How to Build an AI-Native Services Company
This transcript outlines a playbook for building AI-native services companies, where AI does the bulk of work to deliver outcomes in large markets like tax, law, and insurance. The speaker covers market selection, team formation, product building, sales strategy, and P&L structure. The core thesis is that these businesses can achieve software-like margins (50%+) on markets two to three times larger than traditional software.
Summary
The speaker opens by arguing that some of the biggest companies of the next decade will be AI-native services firms—not software companies—rebuilt from scratch in industries like tax, audit, insurance, law, and healthcare. Unlike co-pilot tools, these companies deliver outcomes directly to customers, with AI doing most of the work.
On market selection, the speaker identifies four key traits of ideal markets: low trust (work is already outsourced, so no behavior change is required), low judgment at the task level (most steps are automatable), a high intelligence threshold (the work must be hard enough to require models plus humans), and regulatory environments that can serve as moats rather than obstacles. Example markets include tax, audit, insurance, mortgages, and parts of healthcare and logistics.
For founding teams, the speaker emphasizes three critical attributes: domain fluency (direct experience is ideal but learned is acceptable), model fluency (deep understanding of what frontier models can do today and tomorrow), and operational rigor (comfort with metrics like variance, throughput, and cycle times). General Legal is cited as an example of a team combining law firm experience with technical leadership and operational thinking like shift work to reduce cycle times.
On product building, the speaker stresses that in AI services, humans are the interface to the customer—not the software. The product's job is to help humans scale nonlinearly. Key priorities include applying an operations mindset, obsessing over variance (inconsistent outputs destroy customer trust and cause churn), and ensuring that revenue does not scale one-to-one with headcount.
For sales and customer success, the speaker warns against the 'early demand trap'—signing too many pilot customers before the product can handle them, leading to a reliance on humans rather than scalable systems. Pricing should focus on outcomes (per-unit or outcome-based) rather than cost-plus or simple undercutting, as the competition is the cost of labor, not other software.
The P&L section walks through revenue, COGS (model costs, hosting, and humans in the loop), and operating expenses. The speaker introduces the concept of 'AI operating leverage'—as the product matures, COGS should drop and gross margins should rise toward software levels (50%+), but on much larger total addressable markets than pure software plays.
Finally, the speaker advises against acquiring existing services businesses to shortcut market entry, arguing that product-market fit cannot be bought and that legacy businesses carry misaligned expectations around metrics, hiring, and performance.
Key Insights
- The speaker argues that the best markets for AI services are ones where work is already outsourced and customers care only about the final outcome—meaning founders are displacing a vendor rather than asking customers to change behavior, which is a critical advantage because the budget already exists.
- The speaker contends that variance—non-uniform outputs from the service—is the existential threat to AI services businesses, claiming customers will fire a provider for inconsistency faster than for being slower or more expensive than incumbents.
- The speaker introduces the 'Sam Altman test' as a framework: founders should ask whether their service gets stronger as models improve, or whether better models commoditize them—arguing that only businesses in the first camp are safe long-term.
- The speaker describes the 'early demand trap,' where signing too many pilot customers at launch overwhelms a team's ability to serve them, locking the company into using humans rather than building scalable product systems.
- The speaker argues that the P&L opportunity in AI services is that traditional services firms cap out around 30% margins, but AI operating leverage—where COGS drop as the product matures—can push margins toward 50%+ on markets two to three times larger than what pure software companies typically address.
Topics
Full transcript available for MurmurCast members
Sign Up to Access