Your newest competitor runs at $4.4M revenue per head

Somewhere in your category, a team of fifteen is building the AI-native version of your product. They ship weekly, price under you, and don’t carry your cost structure. This isn’t a scare story — it’s the observable 2026 baseline.

The new speed

The reference points are hard to argue with:

• One AI-native firm reached $100M ARR in 8 months.
• Lovable hit $200M ARR with 45 people — roughly $4.4M in revenue per employee, several times the efficiency of a traditional SaaS company.

Efficiency like that isn’t just impressive; it’s strategic. A company running at $4M+ per head can undercut your pricing, grow on word-of-mouth quality instead of paid spend, and still print better margins than you.

The squeeze from both sides

Incumbents are getting compressed from two directions at once:

Differentiation erosion. Foundation models replicate features fast. The workflow that took your team two years is now a capable v1 in a quarter. Buyers have noticed: most now expect AI capabilities built in, and treat their absence as a reason to look at new entrants. Nearly all buyers say they want vendors to clearly demonstrate AI features in the pitch itself.

Margin erosion. Bolting AI into your own product isn’t free. Inference costs are showing up in the aggregate data — median gross margins down 6+ points where AI features ship. You pay to keep up, and the payment comes out of the margin that funded your growth.

What still defends

An honest inventory — some classic moats are weaker, a few are stronger:

1. Workflow embedment. Deep integration into how a customer’s team actually operates remains the hardest thing to rip out. AI-native rivals can copy features; they can’t copy your position inside the customer’s process. This is also why retention infrastructure matters more now, not less — embedment is a moat only if you notice when it’s eroding.
2. Proprietary data loops. If your product gets measurably better with each customer’s usage — and that improvement is visible to the customer — replication gets harder every month.
3. Being the answer. Distribution is shifting to AI-mediated discovery, and models recommend brands they can verify across many trusted sources. A new entrant can clone your feature list in a quarter; they can’t clone three years of citations, reviews, and category presence. In an odd twist, brand authority — the softest asset on the balance sheet — is becoming machine-readable and defensible.

Hype, discounted

Some of the panic is overdone. Enterprises still move slowly. Incumbent customer bases are stickier than Twitter believes. And a subset of the $100M-in-months stories will unwind — the same data shows AI products with brutal churn underneath the growth.

But the pricing pressure is real, the expectation of built-in AI is real, and the efficiency gap is real. The correct posture isn’t fear or dismissal. It’s an audit: which of your moats are feature-shaped (weak), and which are position-shaped — embedded workflows, data loops, and being the name the answer engines give (strong). Invest accordingly, before someone with 45 people does it for you.