Why Being a System of Record Isn't Enough for Growth
Retention vs. growth: The widening gap in enterprise software economics
1ServiceTitan's decisive move
ServiceTitan terminated its 9-year integration with Podium, cutting off 1,000 shared customers. Despite this aggressive move, it retained nearly all accounts because contractors rely on it for core operations like job scheduling and customer history. This demonstrates the power of being the System of Record - customers won't leave even when forced to abandon useful add-ons. However, the company's 25% growth rate, while strong for vertical software, pales compared to data platform players like Snowflake (34%) and Databricks (80%+).
2The retention-growth paradox
Systems of Record excel at customer retention but struggle with acceleration. Salesforce's applications business grew just 7% last quarter despite processing 28.6 trillion tokens (up 152% QoQ). Meanwhile, its Data 360 platform handling external records grew 25%. This divergence shows customers aren't abandoning their Systems of Record - they're just spending new money elsewhere. Veeva's 16% growth during a CRM migration further proves retention remains strong while growth requires different capabilities.
3Prohibitive data costs
Salesforce charges $3,000/GB/year for data storage - 500x more than cloud object storage. This pricing model worked when human-generated data volumes were small, but becomes unsustainable with AI agents. A single scoring job in one implementation created 164,523 rows - typical agent workloads generate tool traces, reasoning logs, and retry data that's mostly disposable. Salesforce's Zero Copy solution, where 35 trillion records (67% of total) remain external, acknowledges this reality but forfeits storage revenue.
4API limits before storage costs
Enterprise platforms face agent-scale challenges beyond storage costs. Salesforce allocates API calls per user per day - a model that breaks when AI agents don't correspond to human seats. ServiceTitan's 2026 API terms prohibit AI systems from independently selecting endpoints, creating permission barriers even for paying customers. These constraints reflect a fundamental tension: Systems of Record must choose between controlling agent ecosystems or enabling their growth, with most currently opting for tighter restrictions.
5The control plane battleground
Salesforce and Snowflake are converging on the same strategic layer from opposite directions. Salesforce's Zero Copy approach maintains metadata control while relinquishing physical data storage. Snowflake positions itself as the 'control plane for the Agentic Enterprise' through acquisitions like Natoma. This race to manage distributed agent workflows represents the real growth frontier, separate from traditional System of Record value. The winners will monetize coordination, not just data custody.
6Incumbency doesn't guarantee agent wins
Podium built $100M in AI agent ARR within ServiceTitan's customer base, proving System of Record status doesn't automatically translate to agent dominance. Open platforms become one bidder among many - SaaStr uses Salesforce alongside Artisan, Monaco, and Qualified for different functions. Closed platforms risk making their record the reason customers leave when superior external agents emerge. Marketo's decline shows how workflow rigidity can turn retention assets into migration liabilities over decades-long horizons.
7Enduring System of Record value
Despite growth challenges, Systems of Record maintain crucial roles: resolving data disputes (authority), maintaining audit trails (compliance), and hosting final human approvals (workflow). Veeva's 16% growth during its CRM transition shows regulated industries particularly value these functions. Distribution advantage also persists - incumbent agents get first consideration in existing workflows. These durable advantages support 10-16% growth, but exceeding that range requires capturing agent-driven expansion beyond core record-keeping.
8Strategic implications for vendors
System of Record providers face three imperatives: 1) Their retention moat protects the base but doesn't drive growth, which must come from usage-expanding units. 2) Proprietary storage pricing caps their own agents' viability, forcing separate data layer architectures. 3) They must choose between open ecosystems (competing on agent merit) or closed ones (risking eventual customer defection). The path forward involves monetizing agent coordination while maintaining record integrity - a balance few have mastered.
Original article: SaaStr — Why Being a System of Record Isn't Enough for Growth