, labor regulations, and the ease of doing business typically requires weeks of research or hours of consulting.
Solution: Use a Deep Research feature (available in ChatGPT, Claude, Copilot, etc.) to compile a market comparison table with tax rates, payroll costs, incentives, and regulatory watch-outs — all in hours, not weeks.
How to do it:
Define the target markets and what you need: corporate tax, VAT/sales tax, payroll costs, data privacy rules, etc.
Ask:
“Research [Markets A, B, C]. Create a comparison table with corporate tax, VAT/sales tax, employer social costs, statutory benefits, payroll frequency, data residency rules, import restrictions, typical payment terms, and ease-of-doing-business indicators. Include source links and the as-of date for each metric. Add a 150-word summary with 5 risk watch-outs and 3 assumptions we should validate locally before finalizing the budget.”
Tips:
Require dated sources and at least two reputable citations for key metrics.
Treat this as pre-work — always validate with local advisors.
Stamp the research with “last updated” so you know when it goes stale.
These are just a few ways AI can take pressure off your budgeting process. The value is real — hours saved, errors avoided, and cleaner insights for you and your team.
Closing Thoughts
What excites me most about LLMs is that they’re finally moving from “interesting demos” to “real CFO assistants.” They won’t replace our judgment, but they can save hours, cut errors, and give us sharper starting points.
And since budgeting is the big focus right now, I’ll be sharing more practical cases and step-by-step AI guides over the next few weeks. This is where AI can make the biggest impact — right when the pressure is highest.
👉 One of the October paid-tier toolkits will be a set of ready-to-use budgeting prompts — designed to take you from raw data to clean scenarios faster than ever.
If you’d like to support this newsletter and unlock subscriber-only content, you can upgrade now.
We Want Your Feedback!
This newsletter is for you, and we want to make it as valuable as possible. Please reply to this email with your questions, comments, or topics you'd like to see covered in future issues. Your input shapes our content!
Want to dive deeper into balanced AI adoption for your finance team? Or do you want to hire an AI-powered CFO? Book a consultation!
Did you find this newsletter helpful? Forward it to a colleague who might benefit!
Until next Tuesday, keep balancing!
Anna Tiomina
AI-Powered CFO