For most of my career, I was the checklist.
Not metaphorically. Literally. Every month, I sat down with the bank statements, the GL, the sub-ledgers, and I worked through it. Line by line. Transaction by transaction. I flagged the exceptions, investigated the ones that didn't match, booked the adjustments, calculated the accruals, prepared the journal entries, and assembled the reconciliation package for review. Then I did it again for the next account. And the next.
I'm a CPA. I went into accounting because I wanted to solve hard problems. And I spent years doing work that a sufficiently patient person with a spreadsheet could do with a bit of guidance and a basic understanding of accounting.
I don't say that as a complaint. I say it because it's the honest accounting of what the close actually is and because understanding that is the only way to understand why AI changes everything about it.
What I learned at scale.
I spent time at Zendesk when the company was growing fast. Big team, complex systems, mature finance function. The kind of environment where everything has a process, every process has an owner, and the close still takes two weeks.
That surprised me at first. I assumed that more resources meant a cleaner, faster close. What I learned is that complexity doesn't scale linearly with company size — it compounds. More entities. More accounts. More transactions touching more systems. More exceptions requiring human judgment. The checklist gets longer. You just hire more people to run it.
The work was more sophisticated than it looked from the outside. But the fundamental nature of it hadn't changed. Someone still had to pull the data. Someone still had to do the matching. Someone still had to book the entry.
More resources meant a longer checklist, not a shorter one. You just hired more people to run it.
What I learned from scratch.
Then I went to ENDVR and built the accounting function from nothing. No inherited processes. No existing chart of accounts. No team. Just a blank slate and a set of transactions that needed to be accounted for.
When you build it yourself, you make every decision consciously. You think about what goes on the checklist and why. You figure out which parts require judgment and which parts are just execution. You learn, fast, that the ratio is not what you'd expect.
Most of the close is execution. Pattern recognition. Rule application. It's not that it's easy, it's that it's deterministic. Given these inputs, you do these steps, you get this output. The hard part isn't the work. It's the volume of it, and the zero tolerance for error.
That's not to say the close has no hard parts. The exceptions are hard. The intercompany timing differences, the accrual estimates where the data is thin, the cutoff calls that require someone who actually understands the business. Those still need a human. But the exceptions are maybe 20% of the work and 80% of the stress. The other 80%, the matching, the calculation, the entry preparation, is just execution. And execution is exactly what AI is built for.
That insight sat with me for a long time. Because if most of the close is deterministic, then most of the close is, in principle, automatable. The question was whether anyone would ever build the right system to do it.
Why I joined Kinter.
I didn't join Kinter despite being a CPA. I joined because of it.
The people who should be most skeptical of AI doing this work are the ones who understand exactly what the work is. I've built the checklist. I know which items on it require a human and which ones don't. That map is clearer in my head than it would be for anyone who hasn't sat in the seat.
And honestly? Most of it doesn't require a human. Not the reconciliation. Not the prepaid calculation. Not the journal entry preparation. Not the variance flag. Those things require precision, consistency, and access to the right data. They don't require judgment the way an audit finding does, or a board presentation does, or a conversation with a CFO about why the numbers moved does.
The work that kept people out of accounting: the 11pm close, the deadline grind, the tedium that has nothing to do with why anyone became a CPA is exactly the work AI can do. And do better. Not because it's smarter. Because it doesn't get tired. Because it runs continuously. Because it doesn't skip a step at the end of a long quarter.
The work that kept people out of accounting is exactly the work AI can do. And do better.
What actually changes.
Here's what I want practitioners to understand, because I think the conversation about AI and accounting has gotten muddled.
This isn't about replacing accountants. It's about replacing the part of the job that was never the point of the job. Nobody became a CPA to reconcile bank statements. They became a CPA to understand businesses, to find the signal in the numbers, to help organizations make better decisions. The reconciliation was just the toll you paid to get to the interesting work.
When AI handles the execution layer, the interesting work is all that's left. Flux analysis. Anomaly investigation. Strategic forecasting. The conversations with leadership that actually require what you spent years learning.
I've seen what it looks like when a controller's month isn't consumed by the close. It looks like someone who can actually do their job.
What I'd say to other controllers.
I know the instinct. When someone tells you AI is going to do your reconciliations, the first response is skepticism. Show me. Prove it. I've heard this before.
That's the right instinct. You should be skeptical. The close is not the place for demos that break in production. The books are the backbone of the business, and they need to be right every time, not most of the time.
What I'd say is: hold AI to the same standard you hold yourself. Ask the hard questions. What happens when the data doesn't match? How does it handle exceptions? What's the audit trail? Those aren't objections to AI, they're the questions any good controller asks about any new process.
The answers are getting better. Fast. And the teams who figure out how to work with AI agents rather than around them are going to run circles around the ones who don't.
Hold AI to the same standard you hold yourself. Those aren't objections to AI — they're the questions any good controller asks about any new process.
The checklist isn't going away.
The B-17 still had a preflight checklist after Boeing redesigned the process. The checklist didn't disappear, the human executing every line did. The pilot reviewed it. The system ran it.
That's where we're headed. The close still happens. The reconciliation still happens. The accruals still get booked and the journal entries still get prepared. The difference is who, or what, does it.
I spent my career being the checklist. I'm at Kinter because I think that's the last generation that will have to be.
If you want to see Kinter in action, book a demo.