This week, four posts on this blog and four stories outside it pointed at the same thing from different directions: enterprise software vendors are running their old pricing models and their new ones side by side — the old one still carrying the P&L, the new one absorbing the press releases. Nobody has switched off the old engine. Almost everyone has started the new one. And the productivity story that’s supposed to justify the whole transition isn’t fully showing up in the data yet.
What We Covered
Monday — Three Models, One Product. Goldman Sachs told investors the industry is “increasingly positioning AI workflows as selling a unit of labor,” citing 40+ companies. The same week, Salesforce was running three Agentforce pricing models simultaneously — $2/conversation, $0.10/action, and a $125+/user AELA that is, structurally, a seat licence. The PricingSaaS 500 counted 1,800+ pricing changes across 500 companies in 2025; the winner wasn’t consumption or outcomes but hybrid, surging from 27% to 41% of companies. The seat model is broken. The replacement hasn’t converged. That matters for valuations pricing in a completed transition that has barely started.
Tuesday — The Productivity That Isn’t There. Snap cut 1,000 people on April 15, cited 65% AI-generated code, and saw the stock jump 7%. Q1 2026 tech layoffs hit 78,557 with 47.9% AI-attributed. But the Fed’s April FEDS Note showed 18% firm AI adoption and aggregate productivity still statistically invisible, and an Atlanta/Richmond Fed working paper surveying 750 executives found a “productivity paradox” — perceived gains outrunning measured ones. Goldman and JPMorgan both cut 2026 U.S. productivity forecasts from 2.5% to 1.8%. The seat-compression chain — AI productivity → fewer workers → fewer seats — has a weak first link. Real displacement in routine roles, yes; uniform sector-wide collapse, not yet.
Wednesday — Tokens Don’t Equate to Value. Adobe’s Anil Chakravarthy launched CX Enterprise at Summit with outcome-based pricing — fees tied to completed campaigns and conversion lift. Figma, in the same orbit, was enforcing the opposite: AI credit limits at roughly $0.03 per credit overage, metering the input. Same industry, opposite answers to what a unit of AI value is. Adobe at $247 is down 46% from peak, betting that orchestration across the customer lifecycle captures more value than access did. Figma is betting that margin preservation through cost metering is safer. Neither has proven it yet.
Today — The Foolish Quarter. SAP reported cloud revenue +27% CC to €5.96B, cloud backlog €21.9B +25% CC, and non-IFRS operating profit +24% CC. Five weeks earlier, CEO Christian Klein had called subscription pricing “foolish.” Software licences fell 37%, support fell 11% and the decline is guided to accelerate. The DSAG 2026 Investment Report found only 3% of SAP customers run Business AI productively while SAP’s own disclosure put AI attach at two-thirds of Q4 cloud orders — a >20x gap between AI on paper and AI in production. The 27% growth is partly the 2027 ECC migration cliff, partly the last of the per-user commitments signed before Klein changed his mind.
What the Market Added
The pattern the posts described showed up harder in the outside news. On Tuesday and Wednesday, Google ran Cloud Next ‘26 and rebranded Vertex AI as the Gemini Enterprise Agent Platform — with Workday, Salesforce, ServiceNow, and Box shipping as partner agents inside it. The A2A protocol v1.0 is in production at 150 organisations; Gemini 3.1 Flash-Lite went GA at under half the cost of Flash; Vertex AI Agent Builder launched a new outcome-based pricing tier. If the dispatch layer is the hyperscaler, the SaaS apps’ internal pricing experiments — credits, outcomes, AELAs — matter less than which orchestrator they get routed through. The dispatch-layer thesis this blog floated in early April now has a winner-tier candidate, and it is not a SaaS vendor.
On April 22, Reuters reported that Thoma Bravo is near an agreement to hand customer-experience software firm Medallia to its lenders — Blackstone, KKR, Apollo, Antares — in a ~$5.1B equity wipeout. Acquired in October 2021 for $6.4B, Medallia’s annual debt servicing had ballooned to $300M against roughly $200M of earnings. FS KKR marked the debt at 79 cents on the dollar; Apollo Debt Solutions at 74. This is the first major 2021-vintage software buyout to collapse on the credit wall — and the category is customer experience, exactly what Intercom Fin, Sierra, and HubSpot Breeze AI have been compressing to $0.50–$0.99 per resolution. Distressed software loans across private credit: $46.9B as of February. Post 9’s “Great Sorting” is no longer theoretical.
Two incumbents reported inside 48 hours and, together, form the cleanest falsification case of the quarter. ServiceNow Q1 (April 22): subscription revenue $3.67B, +22% YoY; 50% of net new business now from non-seat pricing; 16 deals >$5M new net ACV, +80% YoY; Now Assist customers at $1M+ ACV +130% YoY; internal 2026 AI target raised from $1.0B to ~$1.5B. Diginomica’s Den Howlett noted dryly that the pricing model looks more like “AI premium on top of seats” than outcome pricing replacing them — which is exactly the AELA pattern from Monday’s post. Between SAP and ServiceNow, two of the three largest incumbent platforms beat Q1 while pivoting pricing in public. Both also disclosed strong attach metrics and, in SAP’s case, thin production metrics. Attach paid for Q1. Production has to pay for 2027.
And at the category below the platforms, UKG cut 950 jobs on April 15 in an “AI-first” restructuring under CEO Jennifer Morgan, ex-SAP co-CEO and ex-Blackstone. UKG sells the workforce-management software other companies use to execute layoffs — and is now using its own category to cut its own workforce. Since H&F took it private, cumulative headcount reduction is roughly a third. A workforce-software vendor firing HR staff on AI grounds is the post-10 “market pays you to shrink” mechanism applied recursively to the vendor’s own P&L.
The Thread
Run the week together and the picture gets sharper, not more contested. The transition away from seat pricing is real — hybrid is already the industry norm, incumbents are publicly disowning their own legacy models, and the hyperscaler has just positioned itself as the orchestrator tier with outcome pricing built in. But the transition is happening on top of the old model, not instead of it. SAP’s 27% cloud growth came from forced migrations and AI attach, not AI production. ServiceNow’s 50% non-seat share of new business is real and still wrapped around seats. Adobe is running outcome pricing and credit-based entitlements on the same platform. Salesforce is running three pricing models at once. Medallia, meanwhile, shows what happens when the parallel run doesn’t get started in time: a 2021 buyout whose debt stack assumed the old model would last five more years than it did.
The Fed data is the constraint. If the productivity to justify the seat cuts doesn’t show up at scale, the transition is partially performative and the incumbent repricing overshoots. If it does show up, the Medallia category is the leading edge of a credit cycle that still has $46B of distressed loans behind it. The hypothesis this blog has been tracking — software loses value because AI commoditises it — holds more tightly for the Medallia end of the market and is actively being falsified at the SAP/ServiceNow end, where growth and margins are intact, but the composition of revenue is shifting in the exact direction the hypothesis predicted. Two distinct outcomes coexisting inside one thesis is not a tidy result, which is why we track an evidence score rather than a verdict. Current hypothesis score: 62%.
Sources: Google Cloud Next ‘26 keynote, Reuters via Yahoo Finance — Thoma Bravo / Medallia, Benzinga — Medallia $5.1B wipeout, Motley Fool — ServiceNow Q1 2026 transcript, Diginomica — ServiceNow Q1 analysis, HR Executive — UKG layoffs, Adobe CX Enterprise press release.