Can you review work you’ve never learned to prepare? When we introduce AI into our firms, that’s basically what we’re asking young accountants to do.
That’s a pretty fundamental change to how we develop people in this profession.
Preparation Was Never Just Preparation
AI agents are starting to do more of the work that traditionally taught young accountants how accounting actually works. The journal entries, reconciliations, returns and other preparation work weren’t just tasks we needed someone to complete. They were how people learned:
What good work looked like
What tended to go wrong
What should tie
What didn’t look quite right
When something just didn’t make sense
Now the agent can do that work faster, but the human still has to make sure it’s right. And reviewing something is a different skill than preparing it.
“Check the AI’s Work” Isn’t a Training Plan
We can’t hand someone an AI agent and tell them to “check its work.” What does that actually mean? What should they check every time? What should tie? What mistakes should they anticipate? When should they trust the output, and when should they start asking questions?
Those are review skills, and we’ve historically developed them after someone had years of experience doing the work themselves. Now we may need to figure out how to develop them much earlier.
The Learning Curve Is Flipping
Instead of spending most of our time producing the work, we will soon be spending more of it:
Reviewing
Questioning
Testing
Interpreting
Explaining
From the worker’s perspective, that could actually be harder. Imagine an agent constantly handing you completed work and knowing you’re responsible for finding the thing it got wrong.
That requires a different kind of confidence and judgment than simply knowing how to follow a process.
We Need to Teach People How to Supervise AI
That means firms need actual procedures for reviewing AI work. We need to teach young accountants:
What they should review every time
Where AI has failed before
What common errors look like
How to test whether an answer makes sense
How to question the AI when they know something is wrong but don’t immediately know why
And they need opportunities to practice finding those errors before the stakes are high. You can’t just teach people to how to use the agent, you have to teach them how to supervise it.
Technical Knowledge Matters, But The Goal Changes
Young accountants still need technical knowledge. They still need to understand why something should tie, what a correct return looks like and what the numbers actually mean. But maybe we need to teach those things with a different goal in mind.
Instead of only teaching someone how to produce the answer, we need to teach them how to interrogate one.
That’s a significant shift.
AI Could Create More Room for Judgment
There’s an opportunity in that shift, too. We spent so much time doing the work that we barely had time to think about the work. If AI handles more of the preparation, accountants can spend more time:
Understanding what the numbers mean
Explaining them to clients
Asking better questions
Identifying what doesn’t fit
Developing judgment
Those are exactly the capabilities firms say they want their people to develop sooner. But that development won’t happen automatically just because we gave them an AI agent.
The Profession Has to Redesign How People Learn
AI isn’t going back in the box. The young accountants of the world are going to have to learn how to work with it, just like I had to learn how to use the internet and Google.
The difference is that AI isn’t just changing the tool they use to do the work. It’s changing the work they need to learn first. Our job is to make sure they’re ready.