A CFO I know called recently, irritated. She had finally gotten her team onto Claude, and they used it to knock out a 27-page policy that had been stuck on her list for months. One afternoon, done. I told her that sounded like a win. “I know,” she said. “But it is not helping me.” It was obviously a raw draft nobody had touched. Padded, generic, clearly unread. They had not written a policy. They had pasted a prompt and forwarded the output.

AI made the easy part of our work free. Drafting, summarizing, building a first model, producing ten pages, all of that costs nothing now. So the value moved to the part that is still hard: cutting, deciding, checking, owning. Most people have not caught up. They are still acting like volume is the job.

What we need is not better prompts. It is a new work ethic. This is also why my corporate clients, especially the bigger teams, have started asking me to add an ethics piece to the Claude training.

In today's newsletter, I put the complaints I hear from both sides next to the new work ethic I now teach.

The first cohort wrapped last week. Participants saved real time, automated workflows they use every week, and learned as much from each other as from me.

I'm running it again. And I made it better.

Cohort 2 is four sessions, 60 minutes each, every Tuesday from July 14 through August 4. Mixed group of finance leaders from different companies. One focused topic per week. Hands-on throughout.

Early bird: $499 through June 30. $799 after. Register here.

The boss: “I could have just done it myself.” She asked for a one-page summary and got six pages of confident filler. Now she reads all six to find the two sentences that matter, then writes the page herself. The tool did not save her time. It handed her a reading assignment.

The boss: “It reads beautifully. I just do not believe it.” Fluent, well formatted, and quietly wrong in two places. The polished-but-hollow version even has a name now, workslop. Fluent and wrong is harder to catch than clumsy and wrong, and the careful people know it.

The team: “My boss built a model in Claude and told me to use it. I have no idea how it works.” A finished model arrives from above with one instruction: run with this. She cannot see the assumptions or trace the formulas, but she is the one who has to defend the numbers. The black box did not disappear. It changed hands.

The team: “He gave us AI and then asked for double.” The tools were supposed to buy time. Instead the quota went up and the deadline moved in. Faster bought more, not better.

Different chairs, same root. Nobody agreed on what good work means now that the easy part is free.

Five rules. They apply whether you are sending the work or receiving it.

1. Send one page, not five. AI made length free, so length stopped proving effort. A six-page memo used to mean someone put in the hours. Now it might mean someone typed one line and hit go. The work is the compression. If the decision fits on a page, the other five are you offloading the reading onto someone more senior.

2. Deliver the answer, not the raw material. The model takes an hour now, so handing one over is not the job. The job is the line that says what to do and why. Bring the recommendation, the three numbers it turns on, and the one risk that could flip it. Raw material is cheap. A point of view is not.

3. You deliver it, you own it. This is the one I repeat most in training. “Claude wrote it” is not a defense, it is a confession that you shipped something you did not stand behind. The tool drafts. You own the result, the numbers, and the call. Ownership does not transfer to the model, ever.

4. Do not send what you have not read. The tell is when someone asks about page two and you go quiet. If you did not read it, you did not do it, you relayed it. Reading your own output before it leaves you is the lowest bar there is, and it is the one people skip most.

5. AI drafts, you decide. The judgment is the job and it does not transfer. Let the model write the first version, run the numbers, lay out the options. Then you choose, and you own the choice. The day you hand the deciding to the tool, you have stopped being worth the seat.

Notice what is not on that list: a new tool. The ethic is not about software, it is about deciding that fast is not the same as good and then working like you believe it. The way to keep your seat in an AI world is not to out-type the machine. It is to do the judgment it cannot.

Here is the part I am most excited about. I built a Claude skill that checks any AI deliverable against the standard, tightens it to a page, confirms the sources, and coaches the person who sent it, before it ever reaches you. Build it once, share it with the team, tune it by role. It is in the subscriber section below. If you have been meaning to upgrade, do it for this one.

AI did not lower the bar for finance work. It moved it. The easy part is free now, so the standard is what you do with the part that is still hard. Send the one page. Check it yourself. Own the call. That is the whole ethic, and right now it is the most employable thing you can be.

If you want your team working this way, that is what I teach in Claude in Action, and increasingly it is the ethics piece teams ask for. Two ways to join. For the whole team, I run it as corporate training; just reply to this email, and I will set it up.

To learn alongside peers from other companies, the next mixed cohort is open for sign-up.

See you next week.

This newsletter is for you, and we want to make it as valuable as possible. Please reply to this email with your questions, comments, or topics you'd like to see covered in future issues. Your input shapes our content!

Want to dive deeper into balanced AI adoption for your finance team? Or do you want to hire an AI-powered CFO? Book a consultation!

Did you find this newsletter helpful? Forward it to a colleague who might benefit!

Until next Tuesday, keep balancing!

Anna Tiomina
AI-Powered CFO