The $6-to-$1 Ratio: Why AI Is Turning Software Companies Into Services Firms

For every dollar a company spends on software, it spends six dollars on services to make that software work. Implementation consultants. System integrators. Managed service providers. Outsourced operations teams. The ratio has held for decades because software has always been a tool, not an outcome. Someone still needs to wield it.

Julien Bek of Sequoia Capital published two essays in early March 2026 arguing that AI collapses this distinction. His thesis: the next trillion-dollar company will be a software company masquerading as a services firm. It will sell outcomes, not seats. And it will deliver those outcomes at software margins because its workforce is largely artificial.

The argument is subtler than “AI replaces software.” AI eats the $6, not the $1.

Copilots Sell Tools. Autopilots Sell Work.

Bek draws a clean line between the current wave of AI products and what comes next. A copilot assists a human doing a task. An autopilot does the task. The business model implications are different in kind, not degree.

A copilot is a feature. It helps an accountant work faster. The customer still needs the accountant. The software vendor sells a subscription. The accountant’s employer pays the salary. The $6 of services remains intact.

An autopilot is a service. It does the accounting. The customer pays for completed tax returns, not for access to a tool. The vendor captures revenue that previously went to the accountant’s employer. The $6 begins migrating.

Bek’s framing matters because it clarifies what “AI disruption” means for specific industries. Software engineering accounts for 50% or more of all AI tool usage. Every other profession is in single digits. That concentration tells you where the copilot model has traction. The autopilot model, the one that actually threatens the $6, is barely started.

The Vendor Swap Versus the Reorg

Here is the sharpest observation in Bek’s analysis: “Replacing an outsourcing contract with an AI-native services provider is a vendor swap. Replacing headcount is a reorg.”

Vendor swaps are procurement decisions. A CTO can approve them. They go through standard evaluation, RFP, negotiation, and contract. The switching cost is moderate. The organizational disruption is minimal.

Reorgs are existential decisions. They require board involvement. They produce legal exposure. They destroy institutional knowledge. They take quarters to execute and years to recover from. As we explored in The AI Workforce Reckoning, companies that frame AI-driven workforce changes as simple efficiency moves underestimate the governance complexity involved.

This distinction explains why the $6 will shift unevenly. The outsourced portion of services spending (IT managed services, business process outsourcing, staffing augmentation) will move first. That is a vendor swap. The in-house portion (internal teams performing services-like work) will move last. That is a reorg.

The TAM for the vendor-swap category alone is enormous. Bek estimates insurance brokerage at $140-200 billion, accounting and audit at $50-80 billion, healthcare revenue cycle at $50-80 billion, IT managed services at $100 billion or more, and management consulting at $300-400 billion. These are not software markets. They are services markets. AI-native firms that can deliver the same outcomes with ten employees instead of ten thousand will capture them at software-like margins.

The Demographic Accelerant

One data point in Bek’s analysis deserves separate attention. The United States lost 340,000 accountants over five years. Seventy-five percent of CPAs are nearing retirement. The profession is not being disrupted by AI. It is collapsing under its own demographics, and AI is arriving just in time to fill the vacuum.

Tax work, Bek notes, is 80-90% intelligence component. Gathering documents, interpreting rules, applying them to specific situations, filing forms. The remaining 10-20% is judgment calls and client relationship. An autopilot handles the 80%. A small team of senior professionals handles the 20%. The result is a firm that does more tax work with fewer people, not because it chose to cut headcount but because the headcount was never going to be available.

This pattern repeats across every profession facing demographic decline. The question is not whether AI will replace workers. In many fields, the workers are already leaving. The question is whether AI-native services firms can absorb the demand that departing workers leave behind.

Meanwhile, Software Is Shrinking

While Bek argues that AI expands what services firms can capture, Tomasz Tunguz of Theory Ventures documents what is happening to the software companies that sit on the other side of the ratio.

Net dollar retention (NDR) measures whether existing customers spend more or less over time. It is the single most important health metric for a SaaS business. In 2022, median NDR for public SaaS companies was 125%. By 2025, it had dropped to 112%. The 25th percentile hit 101% in a single quarter. At 101%, a company is barely growing from its existing base. Below 100%, it is shrinking.

The companies at the bottom of that distribution are recognizable names. Zoom at 98%. Asana at 96%. Bill.com at 94%. These are not failing companies. They are companies whose customers are spending less on their products each renewal cycle.

Tunguz frames this with a provocation: “If Microsoft can lose share in six months, no one is safe.” He is referring to GitHub Copilot’s trajectory. Copilot installs peaked, then declined. Claude Code surpassed it, reaching $2.5 billion in annualized revenue. Claude Code and Codex together blew past 100,000 combined installations. In the AI tooling market, the product that dominated six months ago is already losing ground.

The implication for the $6-to-$1 ratio: if the $1 of software is becoming commoditized and churning faster, the $6 of services becomes even more valuable by comparison. Outcomes are stickier than tools.

The Efficiency Numbers Are Real

Laurie Voss published a data analysis in March 2026 examining how AI startups compare to traditional software companies on operational efficiency. The numbers are striking.

AI startups operate with teams roughly 40% smaller than their traditional counterparts at the same revenue stage. At $10 million in annual recurring revenue, a traditional SaaS company employs 50 to 70 people. An AI startup employs 15 to 20.

Revenue per employee tells the same story from a different angle. AI startups generate $3.48 million per employee compared to $580,000 for traditional companies. That is a 6x multiple.

Series A valuations reflect the market’s belief in this efficiency. AI startups raise at $51.9 million average, compared to $39.9 million for traditional companies. Investors are paying a premium for the operating leverage.

In 2025, AI captured approximately $211 billion of the $425 billion in total venture capital raised. Half of all VC dollars went to AI. The capital markets have already placed their bet on which side of the efficiency divide will win.

These numbers describe the companies building AI-native services. They are the proof point for Bek’s thesis. If you can deliver accounting outcomes with 15 people instead of 70, your margin structure looks like software even though your revenue model looks like services.

The Governance Question Nobody Is Asking

As we explored in The $500 Billion Question, the economics of AI create a paradox. Cheaper capabilities drive broader deployment, which expands the surface area for error, compliance exposure, and unaudited spending. The Jevons Paradox applied to AI tokens.

The same paradox applies to AI-native services firms. A 15-person company delivering accounting services to thousands of clients has extraordinary leverage. It also has extraordinary concentration risk. When the AI makes an error in a tax filing, one human is responsible for the output of what used to be a 70-person team. The error surface scales with the automation. The accountability surface does not.

Bek does not address governance in his analysis. Tunguz does not either. Voss focuses on efficiency metrics. None of them ask the question that matters most for the enterprises buying these services: who is accountable when the autopilot is wrong?

This is the missing piece. The $6-to-$1 ratio will shift. The demographics demand it. The economics enable it. The capital markets are funding it. But the shift will produce a new category of risk that neither the services firms nor their clients have frameworks to manage.

An outsourcing contract with a 10,000-person BPO firm has diffuse risk. Errors distribute across a large workforce with multiple layers of review. An autopilot contract with a 15-person AI-native firm has concentrated risk. The automation layer has no professional judgment. The human layer is too thin to catch everything. The contract probably still references SLAs designed for the old model.

What This Means

Three dynamics are converging.

First, the services market ($6 for every $1 of software) is becoming addressable by AI-native firms that can deliver outcomes at software margins. The vendor-swap portion of this market will move within 24 months. The reorg portion will take longer and cause more damage.

Second, the software market (the $1) is simultaneously compressing. NDR is declining across the SaaS industry. Switching costs are collapsing. Products that dominated six months ago are already losing share. The $1 is becoming less defensible at the exact moment the $6 is becoming more addressable.

Third, the firms capturing this shift operate at 6x the revenue-per-employee of traditional companies. They raise larger rounds. They scale faster. They need fewer people. The efficiency advantage compounds because every new AI capability makes the 15-person team more productive without requiring the 16th hire.

The trillion-dollar question is not whether this shift happens. The demographics, economics, and capital flows all point in the same direction. The question is whether the firms executing this shift build the governance infrastructure to operate at that leverage without catastrophic failure. A 15-person firm doing the work of 70 needs better controls than the 70-person firm ever did. Not worse. Not the same. Better.

The $6 is moving. The only open question is whether it moves into governed systems or ungoverned ones. The answer will determine whether AI-native services firms become the next great category of enterprise value creation, or the next great category of enterprise liability.

---

This analysis synthesizes Julien Bek’s Services: The New Software (Sequoia Capital, March 2026), Tomasz Tunguz’s The Sword of Damocles in Software (Theory Ventures, March 2026), and Laurie Voss’s AI Companies Need Fewer People (March 2026).

Victorino Group helps organizations build the governance layer that lets them capture AI-native efficiency without concentrated risk. Let’s talk.