Interview, Fireside Chat
The Five Year Desert to Product Market Fit & a $5.3BN Valuation with Shiv Rao, Founder @ Abridge
- Vertical AI is projected to reach "table stakes" status for health systems within three to four years, driven by a "tidal wave of adoption" necessitated by a current system unprepared for a "tsunami of healthcare needs."
- In-house model adoption is expected to rise from approximately 40% to 60% or higher over a two to three year period as new open-source models are distilled, though the exact future mix with frontier models remains undefined.
- AI agents are forecast to increase token consumption by 24x within a five-year period, primarily automating "high frequency, low stakes" tasks like medication refills until approaching the "most boring variant of primary care."
- New business models aligning stakeholders around prevention rather than "sick care" are anticipated to emerge within the next three to five years, while "high stakes, high frequency" workflows requiring judgment and taste will remain human-led.
- To manage the compression of decision-to-action times at scale, the organizational structure must become "very very flat" with fewer managers, requiring high-judgment executives who can operate with a "players" culture even as the workforce grows to 450 employees.
- Companies are advised to own their destiny by controlling their P&L and reaching down into the technical stack, as building solely on the consumer side is viewed as a barrier to entering the enterprise market where trust is paramount.
- While inference costs are expected to decrease, total spending will rise due to increased usage, potentially offset by future costs for frontier models that may rise in the coming weeks and months.
- Founders are advised to focus on "tours of duty" to address specific challenges rather than micromanaging, with the goal of raising more capital and impacting more users before a potential public offering, which depends on reaching a specific milestone of impact.
- The industry is expected to see "AI doctors" and teams of background assistants deployed for each clinician to address the estimated need for 30 hours of daily work, utilizing spoken signals as a universal wedge for interaction.
- A key risk involves the market moving too slowly to justify the thesis, which could lead to failure, alongside the danger of companies being unable to achieve the necessary trust levels to transition from consumer to enterprise solutions.