On October 26, Anthropic became the first of the leading AI labs to file publicly for an IPO. The line I keep going back to is the definition of its headline metric: run-rate revenue is the last 28 days of usage fees multiplied by 13, plus one month of subscription fees multiplied by 12.
There’s nothing improper about that, but it shows how far the topline number has drifted from the one an auditor signs. Most technology companies now carry at least three versions of revenue: ARR, GAAP revenue, and cash. As the OnlyCFO newsletter put it, “ARR is a hope, GAAP revenue is a judgment, and cash is a fact.” When people start asking which of your numbers is real, how big the gap is matters less than whether you can explain it.
The market prices doubt, not errors
In February, ICON plc, a profitable global company with more than $8 billion in annual revenue, disclosed that its revenue may have been overstated. Its shares lost more than a third of their value in a single session. The restatements ultimately came to about 1% of revenue.
So an error of about one percent took far more than one percent off the company’s value, because for a while, nobody could say with confidence what the rest of the number was worth. Revenue is the number every other number leans on. If it’s wrong, or right but impossible to explain, everything built on it inherits the problem.
Why doubt is multiplying
Pricing changes constantly. About three in four software companies changed pricing or packaging in the past year, and hybrid pricing became the primary model for 37% of companies, up from 25%. Expiring credits and overages each bring a recognition judgment under ASC 606 that a price list doesn’t answer.
Scrutiny is already here. The PCAOB now requires auditors to evaluate the reliability of information produced by technology, and the SEC has pressed companies on how ARR differs from GAAP revenue.
AI raises the stakes. Finance teams are deploying agents that work from whatever number they’re given. If that number is calculated by rule and traceable to its source, the agents make the team faster. If nobody can trace it, they spread the doubt further.
Two reasonable assumptions that leave revenue exposed
“Our billing system has it covered.” Billing tells you what you charged, not what you earned.
Take a customer who signs a 12-month contract for 100 seats, with three free months before billing starts. In month two, they add 50 seats. If revenue follows the invoice, it keeps accruing on 100 seats. In month four the first invoice goes out for 150, and now the invoice and schedule disagree. Nothing flagged it, so the controller finds it at close.
“Our ERP has a revenue module.” The ERP records revenue, but it was never built to be where revenue gets decided. That takes ongoing judgment: allocating, re-measuring after every change, and recording why each number is what it is. When the ERP can’t hold that work, it ends up in a spreadsheet or custom code, somewhere much harder to trace.
The fix is smaller than most people expect. You don’t have to replace your ERP or your billing system. You have to decide where your revenue number is calculated, and make that place governed and traceable.
Five questions to ask your controller this week
Ask your controller or head of revenue accounting and ask these five questions. If the honest answer to any of them is “I’m not sure,” you’ve learned something.
- Does our revenue number draw on every source, including contracts, billing, usage and the ERP, with each one checked first?
- Does our revenue process work alongside our ERP and billing system, without a re-implementation?
- Could we close revenue today and explain every variance?
- Can a new pricing model go live without adding accounting risk?
- Can we trace every performance obligation, and every change behind it, to its source and the rule that was applied?
Where RightAgents come in
Answering those questions is the job RightRev was built for, and this week we launched RightAgents to take it further. Our revenue engine already calculates each number from the contract by rule, across more than $44 billion in annual customer revenue. RightContext connects any billing system, ERP or usage source and checks the data first. The agents take on the manual work around that number, and a person approves everything before it posts:
- Contract CheckList pulls key terms and obligations from each contract and checks them against an ASC 606 checklist or your own.
- Architect turns a new pricing model or policy into a rule from a plain-English description, and finance approves it before it goes live.
- Use Case Modeling builds and runs test scenarios so changes are checked before they reach the books.
- Anomalies flags unexpected variances, missing prices and odd allocations.
- Reconciliation checks that revenue matches across source systems, RightRev and the general ledger, and explains each difference.
- Flux explains why revenue moved between periods and drafts the commentary.
- Assistant answers plain-English questions about contracts, revenue and balances.
Back to the seat example. RightRev updates the revenue schedule when the contract changes in month two, Reconciliation flags that billing hasn’t caught up, and Flux explains the movement. The controller knows before close, not at it.
The agents also run through MCP, so teams can use them from Claude, ChatGPT and other AI tools within their existing permissions. See what’s new in our agentic capabilities: read the press release.
Settle it before someone else does
The number that matters is the one you can defend, obligation by obligation, to an auditor, an investor or your own board, whenever you’re asked.
Where your revenue number gets worked out is still your decision to make. If you don’t make it, an auditor, an investor or a headline will make it for you.
Somewhere, your next pricing model is already on a whiteboard. The question is whether your revenue number will be ready when it reaches a contract.
