Enterprise software is undergoing an existential crisis. Recent breakthroughs in AI agents are forcing a rethink of how software is built, delivered, and valued. Many practitioners argue that agent capabilities have crossed a “coherence” threshold, making 2026 a high-energy digestion phase for what this unlocks in real workflows. Andrej Karpathy (Previously Director of AI @ Tesla, founding team @ OpenAI) summed it up best:
“LLM agent capabilities (Claude & Codex especially) have crossed some kind of threshold of coherence around December 2025 and caused a phase shift in software engineering and closely related. The intelligence part suddenly feels quite a bit ahead of all the rest of it - integrations (tools, knowledge), the necessity for new organizational workflows, processes, diffusion more generally. 2026 is going to be a high-energy year as the industry metabolizes the new capability.” – @karpathy, x.com
The shift is bigger than adding copilots to existing products. It challenges the old “build vs. buy” framing, compresses the profit pool for traditional SaaS, and accelerates a move toward a new architecture where agents become the primary interface to enterprise data and action.
Beyond “build vs buy”
For years, enterprises treated new needs as a binary decision: build in-house or buy a vendor product. That’s increasingly the wrong question. The real decision is how to leverage AI to drive outcomes, either by augmenting incumbent platforms or adopting AI-native tools that automate end-to-end work. In practice, enterprises that obsess over replacing SaaS with internal clones often pay a hidden tax: lost focus. Rebuilding commodity tooling to save licensing dollars can distract from core differentiation, precisely when AI is raising the bar for speed and iteration. The priority shifts from “own the software” to “own the advantage”. The winners will apply AI to proprietary workflows, data, and customer experiences rather than reinventing generic systems.
This is in no way a defence of legacy SaaS systems, nor a declaration that these companies are oversold. Even bombed-out names here are cheap for a reason. Instead, what we need to think about is how Software 2.0 looks.
Why incumbents rarely disrupt themselves
If in-house rebuilds are risky, why not just wait for existing enterprise vendors to lead the transition? Two frictions dominate: architecture and incentives. Architecturally, much of SaaS 1.0 is a federation of systems of record, each with its own database, schema, permissions, and UI. Those silos were not designed for autonomous, cross-functional agents. Commercially, incumbents have strong incentives to protect per-seat licensing, proprietary data models, and lock-in. The rational move is often to infuse “just enough AI” to defend renewals (assistants, chat features, incremental automation), not to embrace a full agentic redesign that could reduce seats, flatten modules, or make data more portable.
The result is a classic innovator’s dilemma. Enterprises feel the pain of fragmentation: dozens of tools create an “unnavigable web” of knowledge and process. Some companies respond by consolidating data into unified layers (including graph and lakehouse approaches) to break silos and make workflows agent-ready. But few vendors are eager to accelerate consolidation that weakens their own walled gardens.
AI-native challengers and 10x workflows
This opens the door for AI-native startups that are architected around agents from day one. To displace entrenched vendors, the improvement has to be dramatic, and that is exactly the pitch: agents that remove repetitive data entry, reconcile information across CRM, ERP, support, and productivity tools, and resolve requests autonomously. These products are not “slightly better dashboards.” They are systems of action that complete work. Because they are unburdened by legacy code and pricing models, they can ship faster and package modern capabilities like real-time context retrieval, multi-system orchestration, and outcome-driven automation. In a macro environment where software ROI is under scrutiny, the willingness to switch rises if the value gap is large enough.
The profit pool shifts away from SaaS 1.0
The headline is not that enterprise software shrinks, but that value migrates. Some analysts project that autonomous agents could represent over 60% of software spend by 2030, with the overall application market expanding over 65% beyond baseline because AI enables new categories of work.
The pie grows, but the slice captured by classic point-solution SaaS compresses as agent layers absorb tasks once performed inside separate applications. Related forecasts suggest a meaningful share of standalone SaaS apps (on the order of one-third) could be absorbed or replaced by agent ecosystems over time. For incumbents, survival may not be the only question; share of the profit pool may erode even if they remain relevant.
Cracks in SaaS economics: growth, “profits,” and retention
Public SaaS fundamentals have been softening even before agents fully hit the mainstream. Median growth rates across cloud software have fallen from the high-30s% range (circa 2019) to around 20% by late 2025. Many companies improved non-GAAP margins into the mid-teens through cost control, but GAAP profitability has often remained near breakeven, with stock-based compensation driving a persistent gap between “adjusted” and real earnings power. The Rule of 40 briefly looked better via margin expansion, but that strategy runs out of runway if growth continues to decelerate or competitive pressure rises.
Net dollar retention is the key pressure point. Median NDR has trended down from roughly 120% in 2018–2021 toward ~110% in 2023–2025. That seemingly small change has an outsized impact on valuation because it weakens the “annuity with expansion” story that underpins long-duration software multiples.
If agents accelerate switching or reduce expansion, terminal value assumptions compress fast. A few large platforms have reported steadier retention in recent quarters, but the burden of proof has shifted: investors want to see resilience in NDR in an agent-saturated world.
SaaS 2.0 architecture: from silos to a shared data layer
The agent revolution is also architectural. SaaS 1.0 delivered cloud apps, but still as walled gardens. SaaS 2.0 trends toward centralized data foundations (lakehouses, unified warehouses, knowledge graphs) with modular apps and agents operating on top. The goal is to remove silos and capture “decision traces” across the organization so agents can execute multi-step workflows with full context. In this model, users do not necessarily live inside dozens of UIs. Agents become the interface, and applications become orchestrated services.
A new stack: models, harness, outcome
The term “harness” is now all the rage in tech circles. Think of a foundation model (like GPT-4 or Claude) as a powerful but untamed engine. It has immense capability but lacks direction, safety controls, and specificity. A “harness” is everything layered around that engine to make it:
- Useful (applies to a specific domain, like customer support, finance, or sales)
- Controlled (secure, compliant, and safe for enterprise use)
- Connected (able to read/write from relevant enterprise data sources)
- Orchestrated (able to act autonomously within well-defined boundaries)
The harness is the operationalization layer.
The best model doesn’t win… the best harness wins. In most enterprise use cases, multiple vendors rely on similar (or even identical) underlying models. The differentiator is how well they’ve wrapped that model with contextual understanding, data access, workflow logic, and guardrails.
It’s no longer about who has the biggest model. It’s about who has the best system built on top of it.
What I’m watching
My previous piece mentioned how I have shifted a majority of my fund (~80%) towards the physical economy rather than the digital one. Hardware, materials, and power over software. I will remain tilted that way until I gain more confidence in how the application layer gets built. But once this layer takes off, it will capture a majority of the total value created. This piece is part of that exercise.
The winners will be those who deliver real automation, embrace interoperability, and align pricing with value (usage or outcomes, not just seats). For investors and operators, the questions become concrete: Is AI improving retention and expansion? Is the architecture open enough to plug into the broader ecosystem? Are agents reducing total work, not just adding a nicer interface?
In this upgrade cycle, stasis is death.
Strong Convictions. Loosely Held.
–Nicholas Mersch, CFA
The content of this document is for informational purposes only and is not being provided in the context of an offering of any securities described herein, nor is it a recommendation or solicitation to buy, hold or sell any security. The information is not investment advice, nor is it tailored to the needs or circumstances of any investor. Information contained in this document is not, and under no circumstances is it to be construed as, an offering memorandum, prospectus, advertisement or public offering of securities. No securities commission or similar regulatory authority has reviewed this document, and any representation to the contrary is an offence. Information contained in this document is believed to be accurate and reliable; however, we cannot guarantee that it is complete or current at all times. The information provided is subject to change without notice.
Commissions, trailing commissions, management fees and expenses all may be associated with investment funds. Please read the prospectus before investing. If the securities are purchased or sold on a stock exchange, you may pay more or receive less than the current net asset value. Investment funds are not guaranteed; their values change frequently, and past performance may not be repeated. Certain statements in this document are forward-looking.
Forward-looking statements (“FLS”) are statements that are predictive in nature, depend on or refer to future events or conditions, or that include words such as “may,” “will,” “should,” “could,” “expect,” “anticipate,” intend,” “plan,” “believe,” “estimate” or other similar expressions. Statements that look forward in time or include anything other than historical information are subject to risks and uncertainties, and actual results, actions or events could differ materially from those set forth in the FLS. FLS are not guarantees of future performance and are, by their nature, based on numerous assumptions. Although the FLS contained in this document are based upon what Purpose Investments and the portfolio manager believe to be reasonable assumptions, Purpose Investments and the portfolio manager cannot assure that actual results will be consistent with these FLS. The reader is cautioned to consider the FLS carefully and not to place undue reliance on the FLS. Unless required by applicable law, it is not undertaken, and specifically disclaimed, that there is any intention or obligation to update or revise FLS, whether as a result of new information, future events or otherwise.