Medable, a clinical trial platform, is using Google's AI stack to compress drug development timelines. Their goal: 12 years to one year, one day to enrol a patient, one day to activate a trial site. They call it the "1:1:1" vision.
Whether they hit those numbers doesn't matter as much as the fact they're trying. Clinical trials are the slowest, most expensive part of bringing new drugs to market. Anything that makes them faster makes medicine cheaper and more accessible.
What Actually Takes 12 Years
Drug development is slow because it's manual and sequential. You identify a compound. Test it in animals. Find patients who match inclusion criteria. Get ethics approval. Set up trial sites. Train staff. Collect data. Analyse results. Each step waits for the one before it.
The bottleneck isn't science - it's coordination. Matching patients to trials, getting sites approved, ensuring regulatory compliance, managing data across organisations. These are logistics problems dressed up as medical problems.
Medable's platform automates the logistics layer. They use Vertex AI Agent Builder to create AI agents that handle patient screening, site selection, and regulatory documentation. Gemini processes unstructured medical records to find eligible patients. Google Cloud handles data orchestration across trial sites, pharma companies, and regulators.
The result, according to their own case study, is trials that start in days instead of months and enrol patients in hours instead of weeks.
The One-Day Patient Enrollment
Patient recruitment is the longest part of most trials. You need people who match specific criteria - age, condition, medical history, no conflicting medications. Finding them means trawling through hospital records, GP databases, patient registries. Then contacting them, explaining the trial, getting consent.
Medable's system scans electronic health records using Gemini's multimodal capabilities. It reads doctors' notes, lab results, prescription histories - all the unstructured data that traditionally requires human review. Then it identifies candidates who meet the trial criteria and flags them for outreach.
The claim is this takes under 24 hours from query to patient contact. That's plausible if the records are digitised and accessible. It's not plausible if you're working across fragmented healthcare systems or countries with strict data protection laws. But for large hospital networks with unified EHR systems, the maths works.
The One-Day Site Activation
Activating a trial site - getting a hospital or clinic approved to run a study - involves regulatory paperwork, ethics committee reviews, staff training, and equipment setup. This normally takes weeks or months.
Medable's agents automate the paperwork. They generate protocol documents, consent forms, and regulatory submissions based on templates and trial-specific data. They track approval workflows and flag blockers in real-time.
The one-day claim assumes the ethics review is fast-tracked or pre-approved, which isn't always realistic. But even cutting site activation from 8 weeks to 2 weeks is significant. Trials with 50 sites save years of calendar time.
What This Means for Drug Costs
Clinical trials cost between £1 billion and £2 billion per drug. Most of that is time - staff salaries, site overhead, patient management. If you cut trial duration by 50%, you cut costs by roughly the same amount.
Cheaper trials mean cheaper drugs. Or at least, they should. Whether pharma companies pass savings on to patients or pocket them as profit is a different question. But faster trials definitely mean drugs reach patients sooner. For terminal illnesses, that's lives saved. For chronic conditions, that's years of suffering avoided.
The sceptical view: Medable's numbers are aspirational. 12 years to one year is a theoretical best case, not an industry average. Most trials will still take longer because of factors AI can't fix - unexpected side effects, slow patient enrollment despite automation, regulatory delays.
But even half the claimed improvement is transformative. A trial that takes 6 years instead of 12 is still a drug reaching patients twice as fast.
The Regulatory Question
Regulators don't move fast. They shouldn't. Drug approval requires careful review of safety data, and "move fast and break things" is a terrible philosophy when the things breaking are human bodies.
AI can speed up data collection and analysis. It can't speed up the human judgement required to decide whether a drug is safe. That's where the 1:1:1 vision hits a wall. You can compress logistics to days. You can't compress "wait and see if people die" to days.
What AI can do is improve data quality. Better documentation, fewer errors, clearer audit trails. That makes regulators' jobs easier, which might speed up approvals indirectly. But it's incremental, not significant.
Who Benefits
Patients, obviously. Faster trials mean faster access to treatments. But also pharma companies, who spend less money and get to market sooner. And healthcare systems, who save money on expensive late-stage interventions by catching conditions earlier with better drugs.
The risk is concentration. If only large pharma companies can afford platforms like Medable, they get faster and cheaper while small biotech firms get priced out. That reduces competition and innovation in the long run.
Medable's response would be that cloud-based platforms democratise access - small firms can rent capabilities they couldn't build themselves. That's theoretically true. Whether it's true in practice depends on pricing and how much expertise you need to use the tools effectively.
The other risk is over-reliance on automation. If AI agents handle all the logistics, do we lose institutional knowledge about how trials work? Do we create systems that are fast but brittle, where a single vendor outage or AI failure cascades into stalled trials across the industry?
Those are real concerns. But they're problems to solve, not reasons to avoid progress. Slower trials don't help anyone.