You changed how you price. Did anyone tell finance?

For twenty years, software was easy to price: a seat, a license, an annual contract. Then AI showed up and pricing got creative.

Intercom charges for its AI agent per resolved conversation. Salesforce sells Agentforce in credits spent per action. Across enterprise software, companies are adding prepaid credit packs, overages and fees that only kick in when an agent actually gets the job done.

It makes sense. Customers pay for what they use. Sales gets flexibility. Price tracks value.

But one thing didn't change: the rules for booking revenue. ASC 606, the standard companies report revenue under, was written for fixed-price contracts with clear deliverables. Every new pricing model forces a judgment call about what counts as revenue, and when.

Three ways new pricing surprises finance

Is it $10M of revenue, or $3.3M? Say you sell an AI support product built on a frontier model. Customers pay you $10M a year, and about $6.7M of that goes straight to the model provider. So is your revenue the $10M customers paid, or the $3.3M you kept?

It depends on who the customer is really buying from. If they're buying your product (you set the price, own the relationship and decide which model handles each request), you report the full $10M. If you're mostly passing along access to someone else's model, you report $3.3M. Same business, same profit, and one version of the board slide shows 3x the revenue.

You can't book the agent's wins in advance. Charge $4 every time your agent resolves a ticket, and sales will want to count a full year of resolutions at signing. For a new customer whose resolution rate could land anywhere from 20% to 70%, that's not happening. Revenue arrives as the agent succeeds, one month at a time.

Unused credits aren't a windfall at expiration. If history shows 8% of prepaid credits go unused, that revenue is recognized gradually as customers use the rest, not in one lump when the credits expire. And in some jurisdictions, unclaimed balances belong to the state, not to you.

None of these are edge cases. They show up the first month a new pricing model goes live, and they compound at every close after that.

The answers, worked out

Hyphenate put together a guide to the six questions finance teams hit most with usage-based pricing: setup and fine-tuning fees, pay-per-outcome, gross vs. net, prepaid credits, unbilled usage, and sales commissions on contracts with no fixed term.

Each one comes with a realistic scenario, a decision tree showing which facts change the answer, and the ASC 606 citations to back it up, cross-checked against the Deloitte and KPMG handbooks. It's written for the people who have to defend the number to an auditor.

If you set pricing, it's a fast way to see what your choices mean downstream. If you run finance, it's the analysis you'd otherwise build from scratch. Either way, it's worth forwarding.

Where AI actually helps

Getting the answer right on one contract is the easy part. The hard part is applying it the same way across every contract, every month, with a trail an auditor can follow from contract to amendment to usage to revenue.

That's what Hyphenate does. Its AI agents read billing and revenue terms straight from contracts, run invoicing and revenue recognition from that same record, and draft the reasoning behind each call, with citations to the accounting standard and the company's own policies. Straightforward contracts get handled. New or unusual ones go to a person, with the analysis already drafted.

For finance, that means pricing can keep changing without every close turning into a scramble.

Before your next pricing change

If your company is adding credits, overages or its first pay-per-outcome deal, bring finance in before launch, not after the first close. The guide is a good place to start that conversation.

Get the UsageBased Rev Rec guide