The Sentence That Changed How I Think About Software

“Most software has always served to organize our work, never to do our work.”

When I saw this line from Insight Partners in a Harvard alumni debate, I stopped and re-read it three times. Because it distills 30 years of software history into one sentence — and explains why everything is about to change.

For three decades, the industry built multi-billion-dollar empires selling organizational tools. CRMs that organize leads. ERPs that organize processes. Ticketing systems that organize requests. Software structured the flow — but final execution always depended on humans. Software organized. Humans executed.

And in the global market, whoever executes gets the biggest slice. Insight Partners showed the gap: in cybersecurity, labor/services spending is $350B vs $95B in software. In legal: $400B+ in billable hours vs $8B in software. In finance: $400B+ in team costs vs $19B in software.

Software always captured the smaller share. Until now.

The Trillion-Dollar Blind Spot

The convergence data is devastating. The global SaaS market is worth approximately $140 billion (2024). The global wages and services market is worth $11.7 trillion. When software starts doing the work instead of just organizing it, the TAM (Total Addressable Market) explodes 83x.

The software sector lost over $1 trillion in market value in 2026. The hardest hit? Exactly the companies whose core value was providing interfaces for human data entry — the kind of work AI agents are making dispensable.

Salesforce dropped 38% since the start of 2026. Adobe lost over $120 billion in value in seven weeks. ServiceNow and Workday faced double-digit declines. Their core value proposition — “a platform where humans enter data” — is becoming bypassable by agentic workflows.

The Business Model Inversion

AI agents are the first software products in history capable of crossing the line between organizing and executing. An agent doesn’t just structure task flows — it detects demand, analyzes context, prioritizes actions, and executes end to end.

The moment technology takes over execution, the classic “per-seat license” model loses meaning. This is where Outcome-Based Pricing is born.

Bessemer Venture Partners calls it “the AI pricing pivot.” Andreessen Horowitz describes it as “a shift toward outcome-based pricing.” And data confirms: a Pilot study showed per-seat pricing dropped from 21% to 15% of SaaS companies in just 12 months, while hybrid models surged from 27% to 41%.

40% of buyers already cite “seat reduction” as their primary lever to decrease software spending, according to BCG.

Who’s Already Doing It (With Numbers)

It’s not theory. Real companies already generate revenue with the new model:

Salesforce Agentforce: $800 million in ARR by end of fiscal 2026, 29,000 deals closed in Q4 alone. Uses “Agentic Work Units” (AWUs) — discrete units of work completed by agents. 2.4 billion AWUs delivered in Q4 alongside Flex Credits ($0.10 per agent action). The language shifted: “digital labor,” not “software.”

Zendesk AI Agents: charges $2 per automated resolution (pay-as-you-go) or ~$1.50 per resolution with volume commitment. Customer pays only when the problem is resolved without human intervention. Introduced a real-time automation dashboard for customers to verify results.

Intercom Fin: reached nine-figure revenue charging $0.99 per resolved support ticket.

These aren’t experiments. They’re real revenue lines built on a simple insight: when AI agents do measurable work, customers pay for the output.

Fortune Business Insights estimates the agentic AI market at $9.1 billion in 2026, projected to reach $139 billion by 2034 (40.5% CAGR).

The Customer Never Wanted the Hammer

The transition resurrects an old marketing maxim: the customer never wanted to buy a hammer. They always wanted the nail perfectly hung on the wall.

For decades, the industry sold increasingly better, lighter, prettier hammers. But the customer still bore the cost of hiring someone to swing it. AI is the first technology that delivers the nail on the wall directly.

You don’t pay for CRM software — you pay for qualified leads generated by AI. You don’t pay for legal software — you pay for contracts reviewed and validated by the system. You don’t pay for a ticketing system — you pay for problem resolution without human intervention.

As a16z puts it bluntly: “Software is becoming labor. And labor has never been priced by the seat. Labor is priced by output.”

What I Really Think

The direction is irreversible. Charging for results when agents do the work is logically inescapable. The per-seat model was designed for tools used by humans. When the human leaves the execution loop, the seat loses meaning.

But it won’t be binary. As Monetizely’s 2026 guide notes: “The likely end state is not ‘outcomes or bust,’ but a continuum of models.” Some companies will be “outcome-delivery partners.” Others will package automation as traditional subscriptions because their market isn’t ready to transact on outcomes.

The risk is real. If AI fails (hallucinations, context errors, all the limitations I’ve documented on this blog), who pays? In the per-seat model, risk falls on the customer. In the outcome model, risk falls on the vendor. This fundamentally changes incentives — and may be good for quality.

Professionals need to reposition. If software does the work humans did, humans need to move up the value chain. Not execute — direct, supervise, judge. It’s the “designer-director” I described in the Claude Design post. The “pain seeker” from the invisible skill post. The “architect” vs “executor” from early in this series.

Conclusion: Which Side Are You On?

The market is migrating toward whoever sells the final result. AI is here to extract value from the largest market that has ever existed: the labor and services market — $11.7 trillion that software never managed to capture.

The question for leaders and entrepreneurs is direct: are the tools your organization uses or sells still selling hammers, or are they ready to charge for the nail on the wall?

Share if this shifted your perspective:

SaaS is worth $140B. Labor is worth $11.7T. When software does the work, the TAM explodes 83x. The per-seat model can’t survive that math.

Read Also