I was listening to a recent episode of the Pivot podcast with Scott Galloway and Kara Swisher, and I heard the argument that SaaS vendors will be just fine in the age of AI. The reasoning: R&D is only a small portion of their budgets, so AI-driven efficiency gains won’t fundamentally disrupt their economics. They’ll just absorb AI into their tooling and carry on.
I’m not convinced they fully see the magnitude of what’s happening.
Right now, I’m watching multi-month projects collapse into days — delivered with clean architecture, strong unit tests, and minimal rework. My teams are building ontologies on top of enterprise data that allow agents to reason, generate business insights, and automate end-to-end workflows. For some, that may sound advanced. From where I sit, it feels like we’re just getting started.
So here’s my take — in true Pivot fashion.
The Wrong Unit of Analysis
The “SaaS will be fine” argument measures AI’s impact through an input cost lens — R&D as a percentage of revenue. That’s the wrong lens entirely.
The right question isn’t what does it cost to build the software. It’s what does it cost an enterprise to achieve an outcome — and whether SaaS remains the most efficient path to get there.
AI collapses that cost curve in ways that make the R&D ratio argument irrelevant. You don’t need to blow up a vendor’s cost structure to disrupt them. You just need to eliminate the buyer’s dependency on them.
The data is already signaling this. Enterprise SaaS revenue growth decelerated from 21% annually just a few years ago to just 10% in 2024, with Q1 2025 showing outright contraction at -2% quarter-over-quarter. Major SaaS companies including Salesforce, ServiceNow, Adobe, and Workday saw share prices tumble — in some cases more than 40% from recent highs — collectively wiping out close to $390 billion in market value, even as revenues nominally continued growing. The market is pricing in structural disruption before the income statement shows it. That’s worth paying attention to.
The New Framework: Buy the Substrate, Build the Differentiation
Here’s where I want to reframe the conventional “build vs. buy” debate, because I think it’s being asked the wrong way.
The question is no longer should we build software or buy it. That framing is obsolete.
The new question is: Where do you buy leverage, and where do you build moats?
The answer is becoming clear. Enterprises will buy the substrates — foundation models, orchestration platforms, data infrastructure, governance tooling, compliance layers. These are commodities. Competing vendors will fight over them, costs will compress, and there’s no durable advantage in building what dozens of well-funded companies are already racing to perfect.
But the differentiation layer — the agentic workflows that encode your institutional knowledge, your domain-specific ontologies, your customer intelligence, your operational logic — that layer cannot be bought. It must be built. And critically, it compounds. Every workflow refined, every feedback loop closed, every agent trained on your proprietary data makes it harder for a competitor running off the same foundation model to catch up.
When everyone buys from the same AI substrate providers, the competitive advantage shifts entirely to the layer above: how you orchestrate intelligence, how you encode years of domain expertise, how your systems act autonomously within the context of your business. That’s the moat. And you can’t outsource a moat.
Three Structural Shifts Already Underway
1. The Seat-Based Pricing Model Is Breaking
Fixed SKUs and per-seat licensing were built for a world where humans operated software. As AI agents take on more of the work, those structures feel increasingly arbitrary — and expensive.
IDC is already calling it: by 2028, pure seat-based pricing will be obsolete, with 70% of software vendors forced to refactor pricing strategies around consumption, outcomes, or organizational capability metrics. Some vendors are already reporting slower growth in seat count as customer companies become more efficient with AI. The pricing model isn’t just being questioned — it’s structurally misaligned with how work actually gets done now.
Think about what happens when an AI agent executes 10,000 workflow steps that previously required 50 licensed users. The per-seat model produces absurd economics. Buyers are noticing. The vendors who adapt early — moving toward outcome-based or usage-based pricing — will survive the transition. Those who defend seat counts as a revenue line will find their enterprise customers increasingly incentivized to find alternatives.
2. The UI Is Becoming a Supervisory Layer, Not a Work Layer
When agentic systems can execute workflows directly via API, the user interface becomes less central. It doesn’t disappear — but it degrades from the product’s primary value-delivery surface to an oversight and intervention layer.
This is a profound shift for SaaS vendors whose entire product differentiation is the user experience. If agents bypass the UX to hit APIs directly, what exactly are they selling? IDC’s framing describes this as “headless” software modules — process teams designing workflows around end-to-end outcomes rather than application silos. Humans won’t disappear, but they’ll increasingly supervise, guide, and correct agents rather than clicking through interfaces step by step.
3. The “User” as an Organizational Construct Is Dissolving
If software can reason across systems and execute tasks autonomously, you don’t need “users” in the same way you once did. The shape of work changes.
Gartner predicts that by 2028, 33% of enterprise software applications will include agentic AI — up from less than 1% in 2024 — enabling 15% of day-to-day work decisions to be made autonomously. That’s a conservative Gartner forecast. The organizational implication is significant: the “named user license” will eventually feel like paying per typewriter key struck. It’s not just a pricing problem — it’s a fundamental mismatch between the unit of value the vendor is selling and the unit of value the buyer is consuming.
What to Watch: The Leading Indicators
If this thesis is correct, the following signals should appear in sequence. Some are already flashing.
Pricing signals: Watch for acceleration of consumption-based or outcome-based SaaS contracts in new enterprise deals. Watch for public disclosures from major vendors announcing “agent pricing” tiers. Watch for Net Revenue Retention compression at seat-heavy players — Salesforce, Workday, HubSpot specifically. The number of SaaS applications per company has already begun declining: from 112 average in 2023 to 106 in 2024, with mid-sized firms showing a 29% reduction in 2025.
Usage signals: The tell will be declining DAU/MAU ratios at enterprise SaaS products even as ARR holds flat. Agents will log in; humans won’t. Growth in API-layer traffic without corresponding UI interaction growth is the technical fingerprint of this transition.
Capital markets signals: The $390 billion in SaaS equity destruction already noted is the market pricing in structural disruption before it shows up in earnings. Watch for continued multiple compression for seat-based SaaS versus premium expansion for AI-native, consumption-based competitors. Nadella’s “SaaS is dead” comment in December 2024 wasn’t random — it was the CEO of the world’s most valuable software company telegraphing where the industry is heading.
Workforce signals: The job posting mix is already shifting. Declining demand for “SaaS administrator” roles; rising demand for “AI workflow engineer,” “agent governance specialist,” “LLMOps engineer.” When the workforce moves, the software economics follow.
The Lagging Indicators (These Confirm It’s Done)
These will appear 12–36 months after the leading indicators, and by the time they’re visible, the transition will be well underway:
- Actual ARR declines (not just growth deceleration) at seat-heavy SaaS vendors
- M&A activity driven by SaaS vendors desperately acquiring AI-native workflow companies
- Major vendors discontinuing seat-based pricing tiers entirely
- Regulatory and governance frameworks specifically governing enterprise AI agent deployment — the compliance and governance stack I believe becomes the new center of gravity
- Enterprise CIO surveys showing >30% of organizations have eliminated at least one SaaS license category due to agentic replacement
The Honest Counterargument
I want to be fair to the complexity here, because the picture isn’t purely one-directional.
The most significant counterforce is enterprise procurement inertia. Recent data from Menlo Ventures shows that 76% of enterprise AI use cases are now purchased rather than built internally — a complete reversal from 2024. This appears to contradict the “companies will build” thesis. But look carefully at what they’re buying: AI substrates. Foundation models. Orchestration platforms. Not traditional workflow SaaS. The build is happening at the differentiation layer above those substrates — exactly the framework I’m describing.
Additionally, SaaS vendors aren’t standing still. Revenue per employee jumped 17% last year across enterprise SaaS, with some companies posting 47% gains. They’re using AI aggressively to improve their own economics, which extends their survival window while they replatform. The disruption is real, but it will arrive as a series of threshold events — specific use cases and verticals crossing a viability point for agentic replacement — rather than a clean linear decline. Enterprise technology markets have a remarkable ability to slow down even obvious transitions through procurement cycles, multi-year contracts, and entrenched vendor relationships.
The direction is clear. The speed remains genuinely uncertain.
The Center of Gravity Is Moving
Today’s SaaS platforms don’t disappear overnight. But their center of gravity changes — and the companies that correctly anticipate where it’s moving will gain a durable advantage.
The new center of gravity is: compliance, data governance, agent governance, feedback loops, and the foundational platforms that make all of it possible.
The companies that thrive in this environment won’t just buy software off a shelf — they’ll build proprietary agentic workflows on top of purchased substrates. They’ll operate those agents at scale. They’ll close feedback loops that make their systems smarter with every interaction. And they’ll treat the governance and compliance infrastructure not as overhead, but as a competitive capability in its own right.
In that world, the question shifts from “which SaaS vendor should we buy?” to “what is our agentic architecture, and where does proprietary advantage live within it?”
That’s a fundamentally different conversation. And the speed at which enterprise leaders start asking it may surprise people — including the ones on podcasts telling us SaaS will be just fine.
