Skip to content
The Rolling Forecast

If You Got Hit by a Bus Tomorrow...

By ·

On our latest episode of Making Cents with Charlie, I sat down with my friend and mentor Greg Hood, CPA, CMA co-author of Who Moved My Data?, who has spent his career living inside finance data warehouses. We got into AI, and he said something that has stuck with me since.

He told me about a senior accountant who'd built a genuinely brilliant Claude skill. Greg asked him one question: does your controller know this exists? No. Follow-up: what happens to it the day you leave? Nobody has access to his desktop. It goes with him.

Greg has lived this exact scenario before, fifteen years ago, with an Excel macro. He built it, he left the company, and it broke. It took them two months to reclose the books because nobody else knew how to write it.

Greg said: "We've got all these senior accountants writing all these skills in Claude, but no one's actually transferring the knowledge upwards. We've just accelerated the single point of failure. This is a rewrite of VBA and Excel macros from twenty years ago. We're just calling it something else."

This isn't a productivity story. It's the same failure mode running at machine speed.

This isn't only Greg's anecdote either. KPMG's 2026 Global AI in Finance report found active AI use in finance functions jumped from 30% to 75% in two years. Adoption doubled. Governance didn't. Grant Thornton's 2026 AI Impact Survey found 78% of business executives don't have confidence they could pass an independent AI governance audit within 90 days.

Here's what Greg does differently, and what I've started copying. Engineering teams document by default, they centralize everything in something like GitHub, and knowledge transfer happens because the system forces it. Finance teams are mostly still on individual pro accounts with no shared record of anything anyone built. Greg's fix is unglamorous: a monthly team meeting where everyone shows what they built that month, documentation stored somewhere the team actually owns, legal and IT looped in before it spreads.

Greg said: "Your AI that you're doing on your desktop has to be enterprise ready."

Not enterprise ready like a compliance checkbox. Enterprise ready like this: if you got hit by a bus tomorrow, would the work survive you.

Greg and I covered a lot more than this, data rooms that talk back, forecasts built on vibes, why curiosity might be the one skill AI can't hand you. Catch the full episode of Making Cents with Charlie wherever you listen.

Sources

Last updated

If you're reading with more intent than most

How We Work walks through our six stages, including what usually goes wrong at each one.