For 25 years, the software industry ran on a beautiful lie: sell access, not outcomes. License seats, charge monthly, watch NRR compound. The SaaS model turned software companies into annuities — predictable, scalable, wildly profitable. Salesforce hit $34B in revenue. ServiceNow crossed $10B. Workday, HubSpot, Zendesk — the entire stack of enterprise software was built on one foundational bet: humans would always need tools, and tools would always need to be paid for.
That bet is now losing badly.
In Q1 2026, for the first time, enterprise spending on AI agents — systems that autonomously complete tasks — exceeded enterprise spending on new SaaS seats across tracked Fortune 500 procurement data. The inflection point wasn't a prediction. It happened. And the downstream consequences for the $650B SaaS industry are only beginning to register on earnings calls.
The Fundamental Problem With SaaS
Traditional SaaS solved a real problem: it made enterprise software accessible, updatable, and scalable. But its core economic model had a hidden assumption baked in — humans were the execution layer. The software was always a tool. A human still had to log in, click around, interpret the data, take the action.
Agent-as-a-Service breaks that assumption entirely. An AI agent doesn't need a Salesforce license to update CRM records — it is the CRM function. It ingests signals, writes records, sends follow-ups, scores leads, and forecasts pipeline without a single human in the loop. You don't license a seat for an agent. You pay for outcomes — calls completed, tickets resolved, deals advanced.
— Head of Enterprise Technology, Fortune 100 financial services firm, Q1 2026
This is the fundamental disruption. The SaaS pricing model — monthly per-seat recurring revenue — was designed for human workers. When the workers become agents, the pricing model collapses.
How the Pricing Model Breaks
Walk through the math on a hypothetical sales team:
| Function | SaaS Model | AaaS Model | Delta |
|---|---|---|---|
| CRM (Salesforce) | $150/user/mo × 50 users = $90K/yr | $0.12/enriched record, ~$8K/yr total | -91% |
| Customer Support (Zendesk) | $89/agent/mo × 30 agents = $32K/yr | $0.04/resolved ticket, ~$3.2K/yr total | -90% |
| HR (Workday) | $45/employee/mo × 200 = $108K/yr | $0.80/HR event processed, ~$6K/yr total | -94% |
| Marketing (HubSpot) | $3,200/mo enterprise = $38K/yr | $0.003/contact touched, ~$4.5K/yr total | -88% |
| IT Ticketing (ServiceNow) | $120/user/mo × 40 = $57.6K/yr | $0.06/ticket resolved, ~$2.4K/yr total | -96% |
The savings aren't marginal. They're existential for SaaS vendors. A mid-sized company running entirely on AaaS instead of SaaS saves $250K–$2M per year on software alone — before accounting for the headcount reduction that drove the switch in the first place.
The Vendors Who Saw It Coming (And Pivoted)
Not every SaaS company was caught sleeping. A handful recognized the existential threat early and repositioned before the wave hit.
Launched in late 2024, Salesforce's Agentforce represents the most aggressive SaaS-to-AaaS pivot in enterprise software history. Rather than selling seats, Agentforce sells autonomous agent "conversations" — completed customer interactions. Revenue per customer has declined, but Salesforce is betting that stickiness and volume compensate. Early data suggests they're right: Agentforce customers show 4.2x higher retention than traditional seat customers.
ServiceNow has repackaged its entire ITSM platform as an agent-orchestration layer. The play: your AI agents use ServiceNow as the workflow backbone. You're not paying for human seats — you're paying for agent throughput. ACV has dropped 40% per enterprise, but customer count has grown 180% as mid-market companies that could never afford traditional ServiceNow now run it entirely agentically.
Workday's pivot has been the slowest and most painful. HR software assumes humans. Your payroll system was never designed for a workforce where 60% of "employees" are agents. Workday's Q4 2025 earnings call was the first time management used the word "existential" in reference to AI disruption. They've since announced Workday Illuminate — a full agent-native rebuild of their HRIS. Expected timeline: 18 months. Expected to be too slow by most analysts.
Categories represent ARR exposed to agent displacement. Assumes 60–80% function coverage by autonomous agents by 2028.
The New Pricing Primitives
If per-seat is dead, what replaces it? Three models are emerging as the dominant AaaS pricing frameworks, each with different implications for vendors and buyers:
Pay per completed task, resolved ticket, qualified lead, or processed transaction. The buyer only pays when value is delivered. Risk shifts from buyer to vendor. This is the dominant model for customer-facing agent deployments — Intercom, Salesforce, and Zendesk are all moving here fast.
Pay per token, per API call, per inference. This is the AWS/Azure model applied to intelligence — infrastructure pricing for cognitive work. OpenAI, Anthropic, Google, and every major model provider operates here. As models commoditize, margins compress, and the pricing edge goes to whoever runs the cheapest inference.
Pay for a fixed number of concurrent agents, each with defined capability tiers — like paying for a team of 10 analysts vs. 100. This is the model BRNZ uses internally and what we see winning in complex enterprise deployments where headcount-equivalent framing helps procurement teams. It maps to existing budget categories: "we used to have 50 customer service reps; now we have 50 agent-equivalent capacity units."
The Winners: Who Actually Benefits
Not every player loses. The transition from SaaS to AaaS creates massive winners — specifically the infrastructure and orchestration layers that sit below the application.
The real money isn't in building agents — it's in owning the substrate agents run on. The same dynamic that made AWS more valuable than any single web app is playing out again. Model providers, vector stores, agent memory systems, and compute layers are capturing the margin that application-layer SaaS companies used to own.
What This Means for Builders
If you're building a company in 2026, the SaaS playbook is not just outdated — it's a trap. The companies that will matter in 2030 are the ones that understand the fundamental shift: software is becoming labor.
You don't sell tools to workers. You are the worker. Your product isn't software — it's capability. You price on outcomes, not access. You compete on quality-of-work, not feature checklists. Your sales motion isn't "here's a license" — it's "here's a team."
This is precisely the thesis behind every company in the BRNZ portfolio. Not "AI-powered software" — autonomous agent workforces that replace entire functional departments. KENSAI doesn't sell security software; it fields a continuous security operations team. CodeForceAI doesn't license a dev tool; it builds and ships product at human-developer quality, autonomously.
The Uncomfortable Truth for Investors
The venture community has $180B tied up in SaaS multiples that assumed durable per-seat revenue. Those multiples were built on the assumption that software churn was low because switching costs were high. In the AaaS world, switching costs are near zero — you're not migrating data and retraining humans, you're swapping an API endpoint. The defensibility moats that justified 15x ARR multiples are evaporating.
What replaces them? Proprietary data, vertical-specific training, deeply embedded workflow integrations, and — most importantly — trust. The agent vendors who will command premium pricing are those whose agents demonstrably outperform alternatives in measurable, auditable, outcome-linked ways.
That's a fundamentally different competitive moat than "we got here first and our data is locked in your system." It's harder to build, harder to defend, but far more aligned with actual customer value.
The Bottom Line
The SaaS era isn't ending with a bang. It's ending with a quiet, relentless substitution — one enterprise contract at a time, as CFOs realize they're paying for human-shaped software licenses in a world where the humans have been replaced by agents who don't need licenses.
The $2.4 trillion AaaS market isn't a new category sitting next to SaaS. It's a replacement for it. Every dollar in enterprise software spend is now up for grabs by whatever provides the best outcome-per-dollar, regardless of whether it looks like traditional software or not.
The builders who understand this will build the most valuable companies of the next decade. The SaaS founders who don't will spend 2027 wondering why their NRR is collapsing while their product metrics look fine.
Software doesn't have a future. Capability does.
— BRNZ