You Want to Implement AI in Your Accounting Team? Do This First
Accountants are famous for rolling things forward. Episode 2 shows the better way. Hand the workbook to Claude, ask one question, and review what comes back.
What Should Your Accounting Team Do First with AI?
Take one existing workbook and ask AI to study it and give you five recommendations to make it better. You don't need a tool rollout, a policy document, or a committee to start. In Episode 2, John takes a monthly depreciation schedule full of hard-coded values, gives Claude that one prompt, and walks through everything that happens next. Claude replaces the hard-coded numbers with formulas, adds validation to the input columns, builds a category summary, wires up a journal entry, and documents the file for the next person who opens it.
This is the exact exercise John has instructed his own team at Zeroed-In to run. There are so many directions you can take AI that sometimes the hardest part is just starting.
Why Do Accounting Workbooks Stay the Same for Years?
Accountants are well known for rolling things forward. Do the same thing as last month. Do the same thing as last year. Very rarely do we play outside the sandbox.
Once a workbook gets built, it can stay the same for five years, even though the process it supports has completely changed. Instead of rebuilding, band-aids get layered on top of band-aids just to keep it working. Depreciation gets copied and pasted in as hard-coded numbers instead of formulas. Someone fat-fingers an F where an E should be and the SUMIF quietly stops picking it up.
Everyone has inherited a file like this. Episode 2 is John's way of saying there is a better way, and it takes one sentence to find it.
The Shift
The Hardest Part Is Just Starting
"There's so many different ways that you can go with this that sometimes the hardest part is just starting."
So start small. Take a workbook you already own, one that is easy, and give Claude a single sentence to work from. Let it understand what the file actually does, then let it come back with ideas grounded in your real data.
It works because the first win happens inside a file your team already understands, in software they already have open.
Inside the Episode
What You'll See in Episode 2
- The one-sentence prompt that opens the exercise, run live on a real monthly depreciation schedule
- Which of Claude's recommendations John takes, which one he skips, and how he judges each of them
- The mistake John makes on camera, and what Claude does when you keep editing cells while it works
- Where he pushes past the AI's list, and the last small ask that documents the file for whoever opens it next
The 4-Step Exercise John Runs in Episode 2
One workbook, one prompt, and a review pass. This is the exact sequence from the episode, on a real depreciation schedule.
Pick one existing workbook, and make it an easy one
John starts with a monthly depreciation schedule where the depreciation was copied and pasted in as hard-coded values, the kind of file everyone has inherited. Choose something you know well so you can judge every suggestion the AI comes back with.
Ask for five recommendations
The prompt is one sentence, and this is it verbatim. Can you please study the workbook and give me five recommendations that you would make to make this workbook better? Claude reads the file and comes back with ideas grounded in your actual data, not generic spreadsheet advice.
Implement what you agree with, then push further
John accepts four of the five and skips the one he can go either way on. Then he pushes past the list. He brings in his domain expertise to ask for a journal entry that updates automatically from a month dropdown, debiting depreciation expense by category and crediting accumulated depreciation, complete with a check balance.
Ask for a how-to-use guide
Finish with one more small ask. Can you put together a quick page on how to use this schedule? Now your reviewer understands the file, and whoever covers for you on PTO or a sick day has the user guide sitting right in the workbook. Such a small ask and such a big gain.
Let AI Do It, Then You Review It
John reviews every recommendation before he accepts it. He agrees strongly with some, goes either way on others, and skips one entirely. Let the AI do it, then you review it. You own the result.
Key Takeaways from Episode 2
- Don't roll it forward again. Pick one easy workbook and ask AI how to make it better.
- You stay the reviewer. Implement the recommendations you agree with and skip the ones you don't.
- It all starts with starting. Fix one workbook first, then hold weekly touchpoints with the team.
Your Guide
Who's Teaching This Episode
John Ikosipentarhos, CPA
President & Co-Founder, Zeroed-In Consulting
John has built, inherited, and untangled the exact kind of workbooks this episode fixes. Today he helps accounting and finance teams put AI to work on them, and the exercise in Episode 2 is the same one he runs with his own team at Zeroed-In. He has 12+ years of experience in public and private accounting, specializing in month-end close, financial statement preparation, AI and Automation, and is a licensed CPA in California.
Common Questions
Episode 2 Frequently Asked Questions
Start with one existing workbook, not a new tool rollout. Pick an easy file, give Claude a single prompt asking it to study the workbook and recommend five improvements, implement the ones you agree with, and review the result. Episode 2 demonstrates the entire loop on a monthly depreciation schedule, and it is the exact exercise John has instructed his own team at Zeroed-In to run.
Can you please study the workbook and give me five recommendations that you would make to make this workbook better? That is the whole prompt. The AI studies the file and comes back with recommendations grounded in your actual data, which you then accept, modify, or skip.
Claude came back with five ideas. Replace hard-coded depreciation values with formulas. Add a percent depreciated and remaining useful life column. Format zero values as dashes so they are easier to read. Add data validation on key input columns. And add a year-to-date column with a summary section by category. John implemented four of them and then went further by asking for an automatically updating journal entry.
No, and you shouldn't. In the episode John agrees strongly with some recommendations, can go either way on others, and simply skips one. The AI does the work and you review it, the same way you would review a staff accountant's work.
Because someone other than you will open the file one day. It could be your reviewer, or a teammate covering while you are out sick or on PTO. In the episode, one short prompt produces a full user guide covering the asset listing, the depreciation grid, the category summary, the journal entry, how to add a new asset, and what not to touch.
Be careful with that. As John finds mid-episode, Claude does not like it when you are in the cells editing while it is making changes. Give it the file, let it work, and review the output when it finishes.
Episode 2 premieres free on YouTube on September 3rd, 2026 at 9:00 a.m. Pacific Time. Join the series to get the episode delivered to your inbox the moment it's released.
Episode 3, releasing in October 2026, covers Claude in Excel for accounting and finance teams. A new episode is released every month through December 2026.
Keep Going
Your Next Step
Catch Up on Episode 1
Episode 2 builds on the habit from Episode 1. Open AI inside Excel and just ask the question. If you haven't watched it yet, start there and see why the years-long climb from formulas to VBA is now optional.
Watch Episode 1Find Your Team's Place on the AI Skill Tree
Episode 2 gives your team its first workbook win. The interactive AI Skill Tree maps the full journey for accounting and finance teams, so you can pinpoint where you are today and the next move that compounds.
Explore the AI Skill TreeJoin the Series
Receive Each New Episode as It's Released
Most accounting and finance professionals know AI is changing the industry. The challenge isn't whether to adopt it. It's knowing where to start.
This video series walks you through a practical progression from using AI as a research assistant to building custom workflows and AI-powered applications that solve real accounting problems.
Each month, you'll receive a new episode, practical takeaways you can apply immediately, and guidance from our President and Co-Founder, John Ikosipentarhos, as you move from AI curiosity to confident implementation.