The Durability of SaaS in the Agentic Era

We recently shared some perspectives on the “Coming Age of Reason” we are seeing in software. We believe the agentic era will expand what software can do while raising the bar for what users demand. Just as with the cloud era, new category leaders will emerge, and enduring companies will evolve. Against this backdrop, Mayowa and I are sharing some reflections on the SaaSpocalypse and what we are watching next.

The SaaSpocalypse is real: public software stocks trading down meaningfully even after reporting objectively strong quarters. 

The question has largely shifted from “How did you do?” to “How long does this last?” Many investors seem to be underwriting software around two core variables:

1) The lifespan of moats in an AI-accelerated world

AI is seemingly compressing the half-life of feature advantages. When “good enough” can be replicated faster and agents can drive step-function productivity vs. humans operating software, we see markets discounting the duration of premium growth and pricing power.

In this environment, forward-looking contracted revenue/remaining performance obligations (RPO) quality often gets scrutinized:

• Longer time-to-earn means more surface area for renegotiation at renewal
• Higher substitution risk from internal builds, agent layers, or adjacent platform bundling
• More scrutiny on organic vs. inorganic contribution (especially amid prolific M&A) and what backlog really represents

2) The speed to durable, self-funded profitability

Software has historically benefited from investors underwriting revenue multiples. In a more competitive, AI-accelerated environment, we believe the urgency rises to prove SaaS companies can become durably cash generative before the perceived advantage window narrows. This is why we’re seeing the debate shifting toward stock-based compensation (SBC), operating leverage, and GAAP profitability, not just growth.

We're already seeing this in real-time. Atlassian froze hiring for engineers and Meta reduced stock awards for most employees last week. We expect the pattern of headcount restraint and SBC optimization to accelerate across the sector as companies try to hold both truths simultaneously: invest aggressively for the AI transition while demonstrating to the market that the underlying unit economics hold. The market’s message to management teams is increasingly: you don't get the growth multiple back without first proving the cash engine works.

This does not imply software is dead. It implies the market is recalibrating its assumptions around durability. We believe the most durable businesses will build from their strengths towards an agentic future. This will mean an expansion of what software can do, followed by a separation of who can do it at the highest level. 

The underappreciated point: governance

The popular narrative seems to be that AI makes software easier to build and therefore easier to replace. But enterprise buyers don’t pay for “an app.” They pay for a system they can run safely and accountably: security, permissions, auditability, change control, uptime, liability, and integration into a complex web of workflows.

The “software-making-software” loop expands the supply of “good enough,” but we believe it also raises the premium on governance and operational trust. In many ways, governance is the modern expression of the data context layer. It is not just structured data, but the permissions, audit trails, embedded policies and institutional memory that makes data usable inside complex organizations.

History is a useful guide: open source dramatically lowered build costs, but enterprises still paid for supported, governed delivery (e.g., VMWare, RedHat, MongoDB, Confluent, HashiCorp). AI may follow a similar pattern. It accelerates replication, but does not remove the enterprise requirements that sustain willingness to pay. Distribution and switching costs still matter materially in the near term.

That’s why we think the selloff is directionally understandable, but in some cases overweighting fear. There’s also an argument that platforms already sitting in the workflow and governance path can become more valuable if they deploy AI quickly at scale and leverage how deeply embedded they are. As we shared last week, we see two distinct layers forming - The Data Context Layer (platforms with deeply rooted domain expertise, embedded workflow context, and hard-earned trust) and The Agent Layer (AI systems that leverage the data context layer, converting the raw inputs to automate increasingly complex, long-form tasks). The question will be which SaaS incumbents will successfully be able to span between both layers and continue to capture value.

This thesis is being tested in real-time. Earlier this week, OpenAI launched its Frontier Alliance, embedding FDEs alongside BCG, McKinsey, Accenture, and Capgemini to move enterprises from pilots into production workflows, which suggests that model labs could directly compete for the governance positioning we believe incumbents have.

What we’re watching next

It is hard to separate a short-term selloff from a structural rerate, but we believe there is likely truth to the latter as AI-native vs. SaaS competition intensifies. The leading indicators we’re watching:

• Renewal pricing and expansion behavior
• Proof AI is moving from pilots into embedded workflows
• Whether incumbents extend defensibility via less easily replicable advantages (e.g., deeper workflow ownership, partnerships, monetization into new revenue streams that stretch far beyond additional software features, network effects)

2026 is shaping up to be a real test. The speed at which the AI layer is evolving compresses the window for incumbents to react, which is precisely why 2026 feels like an inflection point rather than a gradual transition. If agents become the primary interface and systems of record recede into passive databases, value capture could shift meaningfully.

Our current read from enterprise decision makers is that most are still in “embed AI into existing workflows” mode, not “rip and replace systems of record.” The career-risk asymmetry is real for now.

We believe the market is asking the right questions. It may simply be answering them with too broad a brush by treating all moats as equally fragile in a world where shallow ones compress and deep ones compound.