I’ve been talking to a lot of B2B and AI leaders lately, and there’s a narrative I keep hearing from incumbents that’s starting to worry me:
“We’re the system of record. We’re safe.”
They’re not wrong per se. Not exactly. If you’re the system of record for HR, for finance, for CRM, for supply chain — you’re probably not going to see massive AI-driven churn. Not yet, at least. Ripping out a system of record is painful, expensive, and nobody wants to do it. That’s been true for decades and it’s still true today.
But here’s what I’d push back on: not churning isn’t the same as growing.
The Spend Is Moving. Fast.
Every CIO and VP I talk to right now is telling me the same thing: their incremental budget is going to AI. Not to expanding seats on their existing platforms. Not to upgrading tiers. To AI agents, AI workflows, AI copilots — whatever you want to call them.
The net-new spend in B2B is overwhelmingly flowing to AI. And if you’re a system of record that doesn’t own the AI layer on top of your platform, that spend is flowing around you, not through you.
You’re still getting your renewal. But you’re not getting the growth.
The AI Agent Layer Is the New Expansion Revenue
Think about how the best B2B companies grew over the past decade. It was land-and-expand. You’d get in with 10 seats, then grow to 100, then 1,000. Net revenue retention of 130%, 140%, 150%+ was the hallmark of the best companies.
Where does that expansion come from now? It’s not more seats — headcount is flat or declining at most companies. It’s AI agents doing the work that new hires used to do.
If those AI agents live on your platform, if you own them, that expansion revenue is yours. If they don’t — if a third-party AI layer is sitting on top of your data, orchestrating workflows, and delivering the value — you’re just the database underneath. You’re infrastructure. And infrastructure gets commoditized.
A few examples of what I’m seeing:
- ServiceNow gets this. They’re building their own AI agents aggressively, and it’s showing up in their growth numbers. They’re not waiting for someone else to build the agent layer on top of IT workflows.
- Salesforce is pushing Agentforce hard. You can debate the execution, but the strategy is right — they know that if someone else owns the AI agent layer for sales and service, Salesforce becomes just a database.
- Workday is investing here too, but the question is whether they’re moving fast enough. Every quarter they don’t own the AI agent for HR workflows is a quarter where someone else might.
The companies that are treating AI agents as someone else’s problem? They’re the ones I’m most worried about.
“Low Churn” Is a Slow Death Without Growth
Let me put some numbers on this to make the point concrete.
If you’re a $500M ARR system of record with 95% gross retention but only 105% net revenue retention because you’re not capturing AI spend — you’re growing at 5% net from your base. That’s it. In a market where AI-native competitors are growing 3-4x year-over-year, 5% expansion from your install base is a recipe for irrelevance. Not today. But in 3-5 years.
Compare that to a system of record that owns its AI agent layer and is driving 130%+ NRR because every customer is buying AI agents on the platform. That’s a completely different trajectory.
The gap between those two companies compounds fast.
What the Best Incumbents Are Doing Right Now
The systems of record that will win the next decade are doing three things:
They’re building AI agents natively, not partnering. Partnerships sound safe. “We’ll integrate with all the leading AI providers.” But that means someone else owns the value creation layer. The best are building their own agents, trained on their own data, deeply integrated into their own workflows. That’s where the moat is.
They’re pricing AI agents as expansion, not as features. If you just bundle AI into your existing SKU, you’re leaving all the growth on the table. The companies getting this right are creating new AI-agent SKUs that drive real expansion revenue — $50K, $100K, $200K+ in incremental ACV per customer.
They’re moving faster than feels comfortable. The AI agent that’s 80% good enough today is worth 10x more than the one that’s 99% good enough in 18 months. Because in 18 months, someone else will already own that workflow.
Low Churn Can Partially Mask The AI Threat. Don’t Let It. Own The Agents.
If you’re a B2B leader sitting on a system of record and feeling safe because your churn is low — I get it. That’s a real asset. But it’s a depreciating one if you’re not capturing the AI growth happening on top of your platform.
Your customers aren’t going to churn. But their spend is going to grow somewhere. The only question is whether it grows with you or with someone else.
Low churn without growth isn’t safety. It’s a slow fade.
Own the AI agents. Own the growth. Or watch someone else build the future on top of your data.
