BenchSci vs Microsoft Discovery

The largest technology companies have moved into biopharma, each bringing its own AI for the lab. For R&D leaders, the question is no longer whether to use agentic AI in drug discovery. It is which kind to trust with it: a general enterprise platform for all of R&D, or one built specifically for biopharma R&D.

EMET by BenchSci and Microsoft Discovery are two of the names that come up in that decision. Both are agentic systems that reason over scientific data, orchestrate specialized agents, and cite their sources. Microsoft Discovery reached general availability in June 2026 and spans chemistry, materials science, silicon design, energy, manufacturing, and pharma. EMET is narrower on purpose, built only for biopharma R&D.

This piece measures the two platforms on five things a preclinical organization depends on: the data each system reasons over, the depth of its biology-specific workflows, the service that drives adoption, the freedom to use the best model for each task, and whether the vendor could become a competitor. The two platforms are evenly matched on two of these. On the other three, BenchSci pulls ahead.

What to look for in AI for preclinical R&D

When you evaluate AI for preclinical R&D, the demo rarely tells you what you need to know. The traits that decide whether a platform is still earning its place a year later are structural, and easy to skip when the output on screen looks polished. Five of them do most of the work.

• Proprietary, science-specific evidence to reason over. A model can only reason as well as the evidence beneath it. Curated, structured biology beats a general system pointed at the same public sources everyone else can reach.
• Depth of biology-grounded workflow and integration. Preclinical R&D is spread across thousands of data sources and hundreds of tools. Value comes from a platform that binds that stack into the scientist's daily work and already understands the biology.
• A service model that drives adoption. Buying software changes nothing on its own about how a lab works. People on the ground who build workflows for specific teams decide whether the tool is still in use a year later.
• The freedom to use the best model for each task. No single model wins every scientific job, and the frontier moves constantly. A platform welded to one lab's models inherits that lab's roadmap, pricing, and outages.
• A partner that will not become a competitor. You are handing this system your targets, your assays, and the programs that define your pipeline. Whether the vendor might one day develop drugs of its own matters before that IP leaves your walls.

The sections below take EMET and Microsoft Discovery through each one in turn.

Criterion 1: the data and evidence the platform reasons over

Both platforms build a knowledge graph, and both can fold in a customer's proprietary data, so the graph itself is not the dividing line. The dividing line is whether the platform brings a curated body of biological evidence no competitor can reach, or assembles one from whatever sources you point it at.

BenchSci reasons over a proprietary knowledge graph and closed-access science

Evidence sets the ceiling on any answer. A decade of BenchSci's data work sits under EMET. EMET reasons over a proprietary Biological Evidence Knowledge Graph of 858M nodes and 2.2B relationship edges. Beneath it sit 38M+ publications, 16M of them closed-access papers opened through eight years of licensing with Elsevier, Springer Nature, Wiley, Oxford University Press, and dozens more, alongside a large reagent and model-systems database.

Every data point is curated by PhD scientists, and the corpus lives in EMET and nowhere else. A neuro-symbolic evaluation loop ties every generated output to the verified graph, reaching 95%+ accuracy on biological questions — 2–4x better than frontier models alone — validated across 600+ tests and 8+ benchmarks. Without that biological ground truth, a general system returns answers that look right and are not, and a wrong answer can cost months of wet-lab time.

Microsoft Discovery builds a graph from your data and external sources

Microsoft Discovery is built on a graph-based knowledge engine, and it is genuinely capable. According to Microsoft's launch blog, the engine goes beyond retrieving facts to build graphs of relationships across "proprietary data as well as external scientific research," and it keeps a long-term memory graph so results can be reused across projects.

Microsoft Discovery does not come with a biomedical dataset of its own. It reasons over your data and outside scientific literature instead. At launch, that literature comes from partners like Wiley and Causaly.

That is a sensible design for a platform meant to serve every research domain. But it means the evidence is either data you already own or sources any competitor can license too. Microsoft Discovery can structure evidence well; it does not supply the decade of curated biology that sits under EMET.

Criterion 2: depth of biology-grounded workflow and integration

Discovery does not live inside a chat box. It moves across databases, internal systems, and the bench, and it is specific to biology at every step. The question is whether the platform arrives understanding that work or expects you to build the understanding in.

EMET is a biology-grounded platform built around preclinical workflows

Biology has fragmented across thousands of sources and hundreds of tools, and scientists have become the connective tissue between them, losing 40% to 60% of their time to finding and cleaning data. Closing that gap is what EMET is for. It carries 200+ proprietary scientific skills, and over 250 agents and workflows spanning the full preclinical journey: target identification and validation, hit discovery, lead optimization, safety and toxicology, translational biomarkers, IND-enabling studies

It also acts in the lab through lab-in-the-loop connectivity to automated infrastructure — closing the loop between computational hypothesis and physical experiment. Fusing those pieces into one biology-grounded system is what makes EMET a platform rather than a marketplace, where the value comes from binding the pieces together, not from the pieces themselves. Each new connector and workflow raises the value of the next, so EMET settles into how scientists already operate and grows more useful with use.

Microsoft Discovery is a horizontal R&D platform you shape to your science

Microsoft Discovery is a powerful general R&D platform that works across many industries. It spans chemistry, materials, energy, and pharma. Its best-known result so far comes from materials science, not biology: Microsoft reports it used the platform to screen and identify a novel, non-PFAS immersion coolant for data centers in about 200 hours, then synthesized the prototype in under four months.

That breadth is also the trade-off for a preclinical team. Microsoft Discovery's agents are defined by the customer in plain language, so the biology expertise has to come from your own scientists and partners — the platform's documentation doesn't claim to supply it.

A platform you can adapt to any field is flexible. But your R&D stack is already crowded with tools, and a general platform can become one more thing to set up rather than the biology-grounded layer that ties everything together. EMET arrives fluent in preclinical R&D; Microsoft Discovery gives you the frame and leaves the biology to you.

Criterion 3: the service model that drives adoption

Good software still dies if nobody adopts it. Whether a platform survives its first year rarely comes down to the interface. It comes down to who shows up to bend the tool around each team's real work, and that service layer tends to be the quiet decider.

BenchSci deploys a dedicated scientific team that builds your workflows

A license alone changes nothing about how a lab works, so BenchSci staffs the gap directly. Every deployment comes with a dedicated PhD team who understands your biology and your therapeutic areas — they run training, manage change, and build the connectors and workflows that fit how your people actually operate, tuned per organization rather than pulled off a shelf.

The team stays involved as an ongoing scientific partner. That's why customers build their programs around EMET: when the workflows are built around the organization, the platform gets adopted instead of shelved.

Microsoft Discovery is delivered through Azure and systems integrators

Microsoft Discovery is delivered as an Azure product with enterprise-grade security and identity, and for individual researchers there is now a self-serve desktop app in preview. Azure-native governance and a low-barrier entry point are real strengths.

For scaled rollouts, Microsoft points to systems integrators such as Accenture and Capgemini to build and scale custom deployments. That leaves the biology-specific workflow design and the push for adoption with the customer and its consultants. A systems integrator can stand up a platform — that's a different thing from PhD scientists who know your therapeutic area working directly with your research teams.

Criterion 4: model flexibility and vendor independence

No lab holds the frontier for long, so the platforms worth committing to are the ones free to reach for a better model whenever one appears. This is the first of two criteria where BenchSci and Microsoft Discovery land in the same place.

BenchSci is model-agnostic and routes every task to the best model

Because EMET builds no model of its own, every question can go to whatever performs best on it rather than to whatever a vendor happens to sell. It selects per step: a frontier model for open reasoning, a specialized biomedical model such as ESM-2 or AbLang2 where sequence or structure is the question, and each new frontier model folded in as it ships.

For the buyer that is both a performance advantage and protection against lock-in. The best model wins the task, and when a stronger one ships you get the upgrade automatically instead of paying to switch platforms.

Microsoft Discovery lets you bring and choose your models too

Microsoft Discovery is model-extensible by design. Microsoft says teams can integrate their own models, tools, and datasets alongside partner and open-source solutions, and can even suggest which models the agents should use. It runs on Azure AI Foundry, whose catalog spans many providers and supports bringing your own model, so a Microsoft Discovery customer is not tied to one model family.

The mechanism differs between the two. EMET routes automatically to the best model for each scientific task, while Microsoft Discovery gives you the freedom to assign models within the Azure ecosystem. On the question that decides lock-in, both platforms answer the same way.

Criterion 5: whether your AI partner also develops drugs

In drug discovery the vendor relationship is unusually intimate. You are trusting a partner with your targets, your assays, and the programs that define your pipeline, so it's fair to ask whether that partner might one day develop drugs of its own. This is the second criterion where the two platforms stand on the same ground.

BenchSci has committed never to develop its own drugs

BenchSci stays on the tooling side of the line by design and has committed never to discover its own drugs. For a pharma organization handing over its most sensitive IP, that permanent commitment is a clean trust signal — and it's structural rather than a promise: BenchSci's only path to winning is making your scientists more effective, not building a pipeline of its own. EMET connects to your internal data, models, software licenses, and cloud, so your science and your IP stay yours.

Microsoft is a platform provider with no drug program of its own

Microsoft has no in-house drug pipeline or therapeutics business, and Microsoft Discovery is positioned as tooling for other people's research. Its pharma work runs through customers and partners such as GSK and Ginkgo Bioworks rather than a Microsoft drug program. As a horizontal platform serving every research domain, Microsoft has even less reason than a life-sciences specialist to compete with its customers' pipelines.

So neutrality is not where these two separate. Unlike some frontier labs that have entered drug discovery themselves, Microsoft has stayed a platform provider, and both companies can make the same commitment today. The real separation runs through the other three questions: the data each reasons over, the depth of its biology-specific workflows, and the people who make it stick.

Which platform belongs in preclinical R&D

This was never an argument about raw capability. EMET and Microsoft Discovery are both strong agentic platforms, both reason over scientific data, and both demo well. Across the five criteria that decide a preclinical platform, they're evenly matched on two — model flexibility and neutrality — and on the other three the separation runs in BenchSci's favor.

EMET reasons over a proprietary knowledge graph and closed-access science that exist nowhere else, where Microsoft Discovery structures a graph from data you already own and sources anyone can license. EMET arrives fluent in preclinical workflows, where Microsoft Discovery is a horizontal platform you teach your science. And BenchSci sends in scientists, where Microsoft points you to systems integrators.

Microsoft Discovery is a capable, general-purpose R&D platform, strongest where your problem looks like every other industry's. EMET by BenchSci is built for one industry and goes deep: grounded in proprietary biology and staffed by scientists. When the question is where your programs and your data will live for the next several years, that's the difference that lasts.

See what EMET does on your own targets. Book a demo with the BenchSci team.