Claude Fable 5 launched on June 9. It topped nearly every benchmark, the public version of the Mythos-class model the whole industry spent a week calling a leap, handed to millions of paying customers for free.

But Anthropic switched the model off for everyone, worldwide. The breakthrough of Tuesday was gone by Friday night. It is the first time the US has ever issued an export control directive for access to an AI model.

And this is the week your board is asking why you are not moving faster on AI. You have heard it in the last three meetings.

The KPMG deal, the Cloudflare memo, and Anthropic’s own CFO building skills. The pressure is real and it is pointed at your seat.

That is the thing worth understanding this week.

For two years the conversation about AI in finance has been about capability.

• Which model is smartest.

• Which one closes the books faster.

This week the conversation changed into the one a CFO is actually trained for. It’s not about how good the model is.

It's about who controls it, who controls the data, and what happens when it is gone on a Monday morning through no fault of yours

The Fable suspension is the loud version of a risk that has three quieter parts. Here is what each one means for the close data and the workflows in your systems right now.

Let’s dive in.

The data rule you are about to inherit

Start with the fact that stopped Microsoft.

Every Mythos-class model, which means Fable 5 and everything Anthropic ships at that capability level, carries mandatory 30-day data retention. Every prompt, every output, and every file you hand it for context is held for 30 days on Anthropic’s servers for safety monitoring.

The reason this matters is what it replaces.

Every other Claude model, Opus 4.8, Sonnet, and Haiku, can run under a Zero Data Retention agreement, where your data is gone the moment the session ends.

Fable can’t. And if your company already signed a Zero Data Retention agreement with Anthropic, that agreement does not cover Fable traffic. The most powerful model is the one your existing contract was not written for.

Anthropic has been clear about the guardrails.

The data is not used to train future models. It is deleted after 30 days in almost all cases. The retention exists to catch novel attacks on its most capable system. Those are real mitigations and they matter.

But a CFO does not get to wave away the exceptions, because the exceptions are where the risk lives. Content flagged by the safety classifier can be held longer, up to two years in the cases Anthropic has described. And “almost all cases” is not the language a regulated finance function can build a control around.

This is why Microsoft, which is both a distribution partner for Anthropic and one of the most sophisticated buyers on earth, told its own staff to wait while the lawyers read the terms.

They were not making a statement about the model. They were doing the thing a finance function does. They read the contract before they let the data out the door.

Fable made its own decisions

Buried in Fable 5’s system card, a 319-page safety document, was a paragraph disclosing that the model would quietly downgrade its own answers for certain requests.

• If it detected that a user was doing frontier AI research, it would silently route the request to a weaker model and answer from there, without telling the user. Same interface. Same apparent answer. Lower capability, undisclosed.

Researchers found the paragraph within hours.

The backlash was immediate. And within two days, Anthropic reversed it, apologized, and said plainly that it had made the wrong tradeoff. Flagged requests now visibly fall back to a named model instead of being silently downgraded. The company moved fast and it moved in the right direction, and that part of the story matters as much as the original misstep.

But sit with what the episode revealed, because the specific policy is not the point. The point is that the most capable models now make decisions about your request on their own. They route. They fall back.

They apply safety classifiers that can reshape what you get back. Most of the time that is invisible to you, and most of the time it is benign. The lesson is not that the model is out to get you. The lesson is that a finance function cannot run on a black box it does not get to inspect.

This is the same principle finance already lives by everywhere else.

You do not accept a number you cannot trace to its source. You do not sign a statement you cannot audit. An AI model in your reporting workflow is no different. If the output that goes into a board pack came from a model that may have silently routed, downgraded, or filtered the request, and you have no record of it, you have a number you cannot fully stand behind.

The requirement is disclosure and an audit trail. The model must tell you what it did, and you must be able to show your work.

You can’t rely on one AI model

This is a procurement problem, which makes it the CFO’s problem.

Fable proved it in the cleanest way possible. A model that was approved and in production on Tuesday was gone by Friday, with no warning and no time to move off it.

There are competing stories about why it happened.

Anthropic says the government acted on a narrow jailbreak and is fighting the order. Others say Anthropic refused a fix the administration asked for. You do not need to settle that argument, because the reason it vanished does not change your problem. It vanished fast, for reasons you had no part in and no power over.

And you do not need a government order to get burned this way. Switching AI vendors now costs three to five times the annual license, and nearly three in four companies say losing an AI vendor would disrupt their core operations. Most finance teams are carrying that risk without ever having priced it.

A builder put it well the night of the ban.

He had set his whole weekend aside to build on Fable, and instead he sat down to learn local models, the kind you run on your own hardware. His lesson was that you should never build your whole workflow on something that can disappear with a single letter.

He compared it to the founders who realized too late that they did not own their social media audiences, right before everyone started building email lists they controlled. One algorithm change had once wiped out his startup’s traffic overnight, and this was the same lesson on a bigger scale.

You already run the rest of the business by this rule. You would never put treasury through one bank with no backup, or source a critical part from a single supplier. AI is no different.

So there are a few things a CFO can actually direct.

• Cap how much of your AI spend sits with any one vendor.

• Keep a second model approved and working so losing one does not freeze a workflow.

• Know which processes could move to a fallback tomorrow and which would simply stop.

• And for the ones that cannot stop, understand what it would take to run them on a model you control.

This is the CFO’s call and not IT’s, because the exposure is financial.

A model that disappears is a close that might not land, a board deadline that slips, or a process the company depends on that stops for reasons no one inside the building can fix.

Keeping the work running has always been the CFO’s job. Only the tool is new.

The Bottom Line

For two years the question about AI in finance was how good the model is. This week it became a different question, and it is the one the CFO seat was built for.

Who holds the data, what the model decides without telling you, and whether the work keeps running when the model is gone.

That is a promotion, not a problem.

It moves AI out of the category of a tool the team plays with and into the category of a dependency the CFO governs. Capability is now the easy part. Every frontier model is extraordinary, and the most extraordinary one just got switched off by a single letter. The hard part is the part finance has always owned. Terms, controls, an audit trail, and a fallback for the day the deal changes.

The CFO who runs AI like procurement can look the board in the eye and say yes, we are moving faster, and yes, I know where our data sits and what we do if a model disappears tomorrow.

The CFO who wires the close to the most powerful model because it tops the benchmarks finds out the hard way that the benchmark was never the point.

Move faster on AI. The board is right to push. Just build it so nothing essential dies when one model does.

And that’s all for today.

See you on Thursday!

Whenever you’re ready, there are 2 ways I can help you:

1. Advertise with the AI CFO Office.

2. If you’re building an AI-powered CFO tech startup, I’d love to hear more and explore if it’s a fit for our investment portfolio.

I’m Wouter Born. A CFOTech investor, advisor, and founder of finstory.ai

Find me on LinkedIn

This is not a demo. This is a 90-minute end-to-end walkthrough of how a real CFO actually uses Claude across every part of a modern finance function. Live. Ask questions. Challenge what works and what does not.

Claude Masterclass for CFOs (LIVE)

The last two weeks made the case for itself.

KPMG put Claude in the hands of 276,000 advisors. PwC trained 30,000 of theirs on Office of the CFO workflows. Cloudflare’s CEO listed finance in the Wall Street Journal as the first category AI replaces. Jamie Dimon publicly built a dashboard in Claude in 20 minutes.

Claude is the new Excel.

Every CFO reading this knows the moment arrived. The question is no longer whether to learn Claude. It is how fast.

Glenn Hopper is the real AI CFO leading this session. Decades in finance leadership. Running Claude-powered workflows inside his actual planning process today. He is also the author of AI Mastery for Finance Professionals, coming out later this year, and he advises finance teams on AI deployment through his consulting practice.

This session is the full owner’s manual.

What Glenn will cover

The Claude surfaces every CFO needs to know

• Chat for the daily questions and quick analysis

• Projects for persistent finance work with the right files in context

• Skills for the workflows that compound every month

• Cowork for collaborative work across the finance team

• Connectors for live data from your ERP, BI, and source systems

The CFO workflows running on top

• Monthly close and reforecast

• Board pack assembly and narrative

• Scenario planning and sensitivity analysis

• Audit and SOX response

• Strategic capital allocation

What Wouter will cover

The strategic layer

• How to prepare your finance function to become AI-native

• Where Claude saves time and where it does not

• What to deploy this quarter and what to wait on

• How to build a Closed Loop finance function

• How CFOs should think about the budget shift coming with premium-tier models like Fable 5

What to expect

90-minute live masterclass with Glenn walking through each surface in his actual Claude environment, joined by Wouter Born on the strategic CFO layer. Open Q&A throughout. Honest discussion of what works and what breaks.

Who this is for

CFOs, VPs of Finance, FP&A leaders, and controllers. Anyone running a finance function that has to be ready for what comes next.

Live masterclasses are exclusive to Insider members. Insider Membership is an annual plan. You get access to all the live meetings, events and Masterclasses. Also the recordings.

If you are thinking about getting the most out of the AI CFO Office, now’s the moment to upgrade to Insider Member.

Prices go up after the event (19/06) at 7 p.m. ET

• $600/year → $699/year

Your current rate stays locked in as long as you’re subscribed.

Read more

And cash is materially below forecast, and nobody on the team can tell you why from the P&L, because the P&L is the thing saying the month was good.

Warren Buffett said it cleaner than the rest of us ever have.

“References to EBITDA make us shudder. Does management think the tooth fairy pays for capital expenditures?”

• EBITDA tells a story about profit.

• Cash tells the truth about the business.

• The gap between the two is exactly what a month like this is hiding.

The answer is real, and it is somewhere in your systems. It is just scattered. A piece of it is in AR aging. A piece in the AP run. A piece in inventory, in returns, in the marketplace settlement file, in customer terms, and in last month’s board commentary.

No single report holds it. Finding it means reading a dozen files against each other, and that has always been a job for a senior person and a long week.

Until now, AI did not help much with this. It answered questions. You asked what the AR balance did; it told you. You asked it to explain a variance; it explained the one you had already found. The thinking, the part where you decide which threads to pull and how they connect, stayed with you.

But when Anthropic released Claude Fable 5, the public version of its Mythos-class model, the coverage was all about cybersecurity because that is where Mythos built its name. It can take a pile of small, unremarkable signals and chain them into a vulnerability that no single signal reveals.

Finance has the same shape. The cash miss is never one thing.

It is a channel mix that quietly shifted toward slower-paying customers, plus an AR build, plus inventory bought for the old mix, plus AP going out the door faster than planned, plus settlements that lagged, plus returns running hot, plus a forecast still running on last quarter’s assumptions.

Each one looks like noise. Chained together, they are the whole story.

That is the shift, and it has nothing to do with benchmarks.

• The old use case was: explain this variance.

  

• The new use case is: investigate why the result does not make sense, prove it with evidence, and help me decide what to tell the board.

  

The first is a chatbot but the second is a forensic investigator.

This week is the first time a model could be the second one.

So I built the test.

Let’s dive in.

How to run a forensic-finance investigation with Claude Fable

This is not a prompt. It is a skill.

A clever prompt works once. Next month you can’t trust the same rigor, because the quality was in how you phrased it that day.

A skill encodes the method. The investigation runs the same disciplined way every time, on any data pack, no matter who runs it.

The forensic-finance-investigation skill gives the model a method. Start with the cash problem. Build the EBITDA-to-cash bridge. Test revenue quality. Inspect working capital. Challenge the forecast assumptions. Review the board narrative. Separate signal from noise. End with a CFO-ready conclusion and a confidence level on every finding.

It also makes the model comparison fair.

Test two models on a prompt and you are testing which one guesses your intent. Give both the same skill and the same messy pack, and you are testing which one can actually do the work.

Install the skill, select the data pack folder in Cowork, and give it the job.

COPY/PASTE PROMPT

/forensic-finance-investigation Use the forensic finance investigation skill on the attached finance data pack. Revenue and EBITDA look acceptable, but cash is materially worse than forecast. Produce a CFO-ready investigation memo and a visual forensic finance report. Focus on root causes, quantified evidence, board narrative risks, confidence levels, and recommended actions.

Then you get out of the way and let it investigate.

Here is what came back.

Here’s the forensic-finance-investigation skill (.md) and data pack.zip; download it below

Run the same investigation on your own close this weekend.

Read more

This is not a demo. This is a 90-minute end-to-end walkthrough of how a real CFO actually uses Claude across every part of a modern finance function. Live. Ask questions. Challenge what works and what does not.

Claude Masterclass for CFOs (LIVE)

The last two weeks made the case for itself.

KPMG put Claude in the hands of 276,000 advisors. PwC trained 30,000 of theirs on Office of the CFO workflows. Cloudflare’s CEO listed finance in the Wall Street Journal as the first category AI replaces. Jamie Dimon publicly built a dashboard in Claude in 20 minutes.

Every CFO reading this knows the moment arrived. The question is no longer whether to learn Claude. It is how fast.

Glenn Hopper is the real AI CFO leading this session. Decades in finance leadership. Running Claude-powered workflows inside his actual planning process today. He is also the author of AI Mastery for Finance Professionals, coming out later this year, and he advises finance teams on AI deployment through his consulting practice.

This session is the full owner’s manual.

What Glenn will cover

The Claude surfaces every CFO needs to know

• Chat for the daily questions and quick analysis

• Projects for persistent finance work with the right files in context

• Skills for the workflows that compound every month

• Cowork for collaborative work across the finance team

• Connectors for live data from your ERP, BI, and source systems

The CFO workflows running on top

• Monthly close and reforecast

• Board pack assembly and narrative

• Scenario planning and sensitivity analysis

• Audit and SOX response

• Strategic capital allocation

The strategic layer

• How to prepare your finance function to become AI-native

• Where Claude saves time and where it does not

• What to deploy this quarter and what to wait on

• How CFOs should think about the budget shift coming with premium-tier models like Fable 5

Who this is for

CFOs, VPs of Finance, FP&A leaders, and controllers. Anyone running a finance function that has to be ready for what comes next.

What to expect

90-minute live masterclass with Glenn walking through each surface in his actual Claude environment, joined by Wouter Born on the strategic CFO layer. Open Q&A throughout. Honest discussion of what works and what breaks.

As a paid member, you get full access to all written content. Live masterclasses are exclusive to Insider members.

If you want the live experience, upgrade to Insider. If you are already an Insider member, the registration link is below.

Read more

• The audit partner who signs off your financials.

• The tax advisor who structures your year-end.

• The M&A team running diligence on your next deal.

• The internal audit consultant testing your SOX controls.

KPMG put it plainly on their own alliance page.

"Our alliance with Anthropic engineers AI directly into the platforms, controls, and workflows where professional judgment matters most."

That is the work the CFO has been paying KPMG to do for 30 years. It just got an AI layer the CFO does not have. It means the tax advisor reviews your deal structure with Claude before the meeting.

It means the M&A diligence team will read your last 12 months of board reporting through Claude before they sit down with you.

KPMG works in industries where accuracy, accountability, and trust aren’t optional, and they’re applying the same standard to AI. They’re rolling Claude out to 276,000 people across the business, and using it for client work in tax and private equity. They’re also bringing it into cybersecurity, where it helps find and fix vulnerabilities. That’s what a firm-wide commitment to AI looks like, and we’re proud to be the partner they chose.

How are you using Claude as a CFO?

It is the question the room will ask you.

Cloudflare CEO tells public which employee to fire first

After firing 20% of his employees because of AI, Cloudflare's CEO told the public which employees you should fire first. He sorted his workforce into 3 groups.

1. Builders

2. Sellers

3. Measurers

Then he cut the measurers first.

Measurers, in his words, are middle management, finance, legal, internal audit, and operations. The people who count and check and document the work instead of making it or selling it.

Every CFO reading this just felt something.

Because for the last 30 years, that is exactly how finance has been filed inside the org chart. Overhead. Back office. A measurer.

Matthew Prince is half right.

The part of finance that just measures is in serious trouble. AI doesn't sleep and doesn't need a controller to chase variance explanations across 12 entities. The part Prince is wrong about is treating finance as one undifferentiated bucket.

A finance team that only measures is overhead. A finance team that builds the operating system the company runs on is infrastructure.

There is a difference between a controller running a 12-day close and a CFO who built the data model that lets the CEO see the next quarter before it happens.

• Both technically work in finance.

• Only one of them is a measurer.

The CFOs who survive the next 36 months are not the ones defending the function. They are the ones rebuilding it from the inside.

Read the full memo here.

Anthropic’s CFO just published the playbook for what comes next

Krishna Rao runs the finance function at Anthropic, a company that scaled from $9 billion to over $30 billion in annualized revenue in roughly four months.

Krishna Rao is the CFO who runs the finance function behind that number.

• His team has built 150 Claude skills.

• Statutory financial statements for all Anthropic legal entities are produced through Claude.

• The monthly financial review is 90 to 95% ready before a human edits it.

• Weekly compute utilization reports that used to take hours now take 30 minutes.

• The biggest internal user of Claude inside the finance team is the head of tax.

Not the youngest controller or the FP&A director. But the head of tax.

The senior people, the heads of function, the people with the most institutional knowledge, are the heaviest users. Because they are the ones writing the skills.

Anthropic’s CFO uses Claude for the monthly financial review. I built MFR Skill.md every CFO can install it.

Tech giants are going public

For 10 years, the operating advice for every private company over $500M was the same. Stay private. Avoid the public markets.

That advice just expired.

Cerebras went public this month. Priced at $185. Range raised twice. Stock opened at $320. Settled around $230. Market cap: $50 to $60 billion.

Bill Gurley said the line every CFO needs to read.

“The public markets may be shifting back. A lot of the companies in our portfolios are now thinking about going public at a billion or three billion or five billion. We had this period of a decade where Andreessen was really pushing stay private forever and I see the pendulum swinging back.”

Anthropic, OpenAI, SpaceX, and the rest of the private giants absorbed all the value creation. Public investors never got to participate.

The pendulum is swinging the other way.

If you are a CFO at a private company in the $1 billion to $5 billion range, your board is already having the conversation. The honest question is not whether to IPO. The honest question is whether your finance function can survive one.

Public market scrutiny is a different animal.

Quarterly filings on a schedule that does not care how complicated your business got last quarter. Audit committees that want every control documented. Investor relations that expects answers within 24 hours of any market move.

The CFO who has been automating the close and building real-time reporting for the last twelve months is ready. The CFO running monthly closes in Excel is not.

The window is open. The CFO whose function is ready walks through it. The CFO whose function is not gets asked to step aside before the S-1 gets filed.

The Bottom Line

For 30 years, the value of the CFO seat came from being the most informed person in the room. The CFO ran the numbers. The CFO controlled the systems. Information asymmetry was the moat.

This week, every part of that moat got priced to zero.

• KPMG put Claude in the hands of every external advisor. The CFO is no longer the most informed person at the table.

• Cloudflare’s CEO told the public market that the measurement work inside the function is the first thing AI cuts. The market rewarded him for saying it.

• Anthropic’s CFO published what the alternative looks like. A function where senior people write skills, the close is 95% automatic, and the head of tax is the heaviest internal AI user.

• The IPO window reopening is the test. Public markets sort the functions that built the architecture from the ones that did not.

The CFO seat is being re-priced from informational expert to architectural author. The CFO who controlled the data is being replaced by the CFO who builds the system the data flows through.

Most CFOs reading this got their seat under the old pricing.

The technical accounting, the model fluency, and the relationships with auditors and bankers. Those skills are still necessary. They are no longer sufficient.

Three questions worth sitting with this weekend.

1. What part of your seat’s value today is information the room no longer needs you for?

2. What part of your seat’s value comes from systems your team built that the room cannot get anywhere else?

3. Which number is bigger?

The CFO who can answer those three questions honestly already knows what the next 12 months have to look like.

And that’s all for today.

See you on Thursday.

Whenever you’re ready, there are 2 ways I can help you:

1. Advertise with the AI CFO Office.

2. If you’re building an AI-powered CFO tech startup, I’d love to hear more and explore if it’s a fit for our investment portfolio.

I’m Wouter Born. A CFOTech investor, advisor, and founder of finstory.ai

Find me on LinkedIn

The answer is yes. And I’m going to show you exactly how.

I built a full monthly planning system for a real SaaS company inside Claude. A planning model with customer metrics, headcount planning, full financials, and a balance sheet.

Wait until you see the dashboard at the end. Because Claude handled the entire reforecast in the same Project, it had full context for every chart, every narrative, and every scenario. It's the most comprehensive planning dashboard I've built so far

I took CoolDashboards, LLC monthly planning model, uploaded it to a Claude Project, and ran the full FP&A cycle. Actuals loaded. Forecast updated. Assumptions pressure-tested. Scenarios built. Dashboard created.

In today’s edition, you’ll see how to set up a Claude Project as your persistent FP&A workspace, load a real planning model and new actuals, spot anomalies against the plan, update the forecast, run scenarios with sensitivity analysis, and save the entire workflow so next month takes half the time.

No code, no API or any technical skills needed

Let’s dive in.

How to do financial modeling with Claude as a CFO

Read more

Most CFOs are running an open loop.

You walk into the board meeting. You answer the questions you predicted. You miss two questions you didn’t. You walk out, debrief with the controller, write down what to fix for next quarter, and the note gets buried in your inbox.

That is the open loop most CFOs were trained on. It is also why CFO turnover hit a decade high in 2025.

The closed loop is the opposite. Every board meeting produces a record. Every record produces a pattern. Every pattern compounds into the next prep. The CFO at meeting 50 walks in measurably sharper than the CFO at meeting 1.

I’m sharing with you the full CFO Skill Stack.

1. CFO Brain is the new master skill. Frameworks, voice, playbooks for covenants, M&A, audit committee, AI capex, capital allocation, FP&A standards, and the CFO-to-CEO career roadmap. Talks to you like a Fortune 500 CFO, not a chatbot. Pulls live data from FactSet, S&P Global, and PitchBook when you need it. Refreshes itself every month.

2. CFO Memory is the pattern layer. Captures recurring directors, lenders, auditors, CEO patterns, and deal patterns. Writes itself at the end of every meaningful session. The next session starts with the patterns already loaded.

3. CFO Board Prep is the skill you already have. It predicts director questions, builds evidence-backed answers, identifies missing support, drafts CFO opening narratives, and creates the pre-board risk register. If you missed it last month, the link is at the bottom.

4. CFO Audit Trail is the formal record. Logs every meaningful output with timestamps, sources, confidence grades, and decision rationale. Aligned to PCAOB AS 1215, AICPA AU-C 230, SOX 302/404, COSO 2013. The audit committee respects it. Outside counsel reads it. SEC enforcement defense lawyers ask for it.

You get four files and they are self-improving.

Let me show you exactly how to install them, then walk through the use case I want every CFO reading this newsletter to run before their next board meeting: building a consistent board cockpit dashboard.

Let’s dive in.

Step 1: Create the CFO Brain project in Claude

Go to claude.ai.

Click Projects in the left sidebar.

Click New project. Name it CFO Brain.

Tell Claude what you are trying to achieve. Paste this in the project description:

I'm a CFO. I need a strategic thought partner who thinks like a 
seasoned Fortune 500 CFO. Use the CFO Brain skill, the CFO Board 
Prep skill, the CFO Memory skill, and the CFO Audit Trail skill 
to respond to my questions, analyze my data, push back on my 
framing when I'm solving the wrong problem, predict director 
questions for board meetings, remember patterns across sessions, 
and log every meaningful output to a defensible audit trail.

Click Files and upload all four SKILL.md files.

Download the full CFO SKILL Stack below.

Read more

He walked us through his actual monthly reforecast workflow inside a Claude Project. Live. With a real dataset. Every Insider Member found it super helpful.

The full recording is now available to Insider Members.

A complete monthly reforecast cycle inside a Claude Project, his exact custom instructions, 5 copy/paste prompts, the full dataset, and his honest take on where Claude saves time and where it doesn’t.

We do not share these recordings publicly.

Insider sessions are a small group on purpose. The conversation is open-camera, open-mic, real questions from CFOs and FP&A leaders who do not want their workflows broadcast to the open internet.

That is what makes them work. People ask the questions they would not ask on LinkedIn.

If you are a paid member and want access to the recording and the next live session, you can upgrade to Insider.

Read more

A messy reconciliation shows the cutoff checklist is weak. A variance that shows up two months running tells you which signal you should have caught earlier. A board member asks the one question your deck did not answer, and you make a mental note to fix it next time.

Then next month starts, and you fix nothing. The note is gone. The lesson stayed in your head, and your head is full.

That is the finance loop almost every company runs. The CFO learns. The system forgets. And the team relearns the same lesson it already paid for.

I wanted to see if Claude could close that gap.

Not remember what happened.

But change what happens next.

Here is the difference that matters.

• Closed-loop finance remembers the last cycle.

• Self-improving finance rewrites the next one.

A normal AI workflow gives you an output and stops. A close memo. A variance analysis. A board narrative. Useful, and done.

A self-improving workflow gives you the output and then asks one more question. What should change so next month runs better?

That question is the entire game.

Every close is hiding signals about your own process. A late invoice is telling you the cutoff checklist has a hole. A recurring inventory issue is telling you the review logic is incomplete. A follow-up you have chased three months in a row is telling you the workflow keeps teaching the same lesson because nobody wrote it down.

You already notice these things. The problem is turning them into operating logic before they evaporate.

So I built it with Claude.

Below, I am sharing the following:

• Full Self-Improving Finance System Pack.

• The Claude skill pack that teaches Claude to run the review, the quality gate, and the learning cycle. The decision log that becomes your structured memory and approval layer.

• The starter folder with the close evidence structure already built.

• And the exact prompt that runs the whole thing and hands you CFO-approved learning candidates at the end.

Let’s dive in.

Build a self-improving finance function with Claude

Before the install, understand the shape.

The system has three layers and each one has a job.

1. Google Drive is the evidence layer. Close packages, forecasts, board materials, and review notes live here. The whole finance team can reach them.

   

2. Notion is the decisions layer. Every approved decision and every approved learning lives in a structured log. This is the memory. It is the part that compounds.

3. Claude Cowork is the operator. It reads the evidence, reads the prior decisions, runs the close review, and proposes the improvements. It never writes to the decisions layer on its own. That is the human-in-the-loop boundary.

Download the self-improving finance skill pack.zip.

Read more

Dynamic workflows

This new feature allows Claude to take on even bigger tasks in Claude Code.

Claude can plan the work and then run hundreds of parallel subagents in a single session (and with Opus 4.8, the agents can run for even longer). It then verifies its outputs before reporting back to the user.

For example, Claude Code with Opus 4.8 can now carry out codebase-scale migrations across hundreds of thousands of lines of code from kickoff to merge, with the existing test suite as its bar. You can read more about dynamic workflows.

Available in Claude Code for Enterprise, Team, and Max plans.

Effort control in Claude Web and Cowork

A new control alongside the model selector lets users choose how much effort Claude puts into a response.

On higher effort settings, Claude will think more frequently and more deeply to give you better responses.

On lower effort settings, Claude will respond faster and use up a user’s rate limits more slowly. Users now have this choice… the effort control is available on all plans

Claude is no longer one tool with one speed

Claude is a router. Low effort for a quick first read. High effort for proper variance diagnosis. Max effort for evidence-backed board synthesis. The CFO decides which dial to turn for which job.

This is the move from prompts to workflows.

The 2024 CFO opened ChatGPT and asked a question. The 2025 CFO built a Claude project and stored some context. The 2026 CFO routes finance work across effort levels and parallel agents, deciding when speed is enough and when deeper reasoning is worth the extra time.

That last one is the new operating model.

I’m testing Claude Opus 4.8 with a simple Born2Cycle finance pack through all three effort levels with the same prompt. You will see how the output evolves from a basic summary into a CFO-level interpretation and finally into a board-ready narrative with evidence, risks, open questions, and explicit caution on what not to overclaim.

The lesson is not the tool. The lesson is the routing.

CFOs can build the board-ready financial analysis with Claude Opus 4.8.

Let’s dive in.

How to build a board-ready financial analysis with Claude Opus 4.8

Go to Claude.ai.

Select Opus 4.8 as a model. Pick Low effort.

Upload these 3-month end close files

1. Born2Cycle_Mar2026_Board_Finance_Pack.xlsx

2. Born2Cycle_Feb2026_CFO_Commentary.md

3. Born2Cycle_Mar2026_CEO_Note.md

COPY/PASTE PROMPT

Read more

The AI rewrites the close. It changes an assumption nobody approved. It quietly updates a forecast model, and the audit committee finds out three months later when the number does not tie.

That is not what self-improving finance means.

Or it is not what it should mean.

Self-improving finance does not mean the AI rewrites your process behind your back. It means every finance cycle produces proposed improvements, and the CFO decides which ones become part of the operating system.

• The AI proposes.

• The CFO disposes.

• Nothing changes without a sign-off.

  

Here is the problem that the system solves.

Every close, your team learns something. A variance pattern that keeps recurring. A cleaner way to frame the e-bike margin story for the board. A reconciliation fix that saved two hours. A judgment call about when an inventory build is a bet and when it is a warning.

And every close, that learning evaporates. It lives in the controller’s head. It lives in a Slack thread. It lives in a review meeting nobody wrote down. Next close, you re-learn the same lesson from scratch.

The people get smarter. The system does not. And when the controller takes the job in Denver, his judgment walks out the door.

That is the open loop most finance functions run.

The learning itself never compounds.

The self-improving finance function is the opposite.

Every cycle it captures what was learned as a set of candidates. The CFO approves the ones worth keeping. They go into a decisions log the next cycle reads before it does anything else. The function compounds. The judgment stays in the building even when the people leave.

I just built it on three tools and a SKILL pack.

1. Google Drive holds the evidence.

2. Notion holds the decisions and the learnings.

3. Claude Cowork is the operator that runs the close, proposes the improvements, and stops to ask the CFO which ones to keep.

4. +self-improving finance skill pack.zip.

I am sharing the full install below for every CFO who wants to build an AI native finance function that improves itself.

Your CEO and the Board will be really impressed.

Let’s dive in.

How to build a Self Improving Finance function as a CFO

Before the install, understand the shape.

The system has three layers and each one has a job.

1. Google Drive is the evidence layer. Close packages, forecasts, board materials, and review notes live here. The whole finance team can reach them.

   

2. Notion is the decisions layer. Every approved decision and every approved learning lives in a structured log. This is the memory. It is the part that compounds.

3. Claude Cowork is the operator. It reads the evidence, reads the prior decisions, runs the close review, and proposes the improvements. It never writes to the decisions layer on its own. That is the human-in-the-loop boundary.

Download the self-improving finance skill pack.zip.

Read more

They shipped 15 small business agents. 8 of them for finance.

Invoice chaser. Payroll planner. Month-end closer. Margin analyzer. Tax-season organizer. Contract reviewer. Lead triager. Content strategist. Campaign builder. Business pulse.

Each one is built the same way:

Skills, connectors, workflows.

Skills are the tasks. Connectors are the data. Workflows are the orchestration.

This is the same architecture I’ve been sharing with you in the last letters.

The CFO Skill Stack uses the exact three building blocks. Anthropic just made it the default for 36 million US small businesses.

Anthropic's CFO uses Claude for the monthly financial review

His team built a library of finance skills. Over 70 of them. They built one skill called MFR. The monthly financial review skill. In Krishna's own words:

"It can produce our monthly financial review. It's 90 to 95 percent ready and then all of our discussion becomes about what do we do, what are the implications, not what exactly happened. Because Claude is not just reporting the weather. It's also helping to think about drivers and why the number changed the way it did."

Krishna Rao runs finance for a company that went from $9 billion to over $30 billion in run rate revenue in four months.

The numbers are remarkable.

Statutory financial statements for every Anthropic legal entity are now produced with Claude. A human checks the output. Claude does the work. A weekly revenue or compute utilization report that used to take hours now takes 30 minutes.

That is what the closed-loop MFR looks like when it works.

The finance leader who treats Claude as a calculator is going to be sitting across from a finance leader who treats it as a reasoning engine.

I built the version of the MFR skill for every CFO.

The Closed Loop Finance Function

CFOs can use AI agents and the CFO Skill Stack to build an AI-native company without buying more AI tools. They need to run finance as a Closed Loop. AI improves when it gets feedback. Your finance function improves when it gets feedback. Both stop improving the moment the loop is open.

Most CFOs run an Open Loop. You decide. You execute. You don't measure the outcome. The variance to plan is a postmortem. The lesson dies in your inbox.

That is the operating system most finance functions still run in 2026. It worked in 1996. It even worked in 2016. It does not work now.

AI improves when it gets feedback. Your finance function improves when it gets feedback. Both stop improving the moment the loop is open.

A Closed Loop changes that.

Every decision produces an artifact. Every artifact is legible to AI. The system continuously monitors its output and adjusts.

Decide. Execute. Log. Measure. Learn. Improve.

This weekend I built one on a sample company.

Google Drive holds the evidence.

Close packs. Board decks. GL exports.

Notion holds the structured decision memory. Every material decision logged with date, owner, and category.

Claude Cowork is the operator. It reads the evidence. Pulls the memory. Drafts the close review. Writes new decisions back into Notion at the end of every session.

Most CFOs have 80% of their institutional knowledge sitting in a senior controller's head. That is an open loop. The CFO who builds a closed loop captures it before that controller retires.

The CFO running an open loop is replaceable.

The CFO running a closed loop is not.

The most important CFO story of the year

OpenAI’s CFO Sarah Friar is fighting Sam Altman over how much money OpenAI is spending on AI.

The CFO of the company that built ChatGPT is worried the company is spending too much on data centers and is not generating enough revenue to justify the contracts she has signed.

She wants discipline. He wants scale.

Every CFO is living a smaller version of this exact fight inside their own company right now. The CEO sees AI as the future and wants to spend. The CFO sees the contracts, the burn, and the lack of measurable return.

And the board is watching both.

Friar's job is the hardest CFO job on Earth this quarter. She is funding a buildout that her CEO believes in faster than her balance sheet can defend while preparing the company for an IPO that will price every dollar of unproven AI capex.

Three lessons from how Friar is handling it.

She is not pretending the math works.

She is publicly disagreeing with the CEO. That is the discipline finance is supposed to bring. She is asking for accountability on the contracts she signed.

Every CFO with an AI budget must do the same.

• What did we sign for?

• What are we using?

• What is it producing?

She is preparing the S-1 anyway.

Internal disagreement does not pause the work.

The IPO clock keeps moving.

The Bottom Line

CFOs must stop seeing AI as a tool and start using AI as an operating system. Let every workflow, every decision, every process flow through an intelligent layer that learns and improves on its own.

Oracle paid $29.7 million to hire its first CFO in 12 years.

Every CFO must understand what’s happening here.

Oracle didn't have a standalone CFO since 2014.

Safra Catz held the CEO and finance officer together. For 12 years, that worked. Then they committed $50 billion to AI infrastructure. Took on $100 billion in debt. And suddenly one person couldn't hold both jobs

They hired Hilary Maxson from Schneider Electric. An energy company. Bloomberg's analyst said the choice signals Oracle's priority is building physical things now, not software.

Her package from the SEC filing:

$950K base salary.
$2.5M target bonus.
$26M equity grant.
$250K relocation allowance.

Companies are willing to pay massive money for CFOs right now.

But only if the CFO can do more than close the books.

Oracle needed someone who could manage a capital buildout bigger than most companies' entire revenue.

The CFO job description from five years ago is dead.

And that’s all for today!

See you on Thursday.

Whenever you’re ready, there are 2 ways I can help you:

1. Advertise with the AI CFO Office.

2. If you’re building an AI-powered CFO tech startup, I’d love to hear more and explore if it’s a fit for our investment portfolio.

I’m Wouter Born. A CFOTech investor, advisor, and founder of finstory.ai

Find me on LinkedIn

It reads the financials you upload. It identifies the contradiction in the month. It builds the visual reasoning with cards, tables, and charts so you can see what changed and why. Then it stops.

Most AI tools answer for you. This one asks you.

The skill refuses to produce the MFR deck until you answer four guided multiple-choice questions, each grounded in the evidence the skill built first.

• Is the revenue driver organic demand, pulled-forward promotion, or one-time mix? Is the margin compression structural, transitional, or strategic?

• Is the cash movement seasonal, operational, or intentional?

• Every question forces a framing decision into the open before the deck gets built.

The MFR Skill is part of the Closed Loop Finance methodology.

Every month produces a record. Every record produces a pattern. Every pattern compounds into the next cycle. The CFO at MFR 12 is running a measurably sharper review than the CFO at MFR 1.

This works because the skill is doing what Anthropic’s own CFO described.

"It can produce our monthly financial review. It's 90 to 95 percent ready and then all of our discussion becomes about what do we do, what are the implications, not what exactly happened. Because Claude is not just reporting the weather. It's also helping to think about drivers and why the number changed the way it did."

I tested the MFR Skill on Born2Cycle, a synthetic retail dataset that mirrors a real mid-market finance function. Everything below runs inside Claude Web. No external connectors. No additional tools. Just one project, one skill, two files, and four cycles.

I am sharing the full walkthrough and the MFR Skill below.

Let’s dive in.

How to run a monthly financial review with MFR Skill.md in Claude

Read more

Krishna Rao is the CFO who runs the finance function behind that number. Last week on the Invest Like the Best podcast, he talked about how his team uses Claude to do their own monthly close.

The numbers he gave are remarkable.

Statutory financial statements for all of Anthropic’s legal entities are now produced using Claude. A human checks the output. Claude does the work.

The team has built a library of finance-specific skills, more than 70 of them at last count, sitting in a shared repository that everyone can access.

On top of that library, they built one skill called MFR. The monthly financial review skill. Krishna’s quote on it.

It can produce our monthly financial review. It’s 90 to 95% ready and then all of our discussion becomes about what do we do, what are the implications, not what exactly happened. Because Claude is not just reporting the weather. It’s also helping to think about drivers and why the number changed the way it did.

A weekly revenue or compute utilization report that used to take hours now takes 30 minutes. The biggest internal Claude user inside the finance team is not the 22-year-old who joined with a coding background. It is the head of tax.

That is what the closed-loop MFR looks like when it works.

I built the version of it that every CFO must install.

One MFR skill installed inside Claude. It reads the actuals. It surfaces the signals. It builds the visual reasoning. Then it refuses to produce the deck until the CFO answers four guided questions, each grounded in the evidence the skill built first.

I tested it on Born2Cycle.

Krishna’s MFR skill is built on Anthropic’s books. I built on a CFO-grade synthetic dataset that mirrors a real retail finance function. The architecture is the same. The reasoning workflow is the same. The output shape is the same.

I am sharing the full install below.

Let’s dive in.

Step 1: Create the Monthly Financial Review project in Claude

Go to claude.ai.

Click Projects on the left, then click New Project.

Read more

It was one of the most practical AI-in-finance sessions we have hosted.

Paul built a 3-statement deferred revenue schedule live, in Excel, using Claude. He showed the new Copilot “plan mode” that just shipped. He walked through where AI saves real time in modeling and where it still breaks. He answered every question on the table.

The full recording is now available to Insider Members.

We do not share these recordings publicly. Insider sessions are a small group on purpose. The conversation is open-camera, open-mic, real questions from CFOs and FP&A leaders who do not want their workflows broadcast to the open internet.

That is what makes them work. People ask the questions they would not ask on LinkedIn. Paul answered the questions he would not answer in a public webinar.

The recording stays with the people who were in the room and the Insider Members who paid to be part of this private space.

If you are a paid member and want access to the recording and the next live session, you can upgrade to Insider.

If you are an Insider Member, use the full recording below.

Read more

Tuesday, May 19, 11:00 AM ET.

He is going to walk through it live.

Here’s what he is covering:

1. AI and Excel in real model-build work.
2. What works in finance modeling and what breaks.
3. Where AI saves real time and where the hype outpaces reality.
4. The lessons from a year of testing AI in real models with Claude shaping the output.

This is not a demo. Not a webinar you half-watch with the tab muted.

You can ask questions. You can challenge him. Open Q&A throughout.
For CFOs, VPs of Finance, FP&A leaders, and controllers. Anyone who owns the monthly cycle.

The group is kept small on purpose.

Who this is for

CFOs, VPs of Finance, FP&A leaders, and controllers. Anyone responsible for the monthly cycle.

What to expect

Live walkthrough.

Open Q&A.

Honest discussion of what works and what breaks.

If you want the live experience, you can upgrade to become an Insider Member. If you’re already an Insider Member, the registration link is visible to you below.

Read more

He is going to walk through it live.

What he is covering:

1. AI and Excel in real model-build work.
2. What works in finance modeling and what breaks.
3. Where AI saves real time and where the hype outpaces reality.
4. The lessons from a year of testing AI in real models with Claude shaping the output.

This is not a demo. Not a webinar you half-watch with the tab muted.

You can ask questions. You can challenge him. Open Q&A throughout.
For CFOs, VPs of Finance, FP&A leaders, and controllers. Anyone who owns the monthly cycle.

The group is kept small on purpose.

Registration is for Insider members only.

Who this is for

CFOs, VPs of Finance, FP&A leaders, and controllers. Anyone responsible for the monthly cycle.

What to expect

Live walkthrough.

Open Q&A.

Honest discussion of what works and what breaks.

If you want the live experience, you can upgrade to become an Insider Member. If you’re already an Insider Member, the registration link is visible to you below.

Read more

Q1 closes. The controller pulls actuals. FP&A rebuilds the model. You write the updated commentary. The deck goes to the board. The new forecast goes into the planning folder.

And three weeks later, Q2 closes and the team rebuilds the model from scratch.

That is the open loop most CFOs were trained on.

The closed loop is the opposite.

Every reforecast produces a record. Every record produces a pattern. Every pattern compounds into the next cycle. The CFO at reforecast 12 walks in measurably sharper than the CFO at reforecast 1.

Today I am going to walk you through the closed-loop reforecast end-to-end, using a real dataset. I am sharing the full Closed Loop CFO Skill Stack that runs it.

You get four files and they are self-improving.

Let me show you exactly how to install them, then walk through the reforecast.

Every CFO reading this newsletter must run this.

Let’s dive in.

Step 1: Install the Closed Loop CFO Skill Stack

Go to claude.ai.

Click Customize in Claude.

Go to Skill Manager and click Create Skill then Upload skill.

Tell Claude what you are trying to achieve.

Paste this in the project description:

I'm a CFO. I need a strategic thought partner who thinks like a 
seasoned Fortune 500 CFO. Use the CFO Brain skill, the CFO Board 
Prep skill, the CFO Memory skill, and the CFO Audit Trail skill 
to respond to my questions, analyze my data, push back on my 
framing when I'm solving the wrong problem, predict director 
questions for board meetings, remember patterns across sessions, 
and log every meaningful output to a defensible audit trail.

Click Files and upload all four SKILL.md files.

Download the full Closed Loop CFO SKILL Stack below and the Reforecast Dataset.

Read more

Companies are buying enterprise AI licenses faster than finance teams know what to do with them. This is the session for that gap.

This is not a demo or a boring webinar.

You can ask questions and challenge.

Insider Session: How to use AI in Excel and Claude with Paul Barnhurst (LIVE)

Paul Barnhurst is The FP&A Guy. Microsoft MVP. Host of FP&A Unlocked. Co-host of The Mod Squad. He’s spent 15 years inside FP&A and the last few teaching finance teams worldwide how to model better. Now he runs AI-powered workflows inside his actual model-build process. He’s going to walk you through it live.

What Paul will cover:

• AI and Excel in real model-build work.

• The lessons from a year of testing AI in real models on The Mod Squad.

• Live examples across the platforms he uses, with Claude shaping the output.

He’ll get into:

• What works in finance modeling and what breaks.

• Where AI saves real time and where the hype outpaces reality.

• Open Q&A throughout.

Who this is for

CFOs, VPs of Finance, FP&A leaders, and controllers. Anyone responsible for the monthly cycle.

What to expect

Live walkthrough.

Open Q&A.

Honest discussion of what works and what breaks.

As a paid member, you get full access to all content. The only difference is live sessions. Those are exclusive to Insider members.

If you want the live experience, you can upgrade to Insider. If you’re already an Insider, the registration link is below.

Read more

Q1 closes. The controller pulls actuals. FP&A rebuilds the model. You write the updated commentary. The deck goes to the board. The new forecast goes into the planning folder.

And three weeks later, Q2 closes and the team rebuilds the model from scratch.

The assumption that drove the Q1 reforecast is now lost in someone’s Slack thread. The reasoning for the marketing cut is in a Loom recording nobody rewatches. The decision to hold rent at the reduced rate is documented nowhere except in your own memory.

That is the open loop most CFOs were trained on.

Every cycle the CEO calls the quarter strong before you have read the data. Every cycle the assumption that drove the reforecast lives in a senior controller’s head and nowhere else.

Every cycle a store manager submits a forecast at plus ten percent to actuals and nobody catches it because the pattern from last quarter is in nobody’s spreadsheet. The board asks for a scenario you did not build. The auditor pulls the file six months later and asks how you reached the partial recovery assumption. You remember the logic.

You can’t produce it. This is the reforecast cycle in 2026. Hours of judgment lost in weeks of rework. Every quarter.

The closed loop is the opposite.

Every reforecast produces a record. Every record produces a pattern. Every pattern compounds into the next cycle. The CFO at reforecast 12 walks in measurably sharper than the CFO at reforecast 1.

Today I am going to walk you through the closed-loop reforecast end-to-end, using a real dataset. I am sharing the full Closed Loop CFO Skill Stack that runs it.

You get four files and they are self-improving.

Let me show you exactly how to install them, then walk through the reforecast.

Every CFO reading this newsletter must run this.

Let’s dive in.

Step 1: Install the Closed Loop Finance CFO Skill Stack

Go to claude.ai.

Click Customize in Claude.

Go to Skill Manager and click Create Skill then Upload skill.

Tell Claude what you are trying to achieve.

Paste this in the project description:

I'm a CFO. I need a strategic thought partner who thinks like a 
seasoned Fortune 500 CFO. Use the CFO Brain skill, the CFO Board 
Prep skill, the CFO Memory skill, and the CFO Audit Trail skill 
to respond to my questions, analyze my data, push back on my 
framing when I'm solving the wrong problem, predict director 
questions for board meetings, remember patterns across sessions, 
and log every meaningful output to a defensible audit trail.

Click Files and upload all four SKILL.md files.

Download the full Closed Loop Finance CFO SKILL Stack below and the Reforecast Dataset.

Read more