Most AI implementations in finance start with energy and end with quiet resentment.

The boss is excited. He went to a conference, played with ChatGPT for a week, maybe built a financial model using Claude. It works well for him. So he schedules a meeting, tells the team this is the direction, and sets a deadline.

Three weeks later, nobody is using it.

This edition is about the two directions AI adoption travels in finance teams, why each one tends to break down, and what it actually takes to make either work. In the free section, I share two real cases and a comparison table that maps out the failure modes and the corrective moves for both. In the subscriber section, I get specific: two operational playbooks, one for the CFO trying to move a team toward adoption, one for the FP&A lead trying to get a skeptical CFO to take it seriously.

Before I get into this week's topic, a quick note: I published the first skill in my CFO skills package on LinkedIn last week. I'm building more of these because a pre-built skill is one of the most practical ways to share expertise.

But remember: a skill is only as good as your ability to adapt it. Every skill I build is designed to prompt the right questions and push you to customize it to your organization, processes, and role. If you missed it, here’s the link.
Download the .md file. In Claude, go to Customize→Skills→Create Skill→Upload A Skill

The second skill in my CFO skills package, dedicated to cash flow reporting and forecasting, is in the paid section.

I did not expect the corporate track to fill this fast. I have no more spots in May. This tells me the demand for structured Claude training in finance is real and growing.

The cohort starting on May 26th has five spots left and closes next Tuesday. Three sessions.

You come out knowing which of your actual finance tasks Claude handles well, how to set it up to save time on the work you repeat every week, and how far to push the automations already available to you. Most people find eight to ten hours a week, sometimes more, once they stop using it ad hoc and start using it with a system.

The group is CFOs, fractional CFOs, Controllers, and VPs of Finance. I am really excited about it. Join us here.

Both tracks reopen in July at a higher price.

If you want July instead, or are looking for something custom, schedule a conversation here or reply to this email.

I watched this play out recently at a client. The founder had spent a few hours with Claude: watched some videos, uploaded a set of old financial models, explained how he wanted them cleaned up. He got a solid output. So he sent it to his finance team and asked them to finalize it and start working with it.

The team didn't know where to start. They hadn't built the model, so they didn't trust it. They didn't know what assumptions were baked in, which outputs to check, or what a wrong answer would look like. And they had no time carved out to figure it out. Deliverables didn't pause for an AI rollout.

The CFO was the most resistant. Until she sat down with the tool herself, on a real task. Two weeks in, she told me: "I'm impressed. I didn't expect it to be this straightforward. I can see how much time we could save." That shift didn't come from a mandate. It came from doing.

This is the top-down trap. The mandate arrives before the capability does.

The bottom-up version looks different but fails just as reliably.

I talk to a lot of finance leaders who are quietly using AI on their own. Running reconciliations through Claude, building variance analysis frameworks, automating parts of the monthly close. It's working. But when they bring it to their CFO, one of two things happens. Either the CFO shuts it down: outputs can't be validated, the tool hasn't been approved, there's a concern about data security. Or the CFO is supportive in theory but doesn't allocate time, budget, or structure to scale it. The experiment stays personal. It never becomes a process.

Sometimes it doesn't die quietly. I've seen the opposite too: a team where several people started experimenting independently, each with different tools, different use cases, different prompts. Nobody compared notes. By the time leadership noticed, there were four parallel projects with no shared standards and no consistent outputs. They eventually had to standardize from scratch, which created exactly the kind of top-down resistance they had avoided the first time around.

Neither direction is safe by default. What matters is how you bridge the gap between where the initiative starts and where it needs to land.

Two weeks after I started working with that client, one member of the finance team had automated two financial processes. A third is in progress. It will take more iterations before it sticks. But the direction is clear and the team is moving.

What changed was not the tool. The tool was already there. What changed was the approach.

The table below is how I think about the two modes and where each tends to break down.

Knowing which mode you are in is the easy part. Most implementations don't fail because the team picked the wrong direction. They fail because nobody anticipated where that direction would break down. But here is the thing: it doesn't matter where you are starting from. Top-down or bottom-up, there are moves that work from either position. And in almost every case I have seen, the teams that made progress were the ones that started doing something, even imperfectly, before they felt ready.

In the subscriber section, I get into the how. Two playbooks built around real scenarios: one for the CFO trying to move a team toward adoption without losing their trust, one for the FP&A lead trying to get a skeptical CFO to take AI seriously. Both include specific steps and the no-go signs worth knowing before you start.

Plus - the second part of my CFO Skill package
Here’s Part 1 - CFO Reporting Skill Package

Part 2 - Cash Flow Reporting and Forecasting. It follows the same structure as the Financial Reporting edition: same philosophy table, same first-run detection, same four-round setup flow that walks Claude through your organization before it produces anything.

If any of this maps to something you are navigating right now, I would genuinely like to hear about it. What is the part that is stuck? Is it the mandate, the trust, the tools, the time? Hit reply and tell me. These situations are rarely identical, and I learn as much from your context as I hope you get from mine.

See you next week.

Anna

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