Services as Software: Why AI Companies Are Selling the Work, Not the Tool

For every dollar a company spends on software, it spends roughly six dollars on services. AI companies selling tools are playing in a $600B market. AI companies selling the work are playing in a $6 trillion one.

That gap is now the central fault line in AI.

The most compelling companies being built right now don't make professionals faster. They replace the function. They don't help lawyers draft contracts -- they draft the contract. They don't give analysts dashboards -- they deliver the analysis. The outcome is the product. The professional is optional.

Sequoia's framing is worth reading in full: the next trillion-dollar company will be a software company masquerading as a services firm, powered by AI that sells outcomes instead of seats. Their list of addressable markets -- management consulting ($300-400B), recruitment ($200B+), insurance brokerage ($140-200B), accounting ($50-80B), healthcare revenue cycle ($50-80B), legal ($20-25B) -- reads like a menu of industries about to get restructured.

Outsourcing Is the Wedge

If a task is already outsourced, three things are true: the company accepts external parties can do the work, a budget exists, and the buyer is already paying for outcomes. Replacing an outsourcing vendor with an AI-native operator is a procurement decision. Replacing headcount is a reorg. The sales cycle difference is enormous.

This is happening across verticals simultaneously:

Insurance brokerage: Standard commercial lines are intelligence work on repeat -- shopping carriers, filling forms, matching risk profiles. WithCoverage and others are selling directly to the CFO, not the broker.

Accounting: The US has lost ~340,000 accountants in five years. 75% of CPAs are nearing retirement. Rillet isn't assisting accountants -- it's performing the function.

Healthcare revenue cycle: Medical coding translates clinical notes into ~70,000 ICD-10 codes. Complex, but rule-based. Anterior is building the autopilot.

Legal transactional work: NDAs, contract drafting, regulatory filings. Harvey moved from copilot to autopilot. Crosby started as an autopilot from day one.

But the Interesting Version Isn't Horizontal

The Sequoia piece is excellent on the what. The more interesting question is the where -- specifically, what happens when vertical AI absorbs services in domains that are too complex, too regulated, or too operationally embedded for generic horizontal players.

Two examples from the ADIN network:

Collide is building AI infrastructure for oil and gas operations. Not a chatbot. Not a copilot for petroleum engineers. An operational layer that indexes all of a company's field data, automates regulatory filings, workover procedures, and compliance workflows, and delivers answers to field operators in seconds. This is AI replacing the outsourced engineering services firms that energy companies have relied on for decades -- except it runs 24/7 and improves with every engagement.

Politheon deploys specialized AI agents across all 50 states and Congress to monitor legislation, score regulatory risk, and generate decision-ready intelligence briefs. This isn't a tool for government affairs teams. It's a replacement for the retainer you pay a DC lobbying firm to tell you what's happening. Politheon scans over 2 million legislative documents in real time, assigns risk scores, and synthesizes strategic context with cited sources -- the kind of analysis that used to take a senior analyst a full day, delivered in seconds.

Both are vertical. Both are selling the work, not the tool. And both are attacking markets where the incumbents are services firms, not software companies.

The Bezos Signal

The loudest confirmation that this thesis has gone mainstream: Jeff Bezos is raising $100 billion to acquire legacy manufacturing businesses and rebuild them with AI and robotics.

Not selling software to manufacturers. Buying the manufacturers.

If AI can perform the intelligence work inside a business, the highest-value move is to absorb the business and capture the margin directly. The AI doesn't sit on top of the operation. It becomes the operation.

Bezos isn't selling picks and shovels. He's buying the mine.

The Investment Frame

The work budget is 6x the tool budget. Moving from selling software to selling outcomes expands TAM by an order of magnitude.

Outsourcing is the entry point, ownership is the endgame. The fastest path to revenue is substituting outsourcing contracts. The long-term play -- as Bezos is demonstrating -- is absorbing the business itself.

The moat is operational, not technical. Model quality is converging. Distribution, domain expertise, regulatory compliance, and the willingness to take contractual responsibility for outcomes -- those are the durable advantages. The vertical players who accumulate proprietary operational data in energy, policy, healthcare, and legal will be nearly impossible to displace.

The question worth debating: We've spent a decade unbundling services into software. Are we now watching AI re-bundle software back into services -- but at software margins?