Salesforce hires Anthropic for $5B, Workday awaits Silver Lake

The SaaSpocalypse thesis just cracked on the company most exposed to it. Salesforce announced a $5B Anthropic partnership the same day it reported earnings, Rippling published the first honest benchmark of what AI models actually cost to do real work, and a Silver Lake buyout rumour hangs over Workday's numbers.

Salesforce proves the SaaSpocalypse isn't happening — at least not to established SaaS

Salesforce did two things on August 26 that together break the "AI will replace SaaS" thesis. First, it reported Q2 FY27 revenue of $11.3 billion, up 11% year over year, and raised full-year guidance to $46.1–$46.4 billion — stock jumped 20% on the day, 65% since the June low. Second, it announced Claudeforce, a deep partnership with Anthropic backed by a $5B Salesforce stake and a $300M commit for Anthropic tokens in 2026. Claude gets embedded inside Salesforce as the reasoning layer; Salesforce gets embedded inside Claude as a plugin with 37 prebuilt sales skills. Anthropic CEO Dario Amodei on CNBC: "we are not interested in destroying anyone."

The number that matters for pricing sits further down the earnings release: 3.2 billion Agentic Work Units delivered in the quarter, up 97% QoQ, cumulative 7.0 billion. Agentforce ARR passed $1.5 billion, up more than 240%. An Agentic Work Unit is a billing unit. Salesforce's Agentforce pricing runs two meters in parallel: Flex Credits at 20 credits per standard action (roughly $0.10), sold in packs of 100,000 for $500, and a legacy $2-per-conversation model. Crossover sits around 20 actions per conversation.

Read the two announcements together and the picture is clear. The company that built the modern per-seat playbook is (a) putting a work-volume metric next to ARR because seats no longer explain the revenue, and (b) letting Claude become the front door to its own product. That is not a company being disrupted. That is a company adapting fast enough to co-opt the disruption. Systems of record with two decades of enterprise trust are not the SaaS getting killed by AI. Vibe-coded competitors are not eating Salesforce. What the market is saying this week is that the SaaSpocalypse thesis was always wrong for the incumbents — Salesforce, SAP, Oracle, Workday, HubSpot, ServiceNow. What is actually at risk is vertical and general-purpose tools without proprietary data. Airtable at 1/5 peak value is the visible casualty.

For buyers: if you were dumping horizontal SaaS out of your stack expecting AI-native replacements, this week is your reset. If your Agentforce contract predates Flex Credits, the meter you are on was chosen for you — pull your Digital Wallet action counts, compute actual actions per conversation, and check which model you should be on before renewal. And when Claudeforce hits open beta in September, watch the pricing sheet closely: when "AI added to your seat" moves from add-on to bundled to default, the effective per-seat cost creeps.

Rippling ran the AI pricing experiment everyone else talks about

Rippling's president and CPO Matt MacInnis published a benchmark this month running roughly 2,100 scored agent attempts per model, across 15 models, against real payroll and personnel records in production — salary increases, onboarding checklists, scheduled terminations, pay-run entry. Pass or fail against Rippling's own correctness checks, no partial credit.

The result: Opus 4.6 passed 91.0% at $1,453 to run the full set. GPT-5.5 med passed 89.5% at $1,435. Eighteen dollars and 1.5 points apart. Across the 15 models, full-benchmark cost ranged from $621 to $4,359 — a 7x spread that quality did not track.

Rippling has standing here. On August 7 it shipped an AI Spend Console after its own token bill was projected to consume 40% of R&D headcount budget, growing 80% month over month, with one engineer burning $50,000 a month. Post-controls it fell to about 15% of that budget; July cost 37% of April at similar volume.

For buyers: the premium tier is not buying accuracy on structured enterprise tasks. Benchmark the cheap model on your own workload before accepting a vendor's default, and ask any SaaS vendor billing you for AI work which model it routes to. If it picks the $4,359 model when the $621 one passes at the same rate, you are funding their procurement laziness.

Workday reports Q2 as Silver Lake circles

Workday reported Q2 FY27 results on August 27. Its guidance had called for quarterly subscription revenue of $2.455 billion, up 13%, against full-year subscription guidance of $9.925–$9.950 billion.

The context changed on August 13, when Reuters reported Silver Lake in talks to take Workday private. The stock surged as much as 21% and was halted for volatility, its best day in a decade. At roughly $43 billion of market value pre-move, a completed deal would rank among the largest software buyouts ever. Neither side commented; there is no agreement.

The pricing lens is not the multiple, it is what follows. Software under sponsor ownership follows a consistent pattern: list price rises, discounting discipline tightens, bundles get repackaged, low-margin modules get sunset or repriced. HR and finance systems of record are also the stickiest software an enterprise owns — switching cost is measured in years.

For buyers: if a Workday renewal lands in the next 18 months, the negotiating window is now, while the deal is still a rumour. Push for multi-year price protection and caps on uplift. Those clauses get much harder to win once new owners set the price book.

Also this week

• Salesforce signed a definitive agreement to acquire Regrello on August 18, its third AI deal in days after Waii and Bluebirds. Regrello folds into Agentforce and Slack — more actions, more Flex Credits.
• Bending Spoons made its first post-IPO deal on August 4, agreeing to buy Airtable for $1.285 billion in cash. Airtable peaked near a $12 billion valuation in 2021.
• Adobe is deferring roughly $500 million of ARR from Creative Cloud price increases planned for H2, per its Q2 FY26 call — a rare voluntary retreat on list price.
• HubSpot's Breeze agents have run on outcome pricing since April 14: $0.50 per resolved conversation, $1 per recommended lead, down from $1.00 per conversation regardless of outcome.
• Gartner put $234 billion of enterprise application spend at risk from agentic arbitrage by 2030 — roughly 20% of enterprise SaaS spend, repriced rather than eliminated.

Catch you next week. — SaaS Price Hub