Nitin Madan could have said yes too.

For more than three years he was CFO of a global bank’s India operations center, a hub of about 30,000 people.

His finance team had AI live in the close, in reconciliations, and in reporting. Then he scored his own function on a maturity model he built. The average came out at 1.7 out of 5.

Just using AI is not a level.

Before becoming the CFO at a top bank. He spent around 20 years as a consultant helping companies transform finance. He finished his tenure on September 30 (last month) and is back in advisory work. Here is his model, how to score yourself on it, and the story behind his 1.7.

Let’s dive in.

CFO AI Maturity Model

Nitin builds the model because fellow CFOs keep asking him the same questions at roundtables. His own team shows the problem. It has five use cases live and a few more in progress.

So yes, finance uses AI. But where are you in the AI journey?

His team can’t use AI for forecasting. It can’t use RAG or MCP. A handful of live tools tells him nothing about how mature the function really is.

So he scores it on technical parameters, the building blocks of AI in finance. He tests the model for a week with a few CTOs and CFOs to make sure it resonates.

The model has 12 building blocks in five groups.

Each has 5 levels:

1. Initial

2. Emerging

3. Defined

4. Governed

5. Optimized.

How to score your team

Go through each row, pick the level that describes your team today, and average the 12 scores. Here is each row from level 1 to level 5, in the words of the model.

Models

1. Model routing. One model for every task. Teams pick models manually. Rules route close and AP tasks. Work is auto-routed, and the cost per task is tracked. Live routing against budget and service levels.

2. Local LLM. All data goes to public APIs. Pilots on non-sensitive data. Approved for defined use cases. Payroll and GL run on private models. A hybrid mix, with lock-in risk priced in.

3. Frontier model. The default for every task. Usage is known but not governed. Reserved for complex judgement. Value per call is measured. Spend is capped to the value delivered.

Customization

4. Fine-tuning. Ad hoc tests with no ROI case. Tried before RAG or tools. Only for niche domains such as tax. Retraining cost is in the run rate. Managed as assets with an ROI.

5. RAG. Context is given by hand. Isolated team RAGs. Isolated team RAGs, but access controlled. Cross-team RAGs. Grounded across ERP, EPM, and the data lake.

Capabilities

6. MCP. No MCP, manual copy and paste. Manual public link integration. Standard application integration. Custom integration that can transform data. Cross-team integration with transformation and auditability.

7. Tool use. Text answers only. AI drafts and humans key in. AI runs queries and reconciliations, with approval. AI posts low-risk entries, logged. Straight-through processing.

Agentic

8. Agent-to-agent. No links between agents. Isolated agents per team. Hand-offs within finance. Finance, tax, and legal agents work together. An enterprise agent network.

9. Agent memory. Every session starts at zero. Saved prompts and templates. It remembers policies and the chart of accounts. Governed retention and access. Institutional knowledge is reused.

10. Multi-agent. One model attempts everything. Single-purpose bots. An agent team runs the month-end close. Record-to-report, procure-to-pay, and order-to-cash run with checkpoints. Humans handle exceptions only.

Controls

11. Evals. Sample spot checks. Accuracy tests before go-live. Regression tests on every change. Evals tied to audit evidence.

12. Guardrails. Minimal input and output controls. Policy only, and shadow AI exists. PII rules, limits, and a kill switch. SOX-aligned controls, tested. Policy as code, enforced automatically.

When Nitin did this for his own function in a spreadsheet, some areas came out at 1, some at 2 and one at 3. The average was 1.7.

Then comes the part most people skip.

You do not need to reach level 5 everywhere.

Align it with your enterprise AI strategy. Is it possible? Is it not?

Then look at the business case and what it will cost.

He also said the model will keep evolving.

Next, he plans a version through an enterprise CFO lens.

Here is the story behind his 1.7.

AI failed in the board reporting

Nitin split his journey into stages, starting with board reporting.

In year one, board numbers came about a week after the hard close. Data came from many systems. He mentioned a survey that puts the average company of that size at 897 applications.

His bank ran more than 500, and finance owned about 300 of them. Everything had to flow into a finance data store after reconciliation before anyone could build the pack.

The next step was an analytics tool on top of that data store, with fixed slide formats and a rules engine that wrote the commentary. That cut the work to about a day. But the story was weak. In his words, it was a mathematic commentary.

Generative AI looked like the fix.

The team designed the whole solution. Then it stopped. All finance data at the bank was tagged as secret because it was a listed, regulated company. The security chief would not approve sending it to an AI tool.

“The tools were there, the use case was there, but it did not see the light of the day.”

Where AI did work

The wins came where the bank did not need to feed in the full data of an entity or a business. Nitin listed them:

• Anomaly detection as part of the close, GL sign-off, and liquidity reporting

• Aggregating commentary from around 20 entities into one readable story for the annual report

• Intercompany matching and reconciliation

• Treasury hedge spreadsheets and other Excel and macro tools, using AI to help with code

• Reconciling regulatory reporting to financial reporting

• Explaining risk-weighted assets

He also made a point I had not heard before. Early on, the best user of AI in financial services was the regulator.

“They get all the data in a very clean format.”

They used it to spot gaps between a bank’s reports and send back the errors. So the bank started using AI to find those gaps first.

What stayed off limits: forecasting, management reporting, and anything reported directly to the regulator.

Below, for paid subscribers:

• How to keep a human involved at every step.

• Why AI is a new data risk.

• Why build on RAG.

• How to budget AI.

• How much of the forecast AI can take

• How to start adoption and what to do this quarter.

Read more

He didn’t give us the polished version.

He told us what worked, what got blocked, and why.

Here are a few things he shared:

• Payback takes longer than you’d hope. In his experience, AI’s productivity gain in finance was not overwhelming, and the payback ran about three years.

• Security is the real gate. In a regulated company, finance data is often classified secret, and that alone can stop an AI design from ever going live.

• Where AI does deliver: intercompany matching and reconciliation, anomaly detection in the close, and pulling commentary from many entities together into one story.

• Make this board meeting boring. That’s what he tells his controllers. His rule for AI board decks is that they should be AI-populated, not AI-generated.

• Hold a no-AI day so your team doesn’t forget how to work without it.

Below, members get the full recording, his maturity model, and timestamps so you can jump straight to the part you need.

Read more

The scarce skill now, they say, is

“Designing the operating system the company runs on.”

I agree with half of that sentence.

The operating system part is right.

But the “new CFO” part is backwards.

I believe the new CFO is not the answer if you don’t have an operating system.

What you must do instead is keep your CFO.

And build an AI Operating System.

Read on.

Where a16z nails it

a16z describes a finance person “who has become proficient in Codex or Claude Code and leads the charge on building tools for the team.”

Their best example comes from Anthropic.

The finance team there maintains a library of 70+ finance-specific AI skills they treat like production code.

One skill produces a 90-95% complete monthly financial review in about 30 minutes. That is the right mental model: skills you version, test, and reuse, rather than prompts you paste.

They’re also right about a trusted chain from every number to its source, about continuous planning, and about who gains most from AI: at Anthropic, the heaviest user on the finance team is the head of tax.

You need a new CFO; it sounds like you need a16z portfolio

The finance leaders in the piece run Anthropic, OpenAI, and a16z, growth-stage companies. The tools named are seven a16z portfolio companies: Rillet, Concourse, Lio, Stuut, Sphere, Petual, and CFO.ai.

That’s fair game, and a16z labels them openly as a few examples from its portfolio. But the playbook comes from companies born on a clean stack.

Most CFOs live on an ERP older than some of their analysts, auditors who sign the opinion, a board that expects its pack every quarter, and SOX controls that took years to earn.

Your CFO holds the context

An incumbent’s greatest untapped advantage is 30 years of proprietary data, operating experience, and domain expertise.

The rule for every AI decision fits on a sticky note:

“Own the context; rent or replace the technology.”

In finance, context means knowing why one big contract gets recognized the way it does, which cost allocation the auditors pushed back on three years ago, and what the board really means by adjusted.

Your sitting CFO carries all of it.

A new hire starts at zero, and so does every agent they build.

So swap the tools freely.

Keep the person who knows what the numbers mean.

AI isn’t replacing CFOs

It’s changing what CFOs consider finance work.

Every company runs on two layers.

• Layer 1 is the work: reconciling, reporting, closing, forecasting.

• Layer 2 is the decisions: where to allocate capital, how to price risk, what to cut, what to invest in.

For 30 years, the finance career ladder was built around layer 1.

You started as a doer of the work. You earned trust through repetition. You moved up by doing more volume, faster, with fewer errors.

That ladder is now being cut from the bottom.

Finance didn’t become less important. The doer layer is being eaten by agents, and what’s left is the part that mattered all along: finance as decision design.

a16z is half right here.

• Right that the work layer shrinks; the leanest teams they see target finance at 2% of headcount.

• Wrong that the answer is a new CFO. The decisions layer is exactly where your sitting CFO’s context and judgment live. Finance jobs will shrink in the areas built on repetition. Finance authority will grow in the areas built on judgment.

A CFO quoted by a16z warns that juniors:

“lack the experience and context to know what ‘great’ looks like.”

If the bottom rungs are gone, how does anyone earn judgment?

Use AI as the tutor, and rotate juniors into exception handling and decision work years earlier. Let them explain the variance an agent flagged, defend a call, and get corrected by someone senior.

That’s the new apprenticeship.

The story of a growth-stage CFO

a16z tells the story of a growth-stage CFO whose team built 50 agents in two weeks. Their own verdict:

“They weren’t all useful.”

A count of agents says nothing about value.

McKinsey’s 2026 State of AI survey explains why.

Among large companies (over $1 billion in revenue), 40% report scaling AI agents. Only 6% say AI generates at least 5% of EBIT, “about the same as the prior year.”

The fix is discipline.

One to three AI operating systems, not one for each use case. A finance team with 50 agents and no shared context ends up with 50 new things to reconcile.

Controls go first, not sixth

a16z lists controls sixth of seven shifts and calls them…

“a critical feature of the product.”

For a CFO, controls are the floor everything stands on.

Remove the standing monthly review

McKinsey’s management shifts translate straight into finance.

Manage by intervention, not by calendar.

Remove standing reviews and define the handful of signals, thresholds, and exceptions that should trigger leadership intervention.

A margin breach or a covenant drifting toward its limit calls the meeting.

Run continuously.

McKinsey reports early deployments sped up planning cycles by up to 10x, reached first value in about 12 weeks, and cut time to ROI by 50% as the system moved into new domains.

Spend time on judgment.

An AI OS is not an executive dashboard, which reports on what has happened. Your team stops shipping numbers up the chain and starts deciding what they mean.

Token unit costs fell more than 95%, yet AI spend at many companies rose more than three times. “

The cheapest intelligence does not necessarily lower costs if lower quality creates expensive mistakes.”

Someone has to own AI economics.

That belongs with the person who already owns the P&L.

Paid subscribers: read on for the AI Operating System for CFOs, a five-layer blueprint, and the first moves I’d make this quarter.

The AI Operating System for CFOs

a16z asks you to change your CFO. McKinsey asks you to change the system. For finance, I map the architecture into five layers with a cockpit on top.

Read more

That has changed fast.

Google launched Gemini 4 Argon for business on October 13.

If your company runs on Google Workspace, you probably already pay for it, and someone on your team will soon ask whether it can build the budget.

Google says it outperforms other frontier models across domains and ranks #1 on the Vals Index, a benchmark that scores AI on real work in finance, tax, legal, and coding.

It also promises it communicates clearly, keeping you in the loop.

Google pitches Argon for work where mistakes are expensive.

• Supply chain planning with many moving parts is on that list, and so is financial modeling where every assumption has to be checked.

• Early partners are said to have made better decisions using it in strategic planning, work that used to need hours of expert review.

On paper, that makes it the smartest finance model you can buy.

That’s a big promise.

So I did what any CFO would do with a big promise.

I tested it. I gave it a real finance job and put it next to the other two frontier models, Claude Opus 5.5 and ChatGPT on GPT-6.

Let’s dive in.

The test

I didn’t want a toy example.

I wanted the kind of model that keeps an FP&A team busy for weeks.

So I invented a company called Velo Works, an e-bike maker with 48 models, hundreds of parts, and eight production lines. I asked three AIs, Gemini 4 Argon, Claude, and ChatGPT, to build its full three-year plan in Excel.

That meant starting with how many bikes it would sell, working out the parts and factory time it would need, and ending with the P&L, cash, and balance sheet.

Then I did what real life does.

I hit the plan with shocks: a sudden jump in demand, a supplier raising prices, and the paint shop going down for three months.

Finally, I asked a question with no formula behind it:

• What happens if the company stops selling Kids bikes?

I made the company up … I know every right answer in advance. So I can check every single number, not just eyeball, whether it looked sensible.

Claude: the analyst you’d hire

Claude built the whole plan in under an hour, and every number was right.

Then I changed some of the inputs, just like you would in a real budget review, and the whole model updated correctly.

That’s the test that separates a real model from a pretty one.

What impresses me most is how it works.

It showed every step as it went, so I could watch the model being built. It caught a gap in my instructions that I hadn’t noticed. It checked its own work before handing it over. And when an answer depended on an assumption, it told me so instead of hiding it.

Claude shows every step as it works.

That’s how a good senior analyst behaves.

ChatGPT: brilliant in the chat but broken in Excel

ChatGPT came next. Every number it gave me in the chat was right, and it even caught and fixed one of its mistakes mid-answer.

Then I opened its file in Excel, and a big chunk of the P&L showed errors.

The thinking behind the model was right, but the file it handed over didn’t work in real Excel. It’s like an analyst who explains the model perfectly in a meeting and then sends you a broken spreadsheet.

ChatGPT’s file in Excel: broken P&L.

It also struggled with the size of the model.

At one point it couldn’t even open a preview of its own work.

ChatGPT couldn’t even preview its own file.

Gemini Argon: the favorite

Which brings me to Gemini, the model I’m most excited about.

Read on to get the verdict.

Read more

This is not a demo.

It’s a one-hour conversation with the Barclays CFO and Wouter Born.

Live. Ask questions.

You can challenge what works and what doesn’t.

AI masterclass for CFOs (LIVE)

Nitin Madan built AI into board reporting at Barclays anyway: agents, connections into company systems, and a memory that reads previous board decks before it writes the next one.

Nitin has run finance at scale, not just advised on it.

Nitin Madan was MD and CFO of Barclays GCC India, an entity of about 30,000 colleagues with $1.3bn in revenue. At the same time, he was Head of Finance Delivery, Global Head of Data Ops, Global Head of Management Reporting, and India Head of FP&A and Finance Transformation.

Under him, the India finance team grew to 60% of Barclays' global finance team.

Before that, he was a partner at KPMG India, where he built the finance transformation practice from zero into one of the firm’s most profitable.

He also worked at EY and Accenture Strategy and set up the Vodafone and Clariant captive centers in India.

What Nitin will cover

The board reporting story

• The agents behind it, and what each one does

• Connecting AI into company systems at bank scale

• A memory that reads previous board decks

• What it took to make it work, and what broke

Where CFOs actually want AI

• Planning

• Reconciliation, where every large bank runs 200 to 500 reconciliation teams

• Performance management and KPIs

The ROI question every CFO is asking

• Why finance rarely tops the ROI list for AI spending

• Why finance still can’t afford to be the laggard function

Open questions

• Your questions, answered live

Who this is for

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

What to expect

A one-hour live conversation between Nitin and Wouter Born. No slides required, open Q&A throughout, and an honest look at what works and what doesn’t.

As a paid member, you get full access to all written content and the session recording. Live sessions are exclusive to Insider members.

If you want the live experience, upgrade to Insider.

If you’re already an Insider member, the registration link is below.

Read more

I tried the new Claude Dashboards and Claude Motion.

Here’s what it does, what I built with it, and where an enterprise CFO should slow down. Claude Dashboards turns a plain-language question into a live dashboard.

Once you have the dashboard, Claude Motion can create videos. All of it built in minutes and tweaked by chatting.

Demo video below.

You can connect to a data platform such as Amazon Redshift, BigQuery, ClickHouse, Databricks, or Snowflake. You ask a question. Claude writes the queries, builds the charts, and keeps them current as the data changes.

It also works with other connectors, so a dashboard of Salesforce opportunities is fair game.

It’s in beta on paid plans: Pro, Max, Team, and Enterprise, per the Claude Help Center.

Let’s dive in.

How to use Claude Dashboards as a CFO

I’ll share all the prompts and files so you can replicate this instantly.

I opened with this prompt; it connects live with my ERP:

Read more

It’s a fair question.

It’s also the wrong one. The better question is, if anyone can build a dashboard in minutes, whose number goes in the board pack?

I spent this morning with the new Claude Dashboards.

Here’s what it does, what I built with it, and where an enterprise CFO should slow down.

Claude Dashboards turns a plain-language question into a live dashboard.

You connect to a data platform such as Amazon Redshift, BigQuery, ClickHouse, Databricks, or Snowflake. You ask a question. Claude writes the queries, builds the charts, and keeps them current as the data changes.

It also works with other connectors, so a dashboard of Salesforce opportunities is fair game.

It’s in beta on paid plans: Pro, Max, Team, and Enterprise, per the Claude Help Center.

3 reasons why a CFO should care

There are three features that map straight onto finance frustrations.

1. The BI queue.

   Today, a new cut of the numbers means a ticket, a data analyst, and a wait. Dashboards moves the first version to the person asking. Anthropic pitches it for quick, exploratory questions. That’s exactly the backlog that clogs finance teams.

2. “Where does this number come from?” Every chart shows the SQL query behind it and when its data last refreshed. Click any number to see its query, or ask Claude to explain it in plain English. For a CFO, that’s the difference between a pretty chart and an auditable one.

3. “Why does sales see a different number?” When you share a dashboard, it runs on each viewer’s own connections. Anthropic’s sharing guide is direct: if a viewer can’t access a data source, that chart shows an error instead of your data. Two people can open the same dashboard and see only what their access allows.

Let’s dive in.

How to use Claude Dashboards as a CFO

Before we start, take a look at the dashboard I just built.

• CFO dashboard

Let’s move on and build a CFO dashboard.

I opened with this prompt; it connects live with my ERP:

Read more

Your CEO forwards you another LinkedIn post about AI closing the books.

“Why aren’t we doing this yet?”

Your auditor asks a different question.

“Who approved this entry?”

Both are fair. Most AI demos only answer the first one.

Today I will answer both live.

I closed Q3 2026 for Born2Cycle, our fictional demo company, in a normal Claude.ai chat. Sonnet 5.5 set to high. I also used the Live Intuit QuickBooks Online and Ramp connectors.

With three custom skills and three prompts.

• No file uploads.

• No new software.

• No export gymnastics.

This is a new case. New quarter, new Born2Cycle dataset, live connectors instead of uploaded files, question cards instead of guesses, and a deck at the end.

AI can run your Close as a CFO.

But it can’t own it.

And that’s the point.

Let’s dive in.

How to close Q3 inside Claude with QuickBooks and Ramp as a CFO

Download the 3 skills below and copy/paste the prompt

Read more

You ran the numbers, and you see 12.

You’re not alone. At AI events surveyed by the Open Future Forum, 70% of CEO-seat leaders expected payback within 6 months.

But only 42% of finance leaders agreed.

Most finance answers fell between 6 and 12 months.

The board hears the CEO’s number. You have to prove it.

The next question is coming.

Every CFO is being asked by the board two questions.

1. Where is the headcount reduction?

2. Which specific OpEx line did this actually shrink?

Here is how to answer without calling your CEO wrong.

1. Find out why the numbers differ

Your CEO counts the day the tool works.

You count the day the saving lands in the ledger. That day comes later. The tool needs data you trust. People need to change how they work. Someone needs to check the output.

Proving the return is the top blocker. 54% of respondents name it, far ahead of integration (23%), security (21%), and data (20%) (Open Future Forum, Oct 2026).

Data hurts finance most. About a third of finance respondents name data readiness as their blocker, twice the rate of the CEO seat. You cannot prove a return on numbers your team does not trust.

2. Skip the headcount shortcut

A smaller team is the fastest saving to show.

But it rarely holds.

Workforce reductions may create budget room, but they do not create return.

An FP&A professional we spoke to said. Most of their AI savings came from retooling processes and consolidating work into fewer teams. The AI played a small part.

Redesign the work first. Let headcount follow.

Give the board a line item, not a story.

The faster team does not survive a board meeting.

What survives is a named P&L line.

For every AI use, build one row:

• Count the real cost

  Include licenses, usage, and the hours people spend reviewing output. If you cannot fill this in, you have your first finding.

• Name one P&L line

  Contractors, overtime, audit fees, support hours, and days to close. If no line moves, say so. Some value is real and invisible to the P&L. Say that plainly.

• Date the baseline

• Without a before number, you can’t prove an after.

• Report a number

  People like it counts as unproven.

• Name the checker

  Anthropic’s CFO says Claude produces the statutory statements for every legal entity. A person reviews each one.

Name one more thing: an owner. Open Future Forum asked investors what they see inside portfolio companies. Where the CEO owns AI, 9% saw nothing measurable. Where nobody owned it, 41% did.

Put a name next to every row of your table.

3. Use this script with your CEO

Say it before the board asks:

“I back the AI plan. I want us to give the board a number we can hit. I mapped each use to a P&L line with a baseline. The first measurable savings land in month [X], not month six. I’d rather show the board a date we meet than one we miss. Can we walk through the table together this week?”

The script supports the strategy. It also gives your CEO a safer story for the board.

Copy what the leaders do.

Barclays announced on 1 October that Claude helps 16,000 colleagues find answers for customers and sorts 120,000 emails a day.

The bank set oversight first, then scaled.

Anthropic’s CFO, Krishna Rao, says the monthly review skill gets him 90 to 95% of the way there. His team now discusses what to do, not what happened.

Both share one pattern: controls and a named checker first, scale second.

This week in finance AI

3 announcements CFOs will hear about.

• IBM, 30 Sept: 62% of 1,500 CFOs now lead AI or technology strategy. Only 6% say finance is transformation-ready.

• Barclays and Anthropic, 1 Oct: Claude already helps 16,000 Barclays colleagues and sorts 120,000 emails a day. Half of its developers use Claude Code by year end.

• Vayu and PwC, 2 Oct: In their CFO Signal Report, 72% of CFOs call billing complexity their top operational friction point.

The Bottom Line

Pressure is high, and controls are thin.

Your board wants the return.

Your auditor wants the explanation.

You answer to both.

In a July Avalara survey, 92% of CFOs and senior finance leaders felt pressure to show an AI return. Only 44% said they were somewhat confident they could explain an AI agent’s actions to an auditor or regulator.

Close that gap this week.

Start with access.

1. List every AI tool with standing access to your ERP, billing system, or bank portal.

2. Record who approved each tool.

3. Set a review date for each.

4. Require human sign-off for anything you cannot undo, such as a new payee.

Do not fight your CEO over six months.

Hand them a better number, with a P&L line behind it, before the board asks.

And that’s all for today!

See you next week.

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

1. Advertise with the AI CFO Office.

2. If you have an exciting AI use case, or you have already put AI into your finance function, I’d love to see it.

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

Find me on LinkedIn, X, and Medium

A financial services plugin marketplace for Claude. Skills, agents, and complete workflows built specifically for finance teams.

For the first time, finance has AI tooling that was actually designed for finance.

The two CFOs must care about most are the Model Builder and the Month-End Closer. The Model Builder is for FP&A workflows. The Month-End Closer is for the work that has resisted every previous wave of AI tools.

That is the one we will test today.

Month-end close is the most painful workflow in the controller’s life.

Beginning balances. Activity detail. Roll-forwards. Tie-outs. GL variances. Items that need sign-off. The work that controllers do at 11pm before the CFO needs the close package.

Every previous AI tool has failed at this work in one of two ways.

• The summarizers read the close files and produced commentary that did not tie to the actual numbers. Useless to a controller.

• The fabricators read the same files and produced confident explanations of variances they did not actually understand. Dangerous to a controller.

What controllers actually need is something that respects the control gate. Surface what is reconciled. Surface what is not. Do not plug gaps. Do not write commentary on numbers that have not been signed off.

That is what Anthropic built.

We ran it on a real Born2Cycle close. March 2026. Three roll-forward areas. Inventory, Accrued Expenses, Debt.

Here is what happened.

Let’s dive in.

The full Anthropic Month-End Closer workflow

Step 1: Install the finance agents inside Claude

Read more

You build the 13-week cash forecast on Monday.

And by Friday a customer has paid late, a vendor has been paid early, and the forecast is already out of date. So you open the spreadsheet and start again.

Why not give this job to Claude Opus 5.5?

I gave it one file with our week 1 cash position, the collections we expect, and the payments already scheduled. It built the 13-week cash flow forecast. Then it tested the forecast, found what was wrong, and fixed it.

But that’s not it.

Plenty of tools can build a forecast. Claude Opus 5.5 checked its own work and improved it, the way a great CFO analyst would.

Anthropic released Opus 5.5 on September 22.

It costs about 40% less to run than Opus 5.

In Anthropic’s test of company earnings reports, where a single made-up figure meant a fail, Opus 5.5 passed 16 of 18 times. Opus 5 and Fable 5.1 didn’t pass once.

Today I’ll show you how to use Opus 5.5 as a CFO.

And set it up with a weekly cash forecast (.skill) and one file (.xlsx).

You install the weekly cash forecast (.skill), upload your week 1 cash file, and get a forecast you can update every week and take straight into your board deck.

Let’s dive in.

How to build a weekly cash forecast with new Claude Opus 5.5 as a CFO

Download the weekly cash forecast (.skill) below

Read more

They shipped 10 ready-to-run agent templates for finance.

• Pitch builder.

• Meeting preparer.

• Model builder.

• Month-end closer.

• KYC screener.

• Five front office agents.

• Five back office agents.

• Each one is built from skills, connectors, and subagents.

I installed the Model Builder Agent to test what it actually does. Most CFOs reading the announcement will think it is a glorified Copilot. It is not.

What you are about to see is Claude trained on how serious finance professionals build models. Formula discipline. Template respect. Cross-statement integrity. Scenario architecture. The visual grammar finance teams expect.

By the end of this newsletter, you will have built a 3-statement financial model for a real company in under 5 minutes.

Let’s dive in.

Step 1: How to install Anthropic's finance agents in Claude

Read more

• Actuals sit in one file.

• Headcount sits in another.

• The revenue drivers live in a Slack thread from August that three people still argue about.

• The P&L is forty tabs, and every time a driver moves, someone rebuilds the bridge by hand.

Budget season is when that workbook becomes the job.

Your job stops being deciding and becomes rebuilding.

Finance gets stuck in the loop.

By the time the exec pack is ready, two drivers move and the YoY bridge is stale again.

This is the path most teams are still on.

So I built Annual-Budget.skill

You can install it once on Claude. Drop in your actuals and drivers. Get a budget pack you can take to the exec team in 15 minutes.

• Drivers.

• P&L build.

• YoY bridge.

• Risks and decisions.

Decide this week.

I added rules for Claude.

Never invent numbers, flag missing drivers, flag stale actuals, and end on the calls that belong on this week’s calendar.

When a driver moves, you can rerun the Skill.

You do not need to rebuild the workbook from scratch every time.

In today’s letter you’ll get the Budgeting Skill and the full prompts.

Let’s dive in.

Download the Claude-Annual-Budget.skill below

The skill breaks down the budget modeling process and includes a template for the Budget Review process. Download it below:

Read more

Your CFO team understands cash, and yet the board is not sure.

The reason is simple.

The data is scattered across your bank export, your AR aging, your AP items, payroll, sales orders, and a tax schedule that lives in someone’s inbox.

The timing is uncertain, because revenue recognized is not cash in the bank, and an invoice issued is not an invoice paid. The assumptions never get written down, so nobody can audit them later.

Scenarios are clumsy, built by hand in a copy of the copy of the model. And the weekly refresh has no reliable trail, so when a number moves, you can’t say why.

PwC says cash forecasting is about when cash actually hits the bank, not when revenue is recognized or an invoice is issued. Every sale has to land in the week the customer is expected to pay, not the week you booked it. That single distinction is where most forecasts quietly go wrong.

And it’s exactly where AI helps, if you use it right.

I built a full 13-week cash flow forecast with Claude.

Claude kept the Excel model as the source of truth, then built an interactive dashboard that lets you stress-test cash, not just look at it.

What impressed me most is what the dashboard didn’t hide.

The dashboard surfaced the exact review items, the threshold breaches, and the scenario trade-offs a CFO needs to see before making a call. It showed the weak spots instead of papering over them. That’s the difference between a dashboard that decorates a number and one you can actually decide from.

Let me show you how I built it using the cash flow model builder skill (.md) and the full prompts I used.

Let’s dive in.

How to build a cash flow forecast with Claude

Step 1: Install the cash flow model builder skill

You’ll get 2 skills and test files.

1. Cash flow model builder skill

2. Cash flow scenario dashboard skill

Download the skills below, then copy/paste the prompt to build a cash flow model.

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

He answered every question on the table.

Insider members 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 full recording is below

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, and 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

When you want an AI to do something specific, say, build a management report or summarize board commentary, you have to prompt it each time.

You set up custom instructions in a project.

But then it only works inside that project.

Outside it? You’re back to copying and pasting the same prompt like a robot.

Claude Skills fixes this completely.

They let you upload specialized knowledge, instructions, best practices, examples, and tone into the model itself. Once uploaded, Claude remembers it.

Anywhere. Anytime. Automatically.

Think of it like The Matrix.

When Neo learns kung fu, it’s not because he memorized a manual. They uploaded it straight into his brain. Seconds later, he knew it. Instinctively.

That’s what Skills does for Claude.

You upload your financial storytelling framework. your voice, structure, report formats, narrative logic, and compliance checks, and Claude knows it. Not just in one chat. Across everything you do.

You can say:

“Draft my CEO letter for Q4 results.”

And Claude instantly applies your Skill, tone, flow, key metrics, and narrative structure without you typing another word of instruction.

That’s the future of how finance works and how you can turn numbers into stories.

Today, I’m showing you how you can build reusable AI workflows in finance to save you days of work.

Let’s dive in.

How To Build Board Decks with Claude Skills in 5 Minutes

COPY-PASTE PROMPT

Read more

Paid Members can access the full recording below

Read more

She said so herself this week, in a post about how she uses Astra as OpenAI’s CFO. By the time she’s awake, the day is already scattered across Slack, email, spreadsheets, dashboards, the calendar, and ChatGPT Work, which pulls it into one update under four headers.

• Who owns it?

• What changed?

• What needs attention?

• What do we need to decide today?

Every CFO I know does a version of that triage by hand.

Phone in one hand, coffee in the other, flipping between Gmail and Slack and a tracker someone updated on Friday, trying to work out which of the six things on fire is actually the one that matters before the 8 a.m. call.

You open your inbox and someone wants approval for $250,000.

The forecast looks good.

The agency needs an answer.

You could approve it and move on.

But in another email, HR says the store has lost four experienced people. In Slack, someone on the floor says customers are waiting 20 minutes for help. Two have already walked out.

You would want to know that before paying to bring more customers through the door. This is the kind of work I wanted to test after reading Sarah Friar’s post about her morning routine.

I wanted to see how much of that I could reproduce.

I went back to Born2Cycle with a different question.

Could Astra read the information arriving between meetings and help the CFO work out what to do next? I gave it three emails, a Slack channel, a calendar, and two finance files. Then I asked it for a morning briefing.

I set the whole thing up, ran it, and read what came back.

This is going to make every CFO’s job easier.

Let’s dive in.

Building OpenAI’s CFO routine with ChatGPT-6 Astra

Step 1: Give Astra your business situation

To get started, go to chatgpt.com and start a new conversation.

Chose work in combination with Astra Medium or High.

Download these 2 Excel files below and import them to Google Sheets

Read more