Our Accountants Start Their Day in a Terminal. Here's How We Got There.

I work at a finance company where the accountants open a terminal before they open their email.

Not the engineers. The accountants.

I’m the Chief of Staff at Till CFO, and I’m not an engineer either. My background is operations, people, and an unreasonable number of spreadsheets. So when I say our controllers and accountants now do real work from a command line every single day, I’m not describing a team of secret programmers. I’m describing ordinary finance professionals who built a new muscle on purpose, with structure.

Here’s the position I’ll defend: AI fluency is not a tools problem. It’s a reps problem.

What Didn’t Work

Demos. Tool demos are exciting for exactly one afternoon. Someone shares their screen, the room says “whoa,” everyone goes back to their actual job, and two weeks later nobody can reproduce what they saw. Watching someone else use AI is like watching someone else go to the gym. Genuinely nothing happens to you.

Licenses didn’t work either. Buying access to a tool and building capability with it are two different line items, and only one of them shows up on the invoice.

What Worked

Daily reps. We run small daily challenges. Everyone participates, regardless of role, and some days the whole thing takes ten minutes. The point isn’t the output. The point is the repetition. Fluency is a motor skill, and motor skills don’t care how inspiring the kickoff meeting was. They care whether you showed up today.

One rule made the difference: nobody gets left behind. The moment AI skills become optional, they become a side project for the already-curious, and the gap between your most and least fluent people turns into a canyon.

Build days. Periodically we clear the calendar and the entire company spends a full day building something real for their own role. Not a tutorial. Not an exercise. A thing they will actually use on Monday. An accountant who automated one reconciliation step in a day learns more than a month of demos can teach, because now the capability is theirs.

Role-shaped work. Fluency sticks when the work is yours. Our controllers built reconciliation workflows. Sales built call prep. I built marketing pipelines. When AI is doing your actual job alongside you, you stop thinking of it as software and start treating it like a junior teammate who needs clear instructions and a verification step. That mental shift is the whole game.

We wrote it down and gave it away. Everything we learned teaching non-technical finance people to work this way became a free public course: finance-engineer-course.tillcfo.com. Publishing it wasn’t charity. Teaching forces clarity, and clarity exposed every step we’d been hand-waving.

The Part That Surprised Me

The people who got fluent fastest weren’t the youngest, and they weren’t the most technical. They were the most curious. The teammate who asks “wait, could it also do this?” laps the teammate with the computer science background who already decided what AI can’t do.

That changed how I think about hiring, honestly. Curiosity is the qualification. The rest is reps.

If You Run a Finance Team

Your people aren’t “not technical.” They’re untrained. Those are different problems with completely different solutions, and only one of them is fixable by next quarter.

Start with one rep a day. Make it small, make it daily, and make it for everyone.

They start their mornings there now.