aiGalen Guan

Deep Dive: Sequoia Capital Trillion-Dollar Quadrant — Why the Next $1T AI Company Won't Sell Software

On March 5, 2026, Sequoia Capital partner Julien Bek published a ~1,400-word article on the firm's website titled "Services: The New Software."

He probably didn't expect it to go viral.

Within days, the article surpassed 1 million views on X, now closing in on 3 million. Over 450K impressions on LinkedIn. Fortune magazine conducted a deep interview. Almost every top-tier VC in Silicon Valley shared and discussed it. This wasn't just a blog post — it was a declaration of investment consensus and a public deal-sourcing manifesto.

The core thesis is captured in a single sentence:

"The next $1T company will be a software company masquerading as a services firm."

In this article, I'll deeply deconstruct Bek's complete analytical framework — the Intelligence vs. Judgement spectrum, the Copilot vs. Autopilot path divergence, the outsourcing wedge as a go-to-market strategy, and the iconic "Trillion-Dollar Quadrant" opportunity map — while also presenting critical counter-arguments from Linas Beliūnas, Deepak Jha, and others.

1. The $1 vs. $6 Economics

Bek's entire argument rests on a deceptively simple number that is often overlooked:

For every $1 spent on software, $6 is spent on services.

The global software market is approximately $700 billion. The global professional services market exceeds $6 trillion. The past two decades of SaaS — from Salesforce to Snowflake — essentially captured value from the $1 side. The AI era opportunity, Bek argues, is to capture the $6 side.

This isn't just about TAM size. It's a fundamental paradigm shift in business models. SaaS companies sell tool seats ($10K/yr per seat). Services firms sell labor outcomes ($120K/yr for an accountant to close the books).

His vivid example:

A company might spend $10K a year for QuickBooks and $120K on an accountant to close the books. The next legendary company will just close the books.

This isn't about building better software. It's about replacing the entire service delivery layer — accountants, insurance brokers, claims adjusters, coders, recruiters — with AI-native firms that look like services companies on the outside but run on software economics internally.

2. Intelligence vs. Judgement: AI Has Crossed the Threshold

Bek draws a sharp line dividing professional work into two categories:

Intelligence work: Complex but rule-based. Translating specs into code, testing, debugging, translating clinical notes into ICD-10 codes, shopping insurance quotes across carriers. "The rules are complex, but they are rules."

Judgement work: Requires experience and taste. Deciding which feature to build next, whether to take on tech debt, when to ship before it's ready, strategic recommendations, culture-fit assessments in hiring. "It requires experience and taste, instinct built on years of practice."

The evidence he provides is specific and compelling: software engineering accounts for over half of all AI tool usage across professions. Every other category is still in single digits. In Cursor, more tasks are now started by agents than by humans.

Why? Because software engineering is primarily intelligence work. AI has crossed the threshold where it can do most intelligence work autonomously and leave the judgement to humans.

This creates a natural sequencing of industry disruption:

  1. Software engineering (already happening)
  2. Accounting/tax/audit (high intelligence ratio + labor shortage)
  3. Insurance brokerage/claims (highly standardized processes)
  4. Healthcare revenue cycle (pure rule-following)
  5. IT managed services (repeatable across identical environments)
  6. Legal transactional work (standardized outputs)
  7. Recruitment (top-of-funnel is intelligence, closing is judgement)
  8. Management consulting (mostly judgement — last to fall)

3. Copilot vs. Autopilot: The Defining Choice

This is the most actionable part of Bek's framework.

A copilot sells the tool. An autopilot sells the work.

Dimension Copilot Autopilot
Customer The professional (lawyer, accountant, broker) The company that needs the outcome
Revenue source Tool budget (~$10K/yr) Work budget (~$120K/yr)
Model risk Every model improvement threatens the tool Every model improvement strengthens the service
Responsibility Professional takes responsibility for output Autopilot company takes responsibility
Pricing Per-seat SaaS Per-outcome (per claim, per filing, per hire)

Harvey sells to law firms. Rogo sells to investment banks. These are copilots.

Crosby sells NDA drafting directly to the company. WithCoverage sells insurance directly to the CFO. These are autopilots.

The critical insight: the work budget in any profession dwarfs the tool budget, and autopilots capture the work budget from day one.

But there's a structural challenge: the innovator's dilemma for copilot companies. They have product, customer knowledge, and distribution. But transitioning to autopilot means cutting their own customers — the professionals — out of doing the work. This creates an opening for pure-play autopilots.

4. The Outsourcing Wedge: Autopilot Go-to-Market

Bek's playbook is refreshingly practical:

  1. Start with outsourced, intelligence-heavy tasks (the wedge)
  2. Nail distribution in that niche
  3. Expand toward insourced, judgement-heavy work as AI compounds

If a task is already outsourced, it signals three things:

  • The company has accepted this work can be done externally
  • There's an existing budget line that can be substituted cleanly
  • The buyer is already purchasing an outcome

Replacing an outsourcing contract = a vendor swap (easy). Replacing headcount = an organizational reorg (hard).

Crosby starting with NDAs is the textbook case: well-defined, primarily intelligence work, already commonly outsourced to external counsel. Budget exists, scope is clear, ROI is immediate, substitution is frictionless.

5. The Trillion-Dollar Quadrant: The 10-Vertical Opportunity Map

This is the most discussed element of Bek's article — the matrix that plots every services vertical along an Intelligence-to-Judgement spectrum (Y-axis) and Outsourced-to-Insourced ratio (X-axis).

The upper-left quadrant (Outsourced + Intelligence-heavy) = the ripest territory for autopilots.

The 10 Verticals (US TAM, outsourced portion):

Insurance Brokerage ($140-200B) — Largest dollar market. Pure intelligence work. Fragmented distribution. WithCoverage, Harper.

Accounting & Audit ($50-80B) — US lost ~340K accountants in 5 years. 75% of CPAs nearing retirement. Rillet, Basis.

Healthcare Revenue Cycle ($50-80B) — Medical coding = translating notes into 70K ICD-10 codes. Pure rules. Anterior.

IT Managed Services ($100B+) — Patches, monitoring, provisioning on repeat. Nobody yet sells "your IT runs" as an outcome. Edra (Sequoia invested), Serval.

Claims Adjusting ($50-80B) — Adjuster workforce aging out. Pace, Strala.

Tax Advisory ($30-35B) — 80-90% intelligence. Multi-jurisdiction complexity deepens data moat. TaxGPT, Skalar, Ravical.

Legal Transactional ($20-25B) — NDAs, contract drafting, regulatory filings. Harvey (copilot→autopilot), Crosby, Lawhive.

Supply Chain & Procurement ($200B+) — 80% of suppliers get zero attention. The wedge is "abandoned work." Magentic, AskLio, Tacto.

Recruitment & Staffing ($200B+) — Largest services market. Top-of-funnel is intelligence; closing is judgement. Wedge: high-volume, low-judgement roles. Juicebox, Mercor, Jack & Jill.

Management Consulting ($300-400B) — Largest TAM but hardest to automate. Mostly judgement. Can consulting be disaggregated into intelligence (data gathering) and judgement (strategy)?

Total TAM across 10 verticals: ~$1.4-1.7T (US alone).

6. Cold Water: The Critics

Bek's argument is elegant, but not bulletproof.

The $0.03 Problem (Linas Beliūnas)

"For every $1 companies stop spending on humans, they spend $0.03 on AI."

This is the sharpest critique. Bek's trillion-dollar framing requires labor budgets to redirect to AI vendors at comparable prices. They don't. They evaporate. When machines do the work, the work gets repriced at machine rates — 97% cheaper. A $200B insurance market doesn't mean $200B in AI revenue. TAM is a ceiling, not a floor.

Margin Compression

Autopilot gross margins are likely ~70% (per Bret Taylor / Sierra data), versus ~90% for pure SaaS. AI inference costs are real. Go-to-market for services remains unsolved. These economics look more like tech-enabled services than pure software.

Judgment Capital Management (Deepak Jha)

The deepest critique comes from a governance perspective. The critical risk isn't that autopilots will replace judgement — it's that autopilots absorb intelligence faster than organizations institutionalize judgement. By the time the gap is visible, the window has closed.

He draws an aviation analogy: when fly-by-wire replaced mechanical controls, pilots lost the haptic cues that warned them of a stall. Synthetic stick shakers were added — but if the next generation never felt a real stall, the stick shaker is just noise. Organizational parallel: as autopilots absorb intelligence-layer decisions, practitioners move from doing the work to supervising the output. The developmental friction through which judgement is built disappears.

The VC Consensus Problem

"The consensus is so complete that you could swap the logos on their published theses and most readers wouldn't notice." — Linas Beliūnas

When YC, a16z, Sequoia, and Bessemer all point in the same direction, founders should be nervous, not excited. Too much capital chasing the same opportunities. The real winners may be in overlooked corners that don't fit the consensus framework.

7. The Bigger Picture: Sequoia's AI Thesis Evolution

Bek's article didn't appear in isolation. It's the culmination of a coherent multi-year AI investment narrative:

  • 2023: "Generative AI's Act 1" by Sonya Huang and Pat Grady — the landscape is mapped
  • 2024: "Generative AI's Act o1" — reasoning/agentic capabilities emerge
  • May 2025: AI Ascent keynote — Konstantine Buhler introduces the Agent Economy; AI opportunity is "at least 10x larger than cloud computing"
  • Aug 2025: "The $10 Trillion AI Revolution" — AI > Industrial Revolution; 144 years compressed to 17
  • Jan 2026: "2026: This Is AGI" — models are now capable enough to handle intelligence work autonomously
  • Mar 2026: "Services: The New Software" — the applied investment strategy, with specific verticals, named companies, and an actionable playbook

8. My Takeaways

Julien Bek's framework went viral not because he discovered something new — the idea of software-ifying services and using outsourcing as a wedge aren't novel strategies. What made it break through was that he drew a clear line and produced an executable playbook.

But I think the most valuable insights aren't in the specific TAM numbers (those will shift, and the $0.03 problem is real). They're in two more fundamental frameworks:

First, the Copilot → Autopilot paradigm shift. If you're an AI founder and your product is currently a copilot, ask yourself honestly: does each base model improvement make your product stronger, or easier to commoditize? If your answer leans toward the latter, you're building a sand castle. Transitioning to autopilot is hard because of the innovator's dilemma — it will hurt.

Second, outsourcing = the wedge. "Replacing an outsourcing contract is a vendor swap. Replacing headcount is an organizational reorg." Every vertical AI founder should tape this to their wall. Your first customer shouldn't need organizational change to buy your product.

The broader perspective: we're witnessing not the evolution of AI tools, but the re-pricing and re-architecting of the entire services economy. When machines can do the same intelligence work at 3% of human cost, the very concept of "the value of services" will be redefined. This isn't a technology question. It's an economic structure question.

The next trillion-dollar company might indeed look more like Accenture than Salesforce.

Sources: