SaaS metrics need rethinking
CRM influencer Matt Slotnick argued that AI software shifting to consumption and commit‑burn models will require teams to redo traditional SaaS metrics. He suggested that platform leaders should rethink how they benchmark revenue and usage when billing ties to consumption patterns. (x.com)
Matt Slotnick said software teams need new scorecards as artificial intelligence pricing moves from fixed seats to usage and prepaid credit drawdowns. (x.com) He was pointing to a billing shift already visible in the market. Salesforce now sells Agentforce with consumption pricing at $500 per 100,000 Flex Credits, or $2 per conversation, alongside some per-user options. (salesforce.com) Salesforce said on May 15, 2025 that each Agentforce action uses 20 Flex Credits, or $0.10, and that customers can move budget between user licenses and Flex Credits under a new Flex Agreement. (salesforce.com) That pricing model changes what revenue means in practice. L.E.K. Consulting wrote on July 24, 2025 that annual recurring revenue and net revenue retention get harder to read when revenue depends on when customers actually use the product, not just what they signed. (lek.com) The same report said companies are starting to track “time to usage,” “usage ramp rate,” and volatility, and to split annual recurring revenue into categories such as committed annual recurring revenue and usage annual recurring revenue. (lek.com) Consultants and vendors are also building new language around commit-and-burn deals, where a customer promises to spend a set amount and then draws that balance down as usage rises. Nue, a revenue software vendor, describes commit burndown as a way to anchor subscriptions, usage, and services to one committed spend even when consumption is uneven. (nue.io) The pressure to rethink metrics is tied to cost as much as billing. Bessemer Venture Partners wrote on February 9, 2026 that unlike classic software as a service, every artificial intelligence query carries real compute and inference cost, and some teams are running at 50% to 60% gross margins instead of the 80% to 90% common in traditional software as a service. (bvp.com) That makes old benchmarks harder to compare across companies. Bessemer said pricing benchmarks are still scarce because artificial intelligence companies use very different charge metrics, while McKinsey said in September 2025 that it studied 150 global vendors because the “most suitable business model is still in question.” (bvp.com) (mckinsey.com) Some software companies have lived with this problem longer than the current artificial intelligence wave. Snowflake, which has long sold consumption-based data services, reported 125% net revenue retention and $9.77 billion in remaining performance obligations as of January 31, 2026, two figures investors use to judge a business whose customer spending can rise or fall with workloads. (investors.snowflake.com) Slotnick’s point is that more software companies may start to look like that. If pricing is tied to actions, tokens, credits, or committed spend that burns down over time, finance teams will have to separate booked demand from realized usage instead of treating every contract like a steady seat subscription. (x.com) (lek.com)
