We’re Ellen and Simone. After 36 years in finance, we’re ready to share what textbooks won’t tell you.

💛 Welcome to CFO Playbook – your practical finance insights delivered bi-weekly. The full read will take approximately 5 minutes. Like what you see? Share it! Use the button below.

READ OF THE WEEK

Token bills jump 50% month over month. Boards ask what the return is. Most finance teams point to a leaderboard or a spend line and hope that counts as an answer. It doesn't.

Nobody has fully cracked this yet, but a pattern is emerging in what the CFOs getting it right do differently. Before we get to ROI, two things finance has to fix first: the line item and the budget logic.

1. The New Line Item That Breaks Your Forecasting

Two years ago, AI sat in the IT budget. Today every department wants its own AI budget. Marketing. Sales. Finance. HR. Operations. The line that used to live with Tech alone now shows up everywhere.

And the bills move in ways your forecasting tools were never built for. Ramp* found that a single prompt template change can triple a customer's AI bill overnight. A new feature launch can move the number by 40% in one month. This is not a normal SaaS line.

*Ramp's April 2026 AI Index, The $1 Trillion AI Spend Blind Spot. Based on AI spend data from 50,000+ Ramp customer companies. More on it in Section 5.

2. Budgets Are Going Up. Something Has to Come Down.

Most 2026 AI budgets that finance teams signed off in January are already burned through or outdated by May. CFOs are not asking whether to grow AI budgets. They're asking by how much:

• 40-50% YoY uplift is the new normal in peer conversations

• Most are revising the plan they made six months ago

• A growing share have a board mandate to AI-enable the org

The catch: AI is OpEx. If AI spend goes up, costs somewhere else have to come down, or you miss your year-end targets. In most companies, that "somewhere else" is hiring. 

One more point, and we'd argue it's the most important one in this newsletter: Don't cut AI access to save budget. 

Why:

• This is the new way of working. People need to use it daily to get fluent.

• AI is absolutely not hype, the way some still believed last year. It's the most radical shift in how we work. The pace is wild, new models and capabilities almost every day. Everyone in your company needs access to this now.

• Without access, your team falls behind. Fast.

• No sanctioned tool means personal accounts. Shadow AI, privacy risk, zero visibility.

• Top talent leaves for companies that provide access.

One distinction worth holding: don't confuse this with keeping every tool. If you run five AI tools and one has zero adoption, cut that one - that's portfolio hygiene, not cutting access. Kill the dead tool, keep the access.

3.  ROI per Use Case – The Wrong Unit

• Tools obsolete themselves every quarter. An agent built in March is often outdated by May. Three-year amortisation doesn't work on a two- to three-month tool cycle.

• AI does different work than the automation tools you knew. Old playbook: rule-based bots, same process every time, easy to measure. AI does judgment work. Drafting, analysing, synthesising. Different output every time. Measure it the old way and you miss most of the value.

• The return is real, but deferred. "ROI doesn't apply to AI" is the wrong lesson - it's the line that preceded every tech bubble. The return has to be there; what's changed is how and when you see it, not whether it shows up at all. So invest ahead of clean proof - being late to fluency costs more than being early and imperfect, but hold the bet to a horizon where the return has to show.

• And it shows up at team level, not per person. One analyst saving two hours doesn't move the needle. Ten people running every commentary, every flux, every board prep through AI does. That compound effect never shows in a single use-case calculation.

4. What Finance Teams Are Actually Tracking

Nobody has the perfect dashboard. Here's what we see working:

• Token leaderboards in the company meeting. Some roll their eyes at this. It's still one of the best early signals you have. You see immediately who's using AI and who isn't. We still hear "we don't have time for this", often from HR and Finance teams. That's a no-go argument in 2026. AI has to be a permanent part of how your team works.

• AI Days. Whole team off for a day. Everyone builds an agent or automates a workflow they hate. A classic for Finance: the monthly close process. Demo at the end. The bonding effect is real. The skeptics convert. The ideas that come out are often the most valuable. One of the highest-return moves we've seen.

• AI as a dedicated P&L line. Clean visibility on where the money goes. If your ERP can break it down by cost center, even better. If not, at least give the total AI spend its own line so you can track the curve.

5. Ramp's AI Index – Worth a Read

Ramp's April 2026 AI Index analyses real AI spend across 50,000+ customer companies. Worth a read.

• 13x growth in monthly AI token spend across Ramp customers since January 2025. Not 13%. Thirteen times.

• ~15% of total software budget now goes to AI tools at the median Ramp customer

• 50.4% of Ramp customers had paid AI subscriptions in March 2026, up from 35% a year earlier

• Anthropic (Claude) has overtaken OpenAI (ChatGPT) in business spend on Ramp's platform. Microsoft Copilot, still the enterprise default, now lets users choose between OpenAI and Claude under the hood.

(Ramp is at a $32B valuation, reportedly raising a new round at $40B, and moving into Europe this summer. They sit on a huge amount of corporate spend data, which is what makes the AI Index credible.)

🔎 CFO Watchlist

Philipp Klöckner at OMR: Beyond the AI Hype

Philipp gave his annual State of AI 2026 talk at OMR two weeks ago. Best overview right now on where AI stands and where it's going.

A few slides worth pulling:

• "It's not AI killing your job, it's the CFO." (Slide 114) Amazon, Microsoft, Meta, Oracle, Intel and others fund AI capex with layoffs.

• Anthropic has overtaken OpenAI in B2B. (Slide 50) 3 out of 4 companies spending on AI for the first time go to Anthropic.

• Real AI adoption is still tiny. (Slide 42) 99.7% of people don't pay for AI. 7 billion have never used it. If your team uses AI daily, you're a front-runner.

• Unicorns need fewer employees. (Slide 115) Headcount at $1B+ companies is steadily declining as AI absorbs operational work.

• The engineer hiring story isn't just "AI killed it." (Slide 113) Software engineer postings are climbing again since Claude Code launched in May 2025, while overall postings keep falling.

• The "AI killed hiring" story doesn't always hold up. (Slides 105-108) Much of the drop started before AI and tracks the end of the near-zero interest rate era, when companies over-hired during Covid. A lot of what we see now is that correction, not AI.

• Admin and legal automate fastest, yet legal jobs are booming. (Slides 111–112) 72% of legal leaders plan to increase headcount in H1 2026. The "AI eats jobs" story isn't playing out evenly.

👉 Keynote on YouTube. Slides. 

🌐 Finance Collective DACH

Your go-to CFO Network (by Simone)

CLOSING REMARKS

Thanks for reading 💛

Send your feedback, suggestions, or requests to feature something in future editions to [email protected]. We’d love to include your input.

Follow us on LinkedIn for more updates!

CFO Playbook reflects our personal opinions, not professional advice.