The Great Platform Reckoning: Will A System Of Record Save SaaS?

The industry is obsessed with the wrong questions. Analysts ask: “Will AI disrupt SaaS?” Vendors ask: “How do we add AI to our product?” CIOs ask: “Which AI tools should we buy?”

The right question is the one PE firms are now asking when they look at portfolio companies, and it is far more dangerous to incumbents.

Does our current technology stack cap our ability to monetize AI, and at what point does that cap become an existential constraint on revenue growth, margin expansion, and competitive positioning?

That reframing changes everything. It turns tech stack rationalization from a cost optimization exercise into a strategic derisking activity…one that has the power to trigger a complete reshuffling of the enterprise platform landscape.

The System of Record & Illusion Of A Moat

The standard defense from incumbent SaaS vendors is: “We are the system of record. We hold the data, and it is heavy. Data’s gravity keeps customers locked in. Therefore, we are safe.”

This was a reasonable argument in 2024. Today, almost $2 trillion in market capitalization has evaporated from the software sector in the SaaSpocalypse. Software price-to-sales ratios have compressed from 9x to 6x. Workday shares are down over 20% year-to-date. Salesforce dropped 28%. Atlassian fell 35%. And on and on.

This quarter’s earnings aren’t the problem. The market is repricing the structural defensibility of being a system of record in an agentic AI world. The system of record isn’t the moat most software vendors believe it to be.

The paradigm most people are overlooking is that being a system of record was never actually a moat. It was a side effect of the real moats: switching costs, workflow embedding, UI-driven habit formation, and data format lock-in. The system of record status was the result of those dynamics, rather than their cause. AI agents are now systematically dismantling each of those underlying moats.

Breaking Down The Castle Walls

Switching costs are collapsing. Amazon, Microsoft, Salesforce, and Palantir are racing to launch AI code generation tools that help businesses migrate massive datasets between platforms. CIOs report they’ve already begun saving money as software switching costs fall.

Federal agencies, truly the slowest to transform or adapt, are testing AI models from Microsoft and OpenAI to extract data from analytical applications operated by contractors like Palantir and Lockheed Martin. The cost of data migration was once the single greatest barrier to platform switching. It is rapidly becoming a commoditized capability.

Workflow embedding is being abstracted. Agentic AI doesn’t need a large UI surface. Agents interface with systems via API, MCP, direct database access, or whatever works.

When an agent can navigate a clumsy interface just as easily as a sleek one (or bypass the interface entirely, as I’ll explain later), the billions invested in UX design become irrelevant to stickiness. That’s one reason Anthropic’s announcement today is a big deal. The company is making rapid advancements in how well models can use software like people do. OpenAI, with its recent takeover of OpenClaw, is likely headed in the same direction soon.

The ‘time-in-tool equals stickiness’ assumption collapses when the user is an agent that doesn’t experience friction.

Data format lock-in is also dissolving. LLMs can now parse unstructured data across silos, making carefully structured proprietary schemas less of a competitive advantage. Users are discovering they can dump raw data into an LLM’s context window and ask questions in natural language.

The app layer is becoming optional. Legacy vendors find themselves charging thousands of dollars for what is effectively a glorified database while a consumer-grade AI does the actual cognitive work of analysis and strategy.

The real moat remains, and one analyst captured it precisely: “The moat was never ‘we store your data.’ It was ‘we’re the system you trust with your data.’” Trust requires governance, auditability, and control. Those elements matter significantly more in an agentic world.

The systems of record that survive will be the ones that make AI interaction trackable, reversible, secure, and auditable by default. The ones that die will be the ones fighting the transition by restricting API access and imposing rate limits. That’s the enterprise software equivalent of the music industry suing Napster. They weren’t wrong about ownership. They were catastrophically wrong about strategy.

See The Complete Course Outline & Enroll Here.

The PE Lens: Tech Stack as Portfolio Risk

What happens when we reframe the enterprise’s technology stack as an investment position? Private equity firms have a clarifying way of looking at business problems. Everything is a risk-adjusted return calculation. Apply that lens to a company’s technology stack, and the analysis becomes brutally clear.

PE firms are now treating technological stagnation as a critical red flag in due diligence. The warning signs are outdated IT systems, resistance to technological change from management, and a lack of data collection or analysis capabilities. These factors suggest companies may struggle to adapt as AI capabilities advance.

Leading firms have adopted what insiders call the barbell strategy. They invest heavily in both AI-enabled companies and AI-resistant businesses (regulated, physical, relationship-dependent) while systematically avoiding the vulnerable middle ground.

That vulnerable middle is where most enterprise technology stacks live.

A 5-Point Tech Stack Derisking Template

When a PE operating partner evaluates technology risk across a portfolio, the framework they use maps directly to how any business should evaluate its own stack.

Revenue Ceiling Risk

The wrong platform does more than waste money. It caps revenue growth. If your CRM can’t support AI-driven personalization at the speed your competitors deliver it, you have bigger problems than those posed by the technology. The business has a market share problem. AI enables competitors to replicate offerings quickly and inexpensively. Any technology that slows your cycle time from insight to action becomes a drag on revenue velocity.

Margin Compression Risk

Enterprise software spending is forecast to rise significantly by 2027, with generative AI as the primary accelerant. But there’s a trap. Vendors lure customers with generous pilot credits, yet scaling to production routinely reveals 500–1,000% cost underestimation. If your tech stack forces you into consumption-based pricing with a vendor whose costs scale nonlinearly, your margins erode with every increment of AI adoption. The platform choice is the margin decision.

Decision Velocity Risk

The emerging competitive differentiator is decision velocity: how quickly smaller decision chains and processes can be automated at scale. A tech stack built around human-in-the-loop workflows at every step physically cannot compete with one designed for delegated autonomy on high-frequency, low-risk decisions. A legacy stack enforces an organizational speed limit.

Data Portability Risk

Enterprise data tolls and API economics are becoming a major pain point. Celonis is suing SAP over data access. Salesforce has been raising prices on applications that tap into its data. Connector fees are looking like the new cloud egress charges. They are a tax on your own data that increases the cost of doing anything new with it. If your vendor treats your data as leverage rather than infrastructure, every AI initiative faces a hidden tax.

Talent and Adoption Risk

PwC’s data reveals that only 14% of workers use AI daily, yet daily users report dramatically better outcomes. 92% report productivity benefits compared to 58% of infrequent users. A tech stack that creates friction for AI adoption isn’t just a technology liability; it’s a people liability. It widens the gap between what AI could deliver and what your organization actually captures.

The Reshuffling: What the New Landscape Will Look Like

Learn More About My Courses & Take The Next Steps.

I teach this framework in most of my courses. The critical takeaway for this article is that value is migrating away from the middle or core. That middle layer of horizontal SaaS (project management, basic CRM, content creation, coding, data entry, etc.) is being flattened. Value is flowing to the edges: compute infrastructure at the bottom and a deeply embedded, business and customer-touching pane of glass at the top.

There will be some overlap, but most of what’s in between faces commoditization. No-code data access is table stakes, but many software platform vendors treat it like a value add. These types of disconnects are stacking up.

Structurally Protected

• Deeply regulated verticals where accuracy, compliance, and auditability matter more than raw intelligence (government software, insurance claims, digital forensics, credit scoring)

• Hardware-dependent data systems (industrial IoT, CAD/physics-based modeling).

• Companies with non-software moats like network effects in the physical, financial, and regulatory world that software alone cannot reproduce.

• Vertical SaaS with genuine proprietary data and domain complexity (electronic health records and pharmaceutical lifecycle management).

Structurally Exposed

• Horizontal SaaS that functions primarily as a UI on a structured database (project management, basic CRM, scheduling, form builders).

• Per-seat pricing models where AI agents directly reduce the number of human seats needed.

• Products where the core workflows are stable, data-driven, and rule-based.

• Vendors who have responded to the AI threat by restricting API access rather than embracing agent interoperability.

Emerging ‘Agent-Native’ platforms are a bit of a wildcard. AI-native startups are funded by capital markets rather than operating cash flow, allowing them to undercut incumbents on price while offering superior automation. The cost curve trajectory of foundation models accelerates downward even as accuracy improves. OpenAI’s latest frontier reasoning model dropped 80% in cost in just two months. This means the gap between what incumbents charge and what agent-native alternatives deliver only widens over time.

The Overlooked Paradigms

The Napster Inversion

The most important dynamic that no one is discussing is the direction of disruption in the platform stack. In previous technology transitions (mainframe → client-server → cloud → SaaS), disruption moved from infrastructure upward toward the application layer. Companies that owned the infrastructure lost, and companies that owned the application experience won.

This time, disruption is moving in the opposite direction. AI is commoditizing the application layer (the UI, workflow, and logic) and pushing value back down toward data infrastructure and out toward outcomes. The companies that spent two decades building beautiful interfaces and sticky workflows are discovering that their most valuable asset is the database underneath. And that asset is far less defensible than they thought.

This is what I call the Napster inversion. The music industry thought it was in the album business. It was actually in the distribution business, and when distribution was commoditized, albums became unbundled tracks on someone else’s platform.

Enterprise SaaS vendors think they’re in the application business. They’re actually in the data custodian and data distributor business. When AI agents can access, migrate, and reason over that data without the application layer, the application becomes an optional wrapper. Commoditization is inevitable.

The Rationalization Cascade

Tech stack rationalization isn’t a one-time event with a single outcome. It’s a cascade of network effects. When a business rationalizes one platform (like replacing a legacy CRM with an AI-native alternative), the implications ripple across the entire stack.

• The data integration layer changes, which forces re-evaluation of the middleware.

• The middleware change reveals that the ERP’s data model is now the bottleneck.

• The ERP evaluation reveals that the BI layer was compensating for bad ERP data.

• Removing the BI layer reveals that reporting was masking operational inefficiency.

Each rationalization decision creates new information about the rest of the stack. This is why PE firms that force one technology change in a portfolio company often end up triggering a complete platform overhaul within 18 months. The first pull on the thread reveals how deeply interconnected and mutually reinforcing technology debt actually is.

This is why I talk so much about workflow reorchestration. The same network effects happen when you start pulling on workflow threads, and that will be another flywheel for tech stack rationalization.

This is what makes tech stack rationalization strategically urgent rather than merely operationally important. It’s not about saving on SaaS licenses. In the process, leaders discover the hidden constraints that their current stack imposes on their business’s ability to capture AI-driven value.

The Competence Trap

The biggest barrier to tech stack rationalization is often organizational. Enterprises have built decades of institutional competence around their current platforms. Salesforce admins, SAP consultants, and Oracle DBAs are organizational identities and power structures.

When you rationalize the tech stack, you’re not just replacing software. You’re threatening the competence base of the people who currently hold organizational power. This is why only 20% of finance leaders report satisfaction with recent technology investments, and why ERP upgrades are cited among the lowest-value initiatives despite their cost and disruption. The platforms are failing organizationally vs technically because the people who manage them have every incentive to preserve the status quo.

PE firms solve this problem by replacing management. Most operating companies don’t have that luxury. The alternative is to frame rationalization not as a threat to existing competence but as an expansion of it. That’s why I emphasize decreasing resistance pain through increased competence trust, intent trust, and process trust.

The Pricing Model Is the Strategy

IDC predicts that by 2028, 70% of software vendors will have refactored their pricing around new value metrics like consumption, outcomes, or organizational capability. Gartner forecasts that 40% of enterprise SaaS will include outcome-based elements by 2026.

However, the forecasters are missing one key element. The transition from seat-based to outcome-based pricing is more than just a pricing change. It’s a business model transformation that most incumbents structurally cannot execute. Their cost structures, sales compensation plans, customer success organizations, and financial reporting are all built around predictable per-seat revenue. Moving to outcome-based pricing requires accepting revenue volatility, investing in measurement infrastructure, and fundamentally redefining what customer success means.

The same structural trap caught legacy media in the digital transition. They could see the future but couldn’t get there because their current operating model was optimized for the past. The SaaS vendors racing to become AI companies face the identical constraint.

Salesforce’s new Agentic Enterprise License Agreement (AELA) is an attempt to bridge this gap, which reveals the tension. “AELA is for customers that have already experimented. They want to go all in so we agree on a flat fee, and then it’s a shared risk.” Shared risk is a fascinating admission from a company that built its empire on eliminating risk from software procurement.

The AI Readiness Tax

Here’s the paradigm that ties everything together, and the one that should anchor any strategic framework on this topic. Every dollar of technical debt in your current stack is now an implicit tax on your AI transformation. One-time migration costs are over. This is an ongoing, compounding tax that widens the gap between you and competitors with AI-ready tech stacks.

1. Integration tax: Time and cost to connect AI capabilities to legacy data.

2. Latency tax: Decision velocity constrained by humans having to audit everything vs. automating low-risk decisions and repetitive, stable workflows.

3. Talent tax: Best AI talent won’t work with legacy stacks, so you pay a premium or get second-tier people.

4. Opportunity tax: Revenue opportunities you can’t pursue because your stack can’t support them.

5. Switching tax: The longer you wait, the more organizational process calcifies around the current stack, making future rationalization harder and more expensive.

The compounding nature of this tax is what makes it so dangerous. It’s an exponential degradation. Companies that rationalize their stacks in 2026 will have two to three years of compounding AI capability gains over companies that wait until 2028. In a market where AI-driven decision velocity is the competitive differentiator, that gap may be insurmountable.

The Strategic Imperative

Tech stack rationalization will become one of the most consequential strategic decisions most enterprises make. The PE lens makes this unambiguous. The wrong platform creates risk across every dimension that matters, from revenue growth to margin expansion, decision velocity, talent acquisition, and the ability to meet rapidly evolving customer expectations.

When PE firms evaluate portfolio companies, technological stagnation is now a red flag on par with declining unit economics or management quality issues. This needs to be the enterprise view on its internal tech stack as well.

The system of record will not save traditional SaaS vendors. What will save some of them is the willingness to transform from closed, UI-centric, seat-licensed platforms into open, agent-interoperable, outcome-priced data infrastructure. The vendors that make AI interaction trackable, reversible, and auditable will survive. The vendors that fight the transition with rate limits and connector fees will be replaced.

For the businesses using enterprise software, the message is equally clear. Rationalize your stack now, while you have leverage, while switching costs are falling, and while the compounding AI readiness tax is still manageable. Every quarter of delay widens the gap between where you are and where the market is moving.

The agents are coming. The only question is whether your platforms are ready to work with them or ready to be replaced by them.