AI has barely touched care delivery and drug discovery, and the next decade will show it
David George argues that AI's transformative phase in healthcare has not yet arrived. His 10-year forecast covers care delivery and drug discovery, two domains where structural change, not just early-stage activity, remains largely absent.
AI has swept through software, finance, and media with enough force that most observers treat its transformative phase as already arrived. David George disagrees, at least for two of the domains where transformation would matter most. In his view, care delivery and drug discovery have not yet been seriously touched, and the decade ahead is where the real movement happens.
George’s argument rests on a contrast between visible early activity and meaningful structural change. He acknowledges that some companies are working in both areas and showing early signs of progress. But that work, in his framing, does not amount to scratching the surface. The implication is that what has been deployed so far in healthcare and drug development is a rounding error relative to what the technology will eventually do. His 10-year window is precise enough to be testable: if care delivery and drug discovery look roughly as they do now in 2035, the call is wrong.
The “early innings” part of his argument is the less contested half. Early signs of progress exist, by George’s own account. What he disputes is whether those signs represent the beginning of structural change or simply the presence of the technology at the margins of two very large industries. His phrasing, that nothing has scratched the surface, is a strong claim: it treats current AI activity in these domains as effectively preliminary, regardless of how it has been described by the companies involved.
We've done nothing to scratch the surface either on care delivery or on drug discovery yet. I mean, there's some companies that are working on it, showing some early signs of progress, but I think the progress that we make there in the next 10 years is going to be massive. David George
Care delivery is the less-discussed half of George’s call, and in some ways the harder one to assess. Drug discovery has clear output metrics: molecules advanced, trials initiated, approvals granted. Care delivery encompasses how patients encounter the system, how clinicians document and decide, how therapies reach the people who need them. Progress there is harder to measure and harder to attribute cleanly to any single technology. George treats it as equally ripe, which implies he sees structural opportunity in the delivery layer, not only in the laboratory.
What makes the call worth taking seriously is the asymmetry it identifies. Most public attention on AI in healthcare has concentrated on imaging, diagnostics, and administrative automation. Those applications are real and in production. George is pointing past them to the parts of the system where AI has not yet changed the basic economics or workflows. The question his forecast raises is whether the same forces that compressed research timelines in other fields will eventually reach care delivery and drug discovery on a similar curve, or whether those domains have friction that makes them genuinely different.
The honest answer is that no current evidence resolves it. George is making a directional bet on rate of change over a long horizon, not reporting a measurement. The companies showing early signs of progress he mentions could represent the leading edge of a wave or could remain isolated experiments. A reader who wants to track the call has a benchmark: not whether AI is present in these domains by 2035, but whether its presence has become massive in George’s sense, meaning structurally consequential rather than supplementary.
The 10-year frame matters because it is long enough for drug discovery timelines to play out and short enough that the people making decisions today will be around to account for the results. If George is right, the implications for patients and for the economics of pharmaceutical development are substantial. If he is wrong, the question is whether the friction in healthcare proved more durable than the technology’s advocates anticipated. Either outcome would tell us something important about where AI’s limits actually are.