The Real Reason Your System of Record Won't Grow: Field Service Data Tells a Story

Imagine running a plumbing business with 14 techs on the road. Your entire operation—scheduling, invoicing, customer history—lives inside ServiceTitan. Last month, they shut off your integration with Podium, the AI agent you rely on for customer communication. Do you rip out ServiceTitan? No. You find another way to keep Podium working, even if it means manual workarounds. You're locked in, and you know it.

Now imagine you're a builder watching this from the outside. What do you build next? The answer isn't a new system of record. It's a tool that sits on top of the existing one and fixes the specific pain points that ServiceTitan—and every other vertical SaaS platform—can't or won't address.

Jason Lemkin's piece on SaaStr, Being a System of Record Helps With Retention. But Alone, It Won't Equal Growth, brilliantly dissects the tension between lock-in and innovation. But he misses the ground-level view from the industries most affected: construction, trucking, and field services. Our data across 25,522 tracked problems reveals a more urgent angle.

Lock-in cuts both ways

Lemkin points out that ServiceTitan's API terms from April 2026 bar AI systems from independently choosing endpoints, effectively fencing off the agent ecosystem. He frames this as a defensive move to protect the record. That's true, but it's also a blueprint for where the opportunities are.

In our database, we track 742 high-severity problems in construction and 930 in trucking & logistics. Many are rated 5/5 severity with explicit willingness to pay. But here's the key: almost none of these problems require replacing the system of record. They require augmenting it.

Take double brokering fraud in trucking. LoadVerify Pro addresses this with a lightweight verification tool that sits on top of existing TMS platforms. Severity 5/5. Willingness to pay: clear. Does it compete with the TMS? No. It patches a hole the TMS left open. That's the pattern.

The data gravity is heavier in vertical SaaS

Lemkin notes that Salesforce's applications business grew just 7% while its data layer grew 25%. He argues that systems of record become expensive databases, and agents produce too much data to store cheaply. But in construction and trucking, the problem isn't data volume. It's data fragmentation.

A contractor juggling ServiceTitan for operations, QuickBooks for accounting, and a hodgepodge of spreadsheets for job costing doesn't care about token counts. They care that their field techs' hours don't match the invoices. They care that change orders get lost. They care that compliance reports take three days to compile.

Our data shows high-severity problems around cash flow, compliance, and safety—areas where AI agents could add value but also where the system of record is often the source of friction. MedCard Navigator tackles DOT medical card compliance with severity 5/5. FairDock Accessorial Resolver handles disputes with Amazon's automated accessorial pay system—another 5/5. These are not world-changing platforms. They are scraping exhaust off a system that was never designed for the job.

The agent sale is won on the edge

Lemkin argues that owning the record doesn't win you the agent sale, citing Podium's $100M AI agent ARR built largely inside ServiceTitan's customer base. He's right, but the corollary is more interesting: the agent sale is won by whoever reduces the daily friction fast enough.

In our tracking of app ideas across 89 industries, the ones gaining traction are not trying to be the next system of record. They are solving one painful problem with a simple, vertical-specific agent. LoadRate Pro, for instance, has 15 signals with average severity 4.1—strong traction for an add-on that doesn't attempt to displace the underlying platform.

This suggests a new investment thesis: don't back the next system of record. Back the exhaust scrubber.

The forced transition window

When ServiceTitan cut off Podium, roughly 1,000 shared customers were suddenly scrambling. In our data, we see similar forced transitions across industries. Amazon's automated accessorial pay system denying detention pay. DOT officers fabricating violations. These moments create windows where incumbents can't or won't move, and users are desperate for a fix.

LegalGuard Recorder, for example, captures evidence during roadside inspections—a direct response to system-driven penalties. It doesn't replace the ELD. It gives drivers ammunition. Severity 5/5. Willingness to pay: high, because the alternative is lost income.

These are not just product opportunities. They are go-to-market opportunities. A tool that can be sold to the 1,000 accounts Podium just lost would have a ready-made customer list and a burning pain. The switching cost is low because the tool doesn't require switching anything.

The real race is for workflow control, not data storage

Lemkin's sharpest observation is that Salesforce's fastest-growing data motion is Zero Copy—the one where data doesn't physically move into Salesforce. He frames this as the system of record giving up storage rent to stay the control plane. That's exactly right, and it's happening in vertical SaaS too.

But in field services, the control plane is the workflow. The system of record owns the workflow, but it doesn't own the edges. AI agents that can plug into the workflow without moving data—answering customer texts, flagging invoice discrepancies, checking compliance—are the ones that win. They don't need to store the data. They need to act on it.

The builders who understand this will build small, focused agents that attach to existing systems like barnacles. They won't compete with ServiceTitan. They'll make ServiceTitan tolerable. And that's a much bigger market than anyone thinks.