On October 8, a16z published "You Need a New CFO." Its point: the scarce CFO skill is now designing the operating system the company runs on.

In plain terms, an operating system here means the connected set of data, rules, software agents, and approvals that turns what happens in the business into numbers and decisions.

That system holds operating data (bookings, hours worked, inventory, headcount) next to the ledger, so every department's numbers run through it.

The operating system a16z describes belongs to the whole company, and the essay implies that finance's job just got bigger.

Our point: the operating system runs on every department's data, so whoever owns it stops being a department. Finance will own it, and that makes finance cross-functional, as it should be.

Finance is the natural owner for two reasons:

  • Every operating number ends up in finance anyway. Bookings become revenue, hours become cost, inventory becomes cash.

  • Finance is the one function that already signs for numbers. A named approver, an audit trail, and a close are habits no other department has.

What it takes for finance to own the operating system.

Four things finance owns once the system runs. None of them is a tool.

1. Finance owns the definitions every department reports in.

  • The problem: every tool keeps its own list of customers, so ten AI tools can still give three answers to "what did we sell last month."

  • The decision: the CRM is right about who the customer is, billing about what they were charged, accounting about what counts as revenue.

  • The cross-functional part: sales, operations, and HR keep their tools and report in finance's definitions. Finance writes the one-page dictionary and holds it.

2. Finance owns the rules, and the rules run in code.

  • Code: revenue recognition, cost splits, bank loan tests, written once and giving the same answer every run.

  • Agents: read documents, flag what looks off, draft entries. Useful, and unreliable at math, so they work around the rules, never inside them.

  • A named person: signs anything that changes the books. Every number traces to a rule and a signature.

3. Finance owns the fit to this business.

  • What has to be custom: which deposits belong to which customer, how a state's excise tax changes margin, which commodity price moves cost of goods. No product ships with it.

  • Where it lives today: three heads. The controller knows the close, IT knows the systems, operations knows the workflow.

  • The ownership: finance pulls all three into one design and holds it, which is a coordination job before it is an engineering job.

4. Finance owns the build without becoming an engineering team.

  • What works: the people who know the business decide what the system has to do, and a builder who knows finance and systems builds it with them. We are that builder for our clients, and the client owns the system once it runs.

The modules so far.

We build this operating system for clients one module at a time. The base is a custom layer on top of the systems a company already runs (ledger, CRM, payroll, and the rest). Each module then takes one department's data, runs it through finance's rules, and puts the result next to the ledger. We laid out the base in August. The three issues since each showed one module running:

Part 2, next week.

With those four owned, finance is in every department's decisions as they happen. Next week: what finance becomes when it runs the company's operating system.

Key Takeaways

  • The operating system runs on every department's data, so finance, which owns it, stops being a department.

  • Ownership is four things: the definitions everyone reports in, the rules in code with a person signing, the fit to this business, and the build.

  • People who know the business, plus a builder who knows finance and systems, get it running.

The Number

This Week in AI-Native Finance

Pigment's Q3 2026 CFO Index, released October 7, found that 83% of 2,000 finance leaders say consumption-based AI costs have pushed spending above expectations, while 13% have stayed within budget. Pigment says average spending ran about 27% above expectations, and expected budget growth for next year fell from 19.1% to 15.6%, the lowest yet. The survey covered the US, UK, France and Germany and was fielded August 19 to September 21.

AWS's pricing team turned an 18-tab Excel model for evaluating customer deals into an AI chatbot, AWS CFO John Felton told Fortune CFO Daily on October 8. Users ask the chatbot what a 20% price drop does to a deal, how different payment terms change it and where break-even sits, and the Excel model can still be exported. A separate team's agent now compares payment terms in all customer contracts with AWS's payment records, where staff previously checked a sample.

SAP announced Joule assistants for finance work at SAP Connect on October 6, saying they use customers' existing data, authorizations and audit trails and that customers control the level of autonomy. SAP plans additional specialized agents for financial close, planning, treasury, billing, tax and receivables, with early adopter programs planned for the first quarter of 2027. Constellation Research reports that little of the announced work is available now.

HubSpot, DNB and FICO each announced workforce cuts on October 6 and describe AI's role in them differently. HubSpot is cutting nearly 660 jobs, about 7% of its workforce, and its CEO says the cuts are not driven by AI-related efficiencies. DNB is cutting about 400 full-time positions and says AI agents already deliver efficiency gains. FICO is cutting about 15% of its workforce and lists AI-driven product development among its objectives.

The Financial Times reported on October 8 that OpenAI told investors its annualized revenue was close to $50 billion at the end of September, about $20 billion less than earlier reports. Axios reports that Anthropic counts the gross amount of certain sales through cloud partners such as Amazon, while OpenAI counts only its share of revenue from partners including Microsoft. Both figures are annualized run rates, not booked revenue.

What We're Reading

  • Maricic's team read 919 earnings calls and found one firm, S&P Global, that put a dollar figure on AI savings, about $9 million a year from its Kensho Scribe tool, while analysts asked for such a number 29 times without getting one. The column says JPMorgan has not given a figure on any of its 13 earnings calls since mid-2023, and it ends on Bank of America's claim of $400 million of AI cost against $800 million of benefit.

  • Martin Mulyadi argues that audit committees do not need to become AI engineers. Management should give them a short quarterly evidence pack on each AI use case that matters to financial reporting, internal control, compliance or forecasting, covering what changed, where the major risks sit, which controls are operating and which exceptions remain open.

  • Contributor Dara-Abasi Ita argues that PE-owned retailers and distributors should build and own their AI systems, because owned systems keep their code and data through a sale while subscriptions are recurring costs that a buyer capitalizes at the exit multiple. The benefit depends on total running costs falling.

About the Authors

Christian Sanford and Kenny Jen are the co-founders and managing partners of QuantFi, where they help PE- and VC-backed companies build investor-grade finance functions powered by AI-native infrastructure.

Christian came up through investment banking at Barclays ($20B+ in M&A and capital markets), buy-side investing at a hedge fund, and fractional CFO work for investor-backed companies. BBA and MSA in Accounting from Texas Tech.

Kenny led financial strategy at Pilot CFO, Pure Beauty, and Emil Capital Partners after starting his career at Credit Suisse. He has overseen finance for consumer, SaaS, and manufacturing businesses, and works directly with founders on pricing, working capital, and capital allocation. BSBA in Finance from Georgetown.