Here is a fact about the pharmaceutical industry that sounds like a rounding error and is actually the whole story: roughly 90% of drug candidates fail. Companies spend a decade and a fortune to learn, most of the time, that a molecule does not work. The obvious response is to make better decisions earlier, and the not-obvious problem is that the information needed to make those decisions is scattered across hundreds of papers, patents, and trial reports, much of it written by competitors, much of it decades old, and a surprising amount of it printed as dots on a chart that no keyword search will ever find.
Someone has to gather all of that. Traditionally that someone is a highly trained scientist, and the gathering takes months. This is the bottleneck Delineate walked up to and decided was, in fact, a business.
Delineate builds AI agents that extract and structure data from the scientific literature at scale. That much you could say about a dozen companies. The interesting part is what the company chose to obsess over: the figures. A lot of the most valuable numbers in a research paper never appear as text. They exist only as a curve, a scatter of points, an error bar on a plot. Delineate pairs custom large language models with computer vision that reads those images the way a trained analyst would, and then - this is the part that matters in pharma - runs the output through a quality-control process, because a confidently wrong number is worse than no number at all.
The result, the company says, is roughly a 15x improvement in processing speed versus industry standard, and studies that shrink from months to weeks. That compression is not a vanity metric. In an industry where saving a single day of development can be worth $1 to $5 million, speed is a strategy. Delineate's pitch to a pharma customer is not "we have a clever model." It is "we will hand you, in weeks, an evidence base that would otherwise have cost you a team and a quarter - and every downstream decision gets sharper because of it."
The technical name for a lot of what this enables is model-based meta-analysis, or MBMA, a standard tool of modern drug development. MBMA leverages published summary data alongside a company's internal data to inform decisions like benefit-risk assessment and comparative efficacy - the kind of judgment calls that determine whether a program advances. The catch, always, is that MBMA starts with a human reading everything. Delineate automated the reading, not the judgment. The scientists still decide. They just decide with more evidence, sooner.
What Delineate is selling, in other words, is not a replacement for expertise. It is leverage on it. The company's own tagline - "quantitative evidence for consequential decisions" - is almost aggressively unglamorous, which is a point in its favor. It is not promising to cure anything. It is promising to make the people who might cure something less blind.