I was sitting in a conference room in Hong Kong last week, listening to a procurement director from a list company describe his company's new software buying protocol. He said something that made my coffee go cold:
"Last year we bought fifty seats of enterprise software. This year? We bought two. Then we connected those two accounts to our internal AI agent. The agent handles what used to take fifty people."
He wasn't bragging. He was warning me.
The per-seat SaaS model—the foundation of the last two decades of software economics—is being dismembered in real time. While you're counting your ARR and planning your next "land and expand" strategy, your customers are quietly replacing your user licenses with autonomous agents that never sleep, never ask for raises, and never generate recurring revenue for you.
This isn't a trend. It's a structural collapse.
The Comfortable Death
a16z partner David George said it plainly last month: the software industry's comfort zone is dead. Public markets have already repriced the sector. The old trick of "slow growth but pretty profits after we deduct stock-based compensation" doesn't work anymore. Investors see through it.
In the next twelve to eighteen months, any company stuck in the middle—growing too slowly for a high valuation, too diluted to be a cash cow—will get liquidated by the market. Not because they're bad companies, but because the physics of value creation has changed.
You have two options. They are extreme. They are mutually exclusive. And "doing a little of both" is a death sentence.
Path One: The Growth Insurgency
This isn't bolting a chatbot onto your legacy codebase. This is committing, in the next 12–18 months, to increase your total revenue growth rate by ten percentage points purely through AI-native products.
This requires organizational arson.
Find your five special forces: Every company has four or five people who create 100x value regardless of their current title. Find them. Give them the charter to "refound" the company. Not a task force—a reinvention.
Purge the bureaucratic immune system: Those five people need to own data architecture, workflow documentation, and CRM context. They need to build the dynamic AI layer that replaces middle management. Watch your VPs carefully this month. Anyone who asks "what's our AI strategy?" instead of showing you shipped code is gone. Replace them with the special forces.
Deploy four-person kill teams: Dedicate 50% of R&D to new AI products. No departments. No scrum-of-scrums. Just "design + product + engineering" micro-squads of four. Start writing code on day one. Constrain headcount, not compute. Communication costs must approach zero.
Shift the monetization physics: If your AI product doesn't allow an agent to autonomously consume and pay for itself, you've failed. Future growth comes from token consumption, usage intensity, automation depth, and outcomes delivered—not from selling seats to humans who might get fired next quarter.
Path Two: The Profit Machine
Not everyone can pull off Path One. If you can't engineer 10-point growth through AI-native products, your only alternative is to become a ruthless cash-generating engine: achieve 40–50% true operating profit (including stock-based compensation) within 12–24 months.
This isn't "trimming 8%" for the headlines. This is redesigning the machine:
Flatten and standardize: Cut middle management. Abolish committees. Reject high-customization professional services. If a customer needs bespoke consulting, they don't fit your economics.
Rationalize the customer base: Raise prices on core workflows with high switching costs. Force low-margin long-tail customers onto high-margin base pricing—or let them churn. Revenue quality beats revenue quantity.
Give your engineers weapons-grade compute: You'll be cutting headcount aggressively (both ICs and executives), but the engineers who remain get $1,000+ monthly AI token budgets. If they don't burn through it, they're not working hard enough. One top-tier engineer with twenty AI agents outproduces a ten-person committee by two orders of magnitude.
Accept moat evaporation: Your traditional UI advantages and integration moats are being dissolved by AI interfaces. Prepare for your core business to become a commodity and price accordingly.
Study Broadcom under Hock Tan: Brutal cost discipline, product simplification, and absolute profit extraction. Become a weapon, not a platform.
What We're Doing at Mercury
We saw the death of per-seat pricing early back in 2021. We refused to die in the middle.
So we fused Path Two's profit discipline with Path One's business model architecture. We rebuilt our pricing and operational philosophy from the substrate up:
Outcome-Based Economics, Not Headcount Taxes
We eliminated per-user billing entirely. Our contracts now run on "Compute & Outcome."
- Infrastructure retainer: Clients pay a fixed monthly base for infrastructure and AI token pools (like a minimum utility bill).
- Value extraction: When our AI processes 10,000 orders autonomously or saves 5,000 hours of support labor, we take a percentage of that quantified business outcome.
The perverse incentive is gone. When our clients downsize their human teams and scale their AI dependency, our revenue grows exponentially. We don't fight their HR department; we become equity holders in their operational efficiency.
The Profit-First Squad Structure
We maintain 60%+ true operating profit (including all options and bonuses). No bloated middle management. No "territory" VPs. Just customer-facing PMs and engineers with massive API budgets.
We don't touch low-customization professional services. We deliver standardized, modular AI infrastructure that agents consume, not humans configure OR high value customization professional service.
The Choice
Software used to be a safe business. You built a tool, sold seats, collected recurring revenue, and raised prices 5% annually until you sold to Salesforce or went public. That arithmetic is now a memento mori.
You have eighteen months. Either build the AI-native product that grows your revenue curve by ten points, or transform into a profit machine that extracts 40% margins while your competitors burn.
"We'll do a little of both" is what your tombstone will say in 2027.
Pick a path (or both). Start today.
— James, CEO, Mercury Technology SolutionsHong Kong, March 31, 2026
