How to Build an AI-Native Services Company

The speaker recommends markets where work is already outsourced, most tasks can be automated, and the work is complex enough to need AI paired with human expertise. Regulation can help create a moat; Panacea, for example, combines experienced FDA consultants with an AI platform. Tax, audit, insurance, mortgages, parts of health care, and logistics are suggested markets. Founders should also ask whether better models strengthen their service or make it easier to replace.

The speaker advises founding with people you know and have worked with. Strong teams combine credibility with buyers, fluency in what frontier models can do, and operational rigor around throughput, cycle times, and standard procedures. The General Legal Team is cited as combining law firm experience with technical leadership and using shift work to reduce client cycle times.

In AI services, the human is the customer's interface, while software helps that person handle more work. The speaker says to build around bottlenecks and track throughput and cycle time as product metrics. Inconsistent service outputs can quickly erode trust, so human involvement must scale less than revenue, and the software needs to work well for the people delivering the service.

Taking on too many pilot customers can overwhelm the team and leave the company relying on manual work instead of building scalable product. The speaker recommends starting with a small handful, selling outcomes rather than seats or tokens, and using pilots to find where AI creates meaningful leverage. Per-unit pricing is one option; outcome-based pricing can align incentives, while cost-plus and simple undercutting are discouraged.

Founders should track model, hosting, and human costs from the start, and be cautious about pilots with zero or negative margins. The long-term goal is for automation to lower costs as the product develops; the speaker describes a possible path toward 50% or higher margins, compared with about 30% for traditional services firms. Buying a services business is generally discouraged, except when it provides a needed regulatory moat such as insurance licensing.