Boring workflows win

There’s a pattern I keep seeing across every AI company that’s actually making money (not just raising it): they’re automating the most tedious work imaginable.

Financial close processes. Invoice reconciliation. Compliance document review. Employee onboarding paperwork. The kind of work where the person doing it is counting the minutes until lunch.

This runs counter to the popular narrative. The exciting AI demos are always creative: generating art, writing novels, composing music, having philosophical debates. And those are genuinely impressive. But the companies building sustainable businesses are pointing AI at spreadsheets, not canvases.

Why? Three reasons.

First, boring workflows have clear completion criteria. You either reconciled the invoices or you didn’t. You either processed all the compliance documents or you missed some. This makes it easy to measure whether the AI is actually working, which makes it easy to sell, and easy to keep selling.

Second, the cost of the status quo is obvious and quantifiable. If a team of six people spends 200 hours per month on financial close, that’s a number a CFO can see on a spreadsheet. “We’ll cut that to 40 hours” is a conversation that closes deals. Try having that same conversation about AI-generated marketing copy.

Third, these workflows have low tolerance for creative interpretation. You don’t want your invoice reconciliation system to get creative. You want it to be accurate, fast, and boring. This is actually great news for current AI capabilities, because reliability on structured tasks is solvable in ways that reliability on open-ended tasks isn’t.

I’ve been tracking evidence for and against this thesis in my knowledge system. So far the evidence is lopsided: Stripe’s autonomous coding agents handle 1,000+ pull requests per week on well-defined tasks. A company called Stacks AI saved enterprises 100,000 hours on financial close. Anthropic’s own enterprise customers get the most value from role-specific agents doing repetitive work, not from general-purpose chat.

The counterexample is worth noting: some companies are successfully automating creative work too. AI photography is replacing human photoshoots. Coding agents are helping with open-ended architecture decisions. So the thesis isn’t that creative AI has no value. It’s that if you’re building a business, pointing AI at boring workflows is the faster path to revenue and the easier path to proving value.

The boring stuff is where the money is. It’s not glamorous, but it works. And in a market full of impressive demos that don’t convert to paying customers, “it works” is a competitive advantage.