The Real AI Agent War Has Three Fronts, Not Two

I stumbled on this piece from Jason Lemkin at SaaStr that crystallizes something every B2B founder above $50M ARR is feeling: you have two full-time jobs now. Keep your installed base happy. And win the AI agent war in your space.

Lemkin is right about the tension. He's right that customers want AI from existing vendors. And he's dead right that your own team will resist the AI buildout through a thousand small objections. But after reading it, I felt like he was describing the Fortune 500 version of the problem. The data from our platform tells a more nuanced story—one with three distinct battlefronts that founders at different scales need to see.

Battlefront 1: The Resourceful Enterprise (Where Lemkin's Model Works)

For companies with $50M+ ARR and hundreds of enterprise customers, Lemkin's framework applies directly. You have the resources to run two parallel businesses. Your challenge is organizational—keeping the CEO's attention split, preventing the installed base from consuming your roadmap, and protecting your AI investment from internal antibodies.

Our data reinforces this. We track 212 problems related to renewal pressure and vendor scrutiny, with an average severity of 4.1/5. Specific complaints like "We're evaluating whether to renew our $500k contract because the vendor hasn't delivered AI enhancements" are common. Enterprise customers are actively benchmarking incumbents against AI-native alternatives. Lemkin captures this dynamic well.

But here's where it gets interesting.

Battlefront 2: The SMB Death Spiral (Where the Real Pressure Lives)

Lemkin assumes resource abundance. What about the SMB and mid-market companies that don't have the luxury of running two parallel businesses? Our data tracks 98 problems specifically from SMBs struggling with AI adoption. The average severity is 4.3/5—higher than enterprise renewal pressure. The core complaint: "We can't afford to maintain our current software and also experiment with AI agents—we have to choose."

These are the customers most likely to churn to an AI-native competitor. Not because they want to, but because they have to. Their current tool is eating their budget, and the AI startup promises 80% functionality at 40% of the price. Even if that promise is half true, the math forces a switch.

For founders building for SMBs, the threat isn't that your installed base will resist AI—it's that they'll disappear before you get a chance to ship it. You don't have the margin to maintain legacy features while building an AI agent. You have to pick, and the wrong pick is fatal.

Battlefront 3: The Regulatory Moat (Where Incumbents Can Breathe)

The article completely misses the regulated industry angle. In healthcare, finance, and legal, AI adoption faces a different bottleneck: compliance. Our data captures 76 problems in regulated industries where the primary blocker is approval, not product. "Compliance teams block AI features for months" and "Our legal team requires manual review of every AI output" score an average severity of 4.5/5—the highest in our dataset.

This is a double-edged sword. If you're an incumbent selling into healthcare, your customers can't quickly switch to an AI-native startup—they need to validate security, privacy, and regulatory compliance first. That gives you time. But if your AI buildout is slow, you risk losing the window when regulations eventually catch up.

For investors and founders in regulated spaces, the playbook is different: your moat is compliance, not just product. Ship AI features that pass regulatory scrutiny first, even if they're less flashy. The AI-native startups will struggle here longer than you think.

The Hidden Reliability Gap

Lemkin cites the Redpoint survey showing customers prefer to buy AI agents from existing providers. Our data confirms this: we have 47 problems in the 'AI Integration' category where users wish their current vendor had better AI so they wouldn't have to evaluate new tools. Average severity 3.9/5.

But here's where I push back on Lemkin's framing. He warns that AI-native competitors "promise to do 80%-200% of what you do, sometimes at 40% of the price." That's frightening. But our data on 34 AI-native competitors reveals that 62% of these tools have hidden reliability issues—problems users discover after purchase, like data privacy gaps, poor integration depth, or flaky agent performance. The average severity of these hidden problems is 4.2/5.

Translation: Many AI agents look great in the demo but fall apart in production. That's a window incumbents can exploit. If you can ship an AI agent that works reliably within your existing platform—even if it's not the most innovative—you can neutralize the pure-plays. The bar isn't "best AI agent." It's "AI agent that doesn't break your workflows."

What This Means for Founders and Investors

Lemkin ends by saying the CEO must hold both jobs. True at scale. But for earlier-stage companies or those in specialized markets, the strategy needs to be more surgical.

If you're selling to SMBs, your risk is churn before you can build. Consider a dual-product strategy: keep your core product lean and ship a separate AI agent that can be sold alongside it, rather than rebuilding your existing product. Let customers buy AI on their own timeline.

If you're in a regulated industry, lean into compliance as a feature. Your customers will forgive slower AI development because they can't adopt fast anyway. Prioritize audits and security certifications over flashy demos.

If you're an investor evaluating a Series A pitch for an AI agent startup, ask about reliability. The demo may show 80% functionality, but our data suggests the remaining 20%—the messy integration work—is where most startups die. Look for teams that understand the hidden complexity of enterprise workflows.

Lemkin's core thesis is sound: the age of one-job CEOs is over. But the battlefronts aren't universal. Know which one you're fighting on.