On 29 January 2026, software had its worst single day since the Covid crash. ServiceNow dropped around 11% despite beating earnings for the ninth consecutive quarter.
Read that twice. Nine straight quarters of beating expectations, punished anyway. Something other than performance was being priced.
What the Market Was Actually Saying
The selloff was not a judgement on execution. It was investors repricing software around a single belief: that AI agents will shrink the number of human seats companies pay for.
If your revenue is a function of headcount, and your customers headcount is about to be a function of automation, then your growth model has a dependency nobody underwrote. That is the whole thesis in one sentence, and you can agree or disagree with it while still recognising it moved several hundred billion dollars.
The Three Arguments Being Made
1. Agents Replace Seats
The aggressive version: agents do the work of junior staff, headcount falls, seat revenue falls with it. Some analysts describe agents as far faster and cheaper than junior staff in narrow tasks, directly cannibalising the seat-based revenue underneath commercial software.
2. AI Dissolves the Moat
The structural version: if AI-assisted development lets a competent team rebuild 90% of a workflow product cheaply, then feature moats disappear and only data, distribution and integration depth remain defensible.
3. Buyers Consolidate
The boring version, and probably the most immediately real. Buyers are cutting tool counts, with reported consolidation in the 20% to 30% range. When budgets tighten and AI features arrive inside platforms you already own, the fourth-place point solution gets cancelled first.
What the Bears Are Missing
The global B2B SaaS market was still valued around $634 billion in 2026, and more than 80% of companies were expected to have AI-enabled applications deployed, up from roughly 5% in 2023. That is not an industry disappearing, it is one changing shape at speed.
Systems of record are also stickier than the thesis assumes. Ripping out the platform that holds your customer history to save on seats is a project few executives survive, and agents need somewhere to read from and write to. That somewhere is usually the incumbent.
What This Means If You Run a Software Business
- Know your seat exposure. What share of revenue is tied to headcount that automation could reduce? Calculate it before an investor does.
- Own the data, not the interface. Interfaces are being rebuilt cheaply. Proprietary data and integration depth are not.
- Make your product agent-accessible. If agents cannot reach your product, they route around it, and that is how quiet churn starts.
- Reprice deliberately. A hybrid of platform fee plus consumption is the pragmatic landing spot for most.
Conclusion
The 2026 repricing was about the seat, not about software demand. Companies whose revenue tracks headcount got marked down, whether or not their own numbers justified it. If you sell software, reduce the share of revenue that depends on how many humans log in, make your product reachable by agents, and lean on the data and integrations that cannot be rebuilt in a weekend.
Frequently Asked Questions
Is SaaS actually dying?
No. Seat-linked growth is under pressure and the delivery model is fine. Software delivered as a service is not the thing being questioned.
Which categories are most exposed?
Products priced per user where the user was doing repetitive work an agent can now do. Least exposed are systems of record and anything embedded in a regulated workflow.
Should buyers expect prices to fall?
Prices for seats, possibly. Total spend, unlikely. Consumption charges have a habit of quietly absorbing whatever the seat reduction saved.