Why now: capabilities, workflow change and the ownership thesis
The argument for why owning service businesses became an attractive way to deploy AI, as made by the investors who fund the category, and what would have to be true for it to hold.
Three enablers
The clearest public statement of the technology argument is Joe Lonsdale's July 2025 essay [1], which builds on 8VC's October 2024 piece on [2]. Lonsdale identifies three practical capabilities that arrived together:
- Interpreting complex documents — contracts, returns, policies, tickets.
- Interacting with legacy systems — the practice-management, ticketing and core-banking software that service firms already run.
- Conducting voice or text conversations — with customers, tenants, patients and suppliers.
Together these let software execute sequences of work rather than single steps. That matters for ownership because an owner can redesign the sequence itself: who is responsible, where data flows, how exceptions are handled. A vendor can only sell into the sequence as it stands 1Attributed opinion.
The feedback mechanism
The most consequential part of the thesis is the loop between practitioners and the system. Experienced accountants, technicians or loan officers identify edge cases and improve procedures; that knowledge is embedded into the software; the software becomes harder to replace than the underlying model. Elad Gil makes a related argument that ownership accelerates organisational adoption because the owner sets incentives 1Attributed opinion. Sequence calls the result "refounding" and describes forward-deployed engineers who feed each deployment back into a shared platform 1Company reported.
The published operating evidence is consistent with the mechanism without yet proving the economics. The Thrive/OpenAI tax pilot describes returns meeting a 75%-correct-field threshold rising from 25% to 86% in six weeks as the system was tuned 1Company reported. Dwelly's maintenance update tracks human minutes per task over time 1Company reported. Both are participant reports.
Why ownership rather than selling software
Sponsors give three reasons, all of which are their positioning rather than established fact:
- Adoption. A tool that shifts work onto someone else stalls. Thrive's own [3] describes a past tax-software effort that failed because document scanning never became part of the administrative team's routine. Owners can change the routine.
- Economics. Margin expansion and growth inside an owned business accrue to the owner, whereas a software licence captures a fraction. General Catalyst frames the target as services businesses with 30–40% margins growing 10–20% — its "Rule of 60" ambition 1Forecast or target.
- Supply of businesses. A generation of owner-operators is retiring. Lonsdale's separate October 2025 essay on the [4] argues that succession pressure creates willing sellers at reasonable prices. Teamshares, Mainshares and similar platforms exist for the same reason without the AI thesis.
What would have to be true
The research turns the thesis into five testable propositions:
- Ownership improves the ability to change workflows.
- The acquirer learns something reusable from each deployment.
- The accumulated operating knowledge survives employee turnover and model changes.
- Saved effort becomes cash flow, better service or revenue capacity — not idle time.
- Competitors cannot immediately bid away the benefit.
Propositions 1 and 2 have the most supporting evidence, almost all company reported. Propositions 3 to 5 are largely untested in public. The economics chapter explains which measurements would settle them.
Research gaps
- Lonsdale's essay cites "10–20x productivity" and margins moving "from the teens into the 50s or 60s" without naming the companies, workflows, sample sizes or margin definition. These are preserved as part of the public thesis, not used as benchmarks.
- No source in the corpus provides a same-cohort, multi-period measurement with fully loaded engineering and inference cost.