In the first week of February 2026, roughly a trillion dollars in software market value disappeared in seven trading days. Forrester, which coined the "SaaS-pocalypse" label, tied the sell-off to how fast AI agents are improving. The market bet that agents would:
- absorb software workflows,
- make per-seat pricing obsolete, and
- let small teams copy platforms that once took years to build.
The market priced the wrong death. Forrester's own numbers show global SaaS spending still climbing, from about $318 billion in 2025 to a projected $512 billion by 2028. The core systems (ERP, CRM, HCM, supply chain) are not going away. No one is vibe-coding a controls-grade record-to-report stack by Friday.
What is dying is the unit of pricing. The seat is going the way of the billable hour, for exactly the same reason.
The seat and the billable hour have the same problem
A seat was always a stand-in. Clients never wanted licenses. They wanted invoices matched, tickets resolved and forecasts produced.
- Per-user pricing worked while users did the work. Headcount was a fair way to meter the value consumed.
- Once an agent does the work, the meter reads zero while the value keeps flowing.
- A finance team that runs source-to-pay matching through agents, at a third of the human seats, uses more workflow value on fewer licenses.
Every seat-based vendor can do that arithmetic, so the repricing is arriving from both sides at once.
From vendors: flex credits that cover both seats and agents, seat-plus-usage hybrids and consumption tiers.
From buyers: pressure that already shows in the spend data. UBS reported in June, from its conversations with enterprise IT leaders, that:
- roughly 60% are now limiting AI spend with guardrails such as token pooling, usage caps and cheaper models, and
- part of the AI bill is reportedly being paid for by trimming SaaS and external IT budgets.
Software spend is not disappearing. It is being cut, moved and re-metered, all at once.
This is the same shift that is repricing professional services, where outcome fees are replacing hours, landing on the software line in the same quarter:
- Hours meter human effort.
- Seats meter human access.
- Agents break both meters at once.
Both were acceptable stand-ins for value while people did the work. It is one repricing, running across the whole industry.
Why this belongs to the CFO, not just the CIO
The strategy story is the CIO's. The financial mechanics are the CFO's, and they get too little attention.
1. Fixed costs are becoming variable.
- Seat contracts were wonderfully predictable: a flat line in the budget.
- Consumption and outcome pricing trade that predictability for a better fit to value. That is often a good trade, but software starts to behave like cloud spend did around 2019: it can surprise you.
- The cost discipline that eventually tamed cloud (FinOps) is now needed for tokens and agent use, before the surprise, not after.
The visibility gap is already documented (KPMG, Global AI Pulse, Q2 2026):
- Only about 26% of organizations can see what their AI actually costs in real time.
- 49% have already scaled back agent deployments once the bill arrived.
- UBS found a single user running $35,000 a month in inference.
Multiply that across an enterprise and the case for tracking AI spend against the value it returns makes itself.
2. The negotiation window is open, and it will close.
Vendors caught mid-transition need reference customers for their new pricing. Buyers have unusual bargaining power for the next twelve to twenty-four months. Forrester's advice is blunt:
- renegotiate enterprise SaaS as agents roll out,
- consolidate to strategic vendors, and
- separate the tools that truly differentiate from the utilities an agent can absorb.
One rule sits above the rest: do not sign a multi-year, seat-committed renewal in 2026 without an agent-conversion clause. That locks in the old meter at the exact moment it is being replaced.
3. Rationalization finally has a trigger.
- Every SaaS estate carries duplication. The average enterprise runs hundreds of applications that do not talk to each other.
- Agent rollout is the audit that seat counts never triggered. Mapping which workflows agents will take over also shows which point solutions exist only to bridge gaps the agents now cover.
- Forrester expects horizontal point-solution vendors to be absorbed or collapse, while vertical vendors with rich data survive. Your renewal calendar should sort vendors the same way.
What the winners do
The winning buyers will not be the companies that slash software spend. Forrester's growth numbers say that is not even possible.
They will be the ones that re-contract around the new unit of value (workflows completed, outcomes delivered, agent-hours used) before those meters become standard without them.
The SaaS-pocalypse is not an extinction event. It is the largest buying opportunity in a decade, and it belongs to the CFO who reprices the contract before the vendor does.
The model was never the moat, and neither was the seat. Value moved to what you own and can meter honestly: the workflow, the outcome, the governed capability. Software is just the latest stand-in to find that out.