Variant-aware oncology literature intelligence.

PubMed indexes roughly 150,000 oncology-relevant publications per year. ASCO, ESMO, and AACR each release 2,000 to 5,000 abstracts annually. ClinicalTrials.gov updates daily with new starts, milestone changes, and arm closures. Medical affairs teams at oncology biotechs are responsible for keeping up with all of it, framing it accurately, and defending the framing to their CMO, the regulatory team, and the field force.

Most teams do not keep up. They cope.

The cope strategies fall into three categories. First, manual review by a senior MSL or medical writer, two to three weeks per surveillance cycle, by which time the literature has moved on. Second, paying for a sweeping enterprise tool. Causaly is the standard there: a strong product with citation-grounded reasoning across all of biology, but priced at the mid-six-figure-and-up enterprise ACV typical of biopharma R&D platforms, which is a heavy lift for a biotech with one or two oncology drugs in development. Third, using a generic LLM to summarize papers, which produces fluent text that often hallucinates citations and is therefore not pharma-defensible.

Elsevier Embase fills part of the gap as the incumbent literature database, but it is a search-first product, not an AI-native synthesis layer. The Embase license starts around $250K and scales from there, and the workflow assumes a librarian or analyst running queries by hand.

The harder constraint is defensibility. Medical affairs is not a SaaS marketing function. Every piece of communication that goes from a medical affairs team to an HCP, an MSL talking point, a congress slide, a medical inquiry response, has to be traceable to a primary source. A summary that paraphrases a paper without linking back to it is not a usable artifact. AI-generated content that cannot survive a medical leadership review is worse than no content, because it costs analyst time to detect and flag.

The market gap is concrete: oncology medical affairs teams need a citation-grounded surveillance product that is oncology-specific (not all of biology), variant-aware (not just keyword search), and available without a top-down enterprise purchase. Engine 3 is built for that gap, whether the team is a lean biotech or an oncology medical affairs group inside a larger pharma.

Causaly is the right competitor to discuss in detail. They are well-funded ($93M raised), they serve 12 of the top 20 pharma companies, and they are the closest in spirit to what Engine 3 does: AI-driven, citation-grounded literature reasoning across biomedical knowledge.

Three honest differences.

Scope. Causaly indexes all of biology. That is appropriate for a top-twenty pharma with diverse R&D pipelines. It is overkill for an oncology biotech that only needs deep coverage of, say, EGFR-mutant NSCLC. Engine 3 is oncology-only and goes deeper inside that scope: variant-aware reasoning, oncology-specific KOL graphs, and conference coverage tuned to ASCO, ESMO, and AACR rather than every biomedical conference.

Pricing. Causaly's enterprise pricing is not publicly disclosed, but is generally understood to sit in the mid-six-figure-and-up ACV band typical of biopharma R&D platforms. That is a multi-stakeholder procurement cycle for any company that does not already buy at that level. Engine 3 starts at $149 per month for an individual MSL who wants to try it for two weeks before bringing the team in. The PLG self-serve tier is the on-ramp Causaly does not have.

Sales motion fit. Causaly's sales motion targets head-of-medical-affairs and CMO buyers. Their inside sales process is built for a 6 to 9 month enterprise procurement cycle. A 50 to 500 employee biotech medical affairs team cannot wait 6 to 9 months to start tracking the literature on their drug. They need a tool the lead MSL can sign up for next Tuesday and get value from before the team budget conversation. The same self-serve on-ramp works for an oncology medical affairs group inside a larger pharma that wants to evaluate oncology-specific depth without a top-down enterprise purchase. That is the gap Engine 3 fills.

Both companies share the citation-grounded approach, which is the right architecture for pharma defensibility regardless of vendor. Engine 3 is not arguing against Causaly's architecture; it is arguing that the same architecture, applied with oncology depth and a self-serve PLG tier, fits a different and underserved market segment.

Engine 3 is not a medical device. It is a knowledge product, not a diagnostic, therapeutic, or treatment-recommendation tool. There is no FDA premarket review exposure, no IDE, and no claim to direct clinical use. Outputs are intended for medical affairs teams reviewing the literature; they are not patient-care recommendations.

The compliance layer that does matter for medical affairs customers is documentation and traceability. Every Engine 3 output, including answers, digests, KOL profiles, and inquiry responses, carries inline citations and provenance back to its primary source, and AI-drafted inquiry responses stay human-review-only before anything goes to an HCP. This is consistent with the documentation standards that ACCME (for CME), EFPIA (for European pharma), and the standard medical-leadership-review processes at US pharma companies expect.

SOC 2 Type II is on the roadmap. Target window for completion is shared with enterprise prospects under NDA. Engine 3 does not handle PHI by default; it works on published literature and account data, neither of which is patient-identifiable, so the self-serve tiers sit outside PHI scope entirely. UNMIRI's underlying HIPAA-ready posture (the AWS BAA covering Engines 1 and 2) is the foundation a future PHI-adjacent Engine 3 workflow would build on; no such workflow ships today, so a BAA is scoped per Enterprise engagement rather than offered as a standing line item. Standard customer agreements include data retention, license attribution for indexed sources, and the same Anti-Kickback-Statute-clean trial-matching posture documented on the CDS API page.

The full subprocessor list, BAA status, and incident-response posture for the underlying platform live on the security overview. Customers who opt into the HIPAA-ready BAA scope at the Enterprise tier inherit that posture; the self-serve tiers are scoped to non-PHI data by default.