Services: The New Software
The next $1T company won't sell AI tools—it will sell completed work, capturing the services budget (6x larger than software) by automating intelligence-heavy tasks that companies already outsource.
Read OriginalMy Notes (4)
For every dollar spent on software, six are spent on services.
The total addressable market for autopilots is all labour spend in a category, insourced and outsourced combined. But the right place to start is where outsourcing already exists.
If a task is already outsourced, it tells you three things. One, the company has accepted that this work can be done externally. Two, there’s an existing budget line that can be substituted cleanly. Three, the buyer is already purchasing an outcome. Replacing an outsourcing contract with an AI-native services provider is a vendor swap. Replacing headcount is a reorg.
Today’s judgement will become tomorrow’s intelligence.
Copilots and Autopilots
A copilot sells the tool. An autopilot sells the work.
Until recently, AI models were still developing intelligence and judgement, so the right approach was to build a copilot first: put AI in the hands of a professional and let them decide what to do with it. Harvey sells to law firms. Rogo sells to investment banks. The professional is the customer, the tool makes them more productive, and they take responsibility for the output.
Today, the models are intelligent enough that in some categories the best place to start is as an autopilot. Crosby sells to the company that needs an NDA drafted, not to outside counsel. WithCoverage sells to the CFO who needs insurance, not to the broker. The customer is buying the outcome directly. The work budget in any profession dwarfs the tool budget, and autopilots capture the work budget from day one.
The higher the intelligence ratio in any field, the sooner autopilots will win.
Intelligence vs Judgement
- Writing code is mostly intelligence. Knowing what to build next is judgement.
- Translating a spec into code, testing, debugging: the rules are complex but they are rules. Judgement is different. It requires experience and taste, instinct built on years of practice. Deciding which feature to build next, whether to take on tech debt, when to ship before it’s ready.
Summary used for search
TLDR
• AI companies face a choice: sell the tool (copilot) and race against model improvements, or sell the work (autopilot) and benefit from every model upgrade that makes your service faster and cheaper
• "Intelligence work" (rule-based, like coding or medical billing) is being automated first; "judgement work" (experience-based, like deciding what to build) stays human—but today's judgement becomes tomorrow's intelligence as AI systems accumulate proprietary data
• The autopilot playbook: start where outsourcing already exists (proven budget, outcome-based buying, no reorg), then expand toward insourced work as AI compounds—replacing an outsourcing contract is a vendor swap, replacing headcount is a reorg
• Biggest opportunities ranked by intelligence ratio and outsourcing prevalence: insurance brokerage ($140-200B), accounting ($50-80B), healthcare revenue cycle ($50-80B), claims adjusting ($50-80B), with specific companies named in each vertical
• Existing copilot companies face innovator's dilemma transitioning to autopilots (means cutting out their own customers), creating opening for pure-play autopilot startups
In Detail
The fundamental insight is that AI companies should stop competing on tools and start competing on outcomes. When you sell QuickBooks, every model improvement threatens to make your product a feature. When you sell "your books are closed," every model improvement makes your service better. The services budget in any profession dwarfs the tool budget—companies spend $10K/year on QuickBooks but $120K on the accountant who actually closes the books.
The piece introduces a framework for evaluating which professions AI will automate first: the intelligence-to-judgement ratio. Intelligence work follows complex but deterministic rules (writing code from a spec, translating clinical notes into ICD-10 codes, drafting NDAs). Judgement work requires experience and taste built over years (deciding which feature to build next, assessing culture fit in hiring, strategic consulting). Software engineering got automated first because it's primarily intelligence work—Cursor usage has shifted from autocomplete to agent-initiated tasks. Every other profession is coming, starting with the highest intelligence ratios.
The tactical playbook is to start where outsourcing already exists. If a task is already outsourced, three things are true: the company accepts external execution, there's an existing budget line to substitute, and the buyer purchases outcomes not tools. The article maps 10+ verticals by intelligence ratio and outsourcing prevalence, with TAM estimates: insurance brokerage ($140-200B) is highly standardized shopping across carriers; accounting ($50-80B) faces a structural shortage with 75% of CPAs nearing retirement; healthcare revenue cycle ($50-80B) is pure rules-based medical coding; claims adjusting ($50-80B) interprets policy language against damage schedules. Each vertical has named companies executing the autopilot strategy (WithCoverage and Harper for insurance, Rillet and Basis for accounting, Anterior for healthcare RCM, Pace and Strala for claims). The wedge is the outsourced intelligence work; the long-term TAM is the insourced judgement work that becomes automatable as AI systems accumulate proprietary data about what good judgement looks like in each domain.