Highlights & Summary

Deployment and integration drove August's AI news. Epic used its Users Group Meeting to build AI agents, predictive analytics, and autonomous coding directly into the EHR, while Ambience began tying its fees to clinical outcomes and R1 moved to acquire prior-authorization automation from Humata Health. In drug discovery, the infrastructure bets grew larger, with Novo Nordisk naming AWS its strategic AI partner, Aureka and Network Bio raising a combined $150M for biological foundation models, and Anthropic reporting that Claude designed validated protein binders against 14 of 15 targets. Diagnostics saw a run of FDA clearances, led by Leica's first standalone AI quality-control software and C2N's PrecivityAD2, the first Alzheimer's blood test cleared for adults as young as 40.

‍

Dexter: Is your team looking to speak with stakeholders, decision makers, or other KOLs? Dexter Experts is a self-service platform that lets you quickly & price effectively schedule 1-on-1 interviews or surveys. Reach out to the team for more info here.

‍

Headlines

AI in Drug Discovery  

1 | Novo Nordisk names AWS its preferred cloud and strategic AI partner to accelerate drug discovery | Partnership

2 | Anthropic reports Claude autonomously designed de novo protein binders against 14 of 15 targets | New Research

3 | Network Bio launches with $50M and an NVIDIA partnership to build a cell-free RNA foundation model | Company Creation

4 | Aureka Biotechnologies raises $100M Series B to build a biological world model for drug discovery | Financing

5 | Eli Lilly's TuneLab AI models go live in Benchling | New Product

‍

‍

AI in Diagnostics  

1 | Leica Biosystems gains FDA clearances including the industry's first standalone AI quality-control software for digital pathology | Regulatory

2 | FDA clears C2N Diagnostics' PrecivityAD2, the first Alzheimer's blood test for adults as young as 40 | Regulatory

3 | Proscia receives new FDA 510(k) clearance for Concentriq AP-Dx with a Predetermined Change Control Plan | New Product

4 | Aiforia secures €20M in EIB venture debt to scale its AI cancer-detection platform | Financing

5 | Imagene AI joins the Proscia Ready alliance to advance AI-powered companion diagnostics | Partnership

‍

‍

AI in Healthcare  

1 | Epic unveils a wave of AI at UGM 2026, headlined by Ergo, Agent Factory, and Cosmos Curiosity | New Product

2 | R1 to acquire Humata Health to bring agentic AI prior authorization into Phare OS | M&A

3 | Nature Medicine study finds ED physicians' use of an LLM decision-support tool fell from 68% to 30% over four weeks | New Research

4 | Ambience Healthcare launches 'The Ambience Standard,' tying AI fees to clinical and financial outcomes | Commercial

5 | MiiHealth AI raises seed round for agentic patient-intake assistant DAINA | Financing

‍

Summaries

AI in Drug Discovery  

1 | Novo Nordisk names AWS its preferred cloud and strategic AI partner to accelerate drug discovery | Partnership

Novo Nordisk and AWS announced a strategic partnership on August 10 that names AWS as Novo's preferred cloud provider and strategic AI partner, with a co-innovation hub in London where both companies' engineers and scientists work to compress the path from drug target to first human dose. Novo will use services including Amazon Bio Discovery and Amazon Bedrock to identify targets, design therapies faster, and connect genomic, imaging, and clinical data, plus Bedrock AgentCore to run AI agents across operations. Financial terms were not disclosed; the deal extends Novo's existing AI push (which already includes an April OpenAI partnership) and builds on results the company says include reduced clinical documentation time and 25,000+ employees using a Bedrock-based generative AI tool.

2 | Anthropic reports Claude autonomously designed de novo protein binders against 14 of 15 targets | New Research

Anthropic published a technical report describing Claude (Opus 4.8 and its Mythos Preview model) running end-to-end de novo protein binder design campaigns guided only by a ~16,000-word protocol prompt, with no human input on individual design decisions. Across 1,320 designs the models produced 354 confirmed binders against 14 of 15 targets, an overall hit rate near 27% versus the 10-15% Anthropic cites as typical, with per-target hit rates of 22.6-35.1% depending on setup. Rather than generating structures itself, Claude orchestrated open-source tools (RFdiffusion, BoltzGen, PXDesign) and reasoned over the results, while external CROs Adaptyv Bio and Twist Bioscience synthesized and tested every design. The results are self-reported and early-stage (binders are only a first step toward drugs, and targets such as maltose-binding protein and BBF-14 defeated the models), but the prompts, data, and full report were released publicly.

3 | Network Bio launches with $50M and an NVIDIA partnership to build a cell-free RNA foundation model | Company Creation

Network Bio launched from stealth on August 19 with $50M led by Section 32 (joined by Thiel Bio, Founders Fund, and Breyer Capital) and what it calls the world's largest patient-tissue training dataset: 500,000+ patients with paired tissue, blood, and longitudinal outcomes sourced from Mass General Brigham, Penn, Duke, and CU Anschutz. The Palo Alto company also announced a $30M collaboration with NVIDIA to build Nexus, described as the first foundation model trained on cell-free RNA, using NVIDIA's Parabricks and BioNeMo Recipes. Network Bio separately disclosed a $30M+ co-development and licensing agreement with an unnamed Fortune 100 healthcare company, which it frames as early commercial validation of the platform rather than clinical validation.

4 | Aureka Biotechnologies raises $100M Series B to build a biological world model for drug discovery | Financing

Aureka Biotechnologies closed a $100M Series B on August 10, led by Granite Asia with a strategic investor and participation from HighLight Capital and existing backers, bringing total funding to nearly $200M. The AI-native techbio company will direct proceeds toward large-scale training of its next-generation biological foundation models for de novo molecular design, structure modeling, and function prediction, and toward upgrading its Lab-in-the-Loop experimental feedback engine. The company frames its strategy as shifting from using AI to make discovery more efficient to using AI to model living systems, and later in August reported winning a blinded AI antibody design benchmark.

5 | Eli Lilly's TuneLab AI models go live in Benchling | New Product

Eli Lilly's TuneLab models became installable directly inside Benchling via its Model Hub registry, extending models trained on over $1B of Lilly's proprietary research data to Benchling's 1,300+ biotech customers. The lineup includes AbLab, a multi-task foundation model that predicts antibody developability (thermal stability, aggregation, viscosity, solubility) from heavy and light chain sequences, and ChemLab, which returns small-molecule ADME/Tox predictions from a SMILES string. Access is reciprocal: participating biotechs contribute their own assay data through TuneLab's federated learning system, which is designed never to expose raw data.

‍

‍

‍

AI in Diagnostics  

1 | Leica Biosystems gains FDA clearances including the industry's first standalone AI quality-control software for digital pathology | Regulatory

Leica Biosystems (a Danaher company) announced multiple FDA 510(k) clearances on August 11, headlined by Aperio iQC DX, the industry's first standalone AI-assisted quality-control software for clinical digital pathology. The software automatically detects six common whole-slide-image artifacts (air bubbles, pen marks, clipped tissue, missing tissue, out-of-focus regions, and image striping) while slides are still on the scanner; in a real-world study at Heidelberg University's Institute of Pathology, the AI caught up to 24% more artifacts than histotechnicians and cut hands-on review time by 69%. The clearances also added the Aperio GT 180 DX scanner for mid-volume laboratories and expanded capabilities for the existing GT 450 DX.

2 | FDA clears C2N Diagnostics' PrecivityAD2, the first Alzheimer's blood test for adults as young as 40 | Regulatory

The FDA cleared C2N Diagnostics' PrecivityAD2 blood test on August 20 for adults 40 and older who present with signs of cognitive impairment, making it the first FDA-cleared Alzheimer's blood test available to patients that young and the third cleared overall. It is the first FDA-cleared test to combine high-resolution mass spectrometry with a proprietary multi-analyte algorithm, measuring plasma Ab42/40 and %p-tau217 ratios and combining them into an Amyloid Probability Score 2 (APS2) from 0-100 to estimate the likelihood of amyloid pathology. The test is intended to aid diagnosis alongside clinical assessment rather than as a screening or standalone tool, with the FDA-cleared version expected to become available later this year.

3 | Proscia receives new FDA 510(k) clearance for Concentriq AP-Dx with a Predetermined Change Control Plan | New Product

Proscia received a new FDA 510(k) clearance on August 6 for its Concentriq AP-Dx primary-diagnosis platform, adding compatibility with Leica's Aperio GT 450 DX scanner, cloud deployment, and a Predetermined Change Control Plan (PCCP). The PCCP lets Proscia validate and add support for additional cleared scanners, image formats, and displays without filing a new 510(k) each time, making interoperability a competitive differentiator as labs mix hardware from multiple vendors. Proscia, which serves laboratories handling 12 million patient cases annually and 16 of the top 20 pharmaceutical companies, also earned the top score for integration support in KLAS Research's 2026 Digital Pathology report.

4 | Aiforia secures €20M in EIB venture debt to scale its AI cancer-detection platform | Financing

Finnish AI-pathology company Aiforia signed a €20M venture debt agreement with the European Investment Bank, delivered in three tranches (starting with €5M) under the EU's InvestEU programme, with each drawdown tied to revenue and other interim targets. The funding will support development of Aiforia's deep-learning platform for digital tissue analysis and tumor detection and expand its commercial operations internationally. Aiforia and the EIB also agreed a synthetic warrant arrangement that avoids immediate share dilution for existing shareholders.

5 | Imagene AI joins the Proscia Ready alliance to advance AI-powered companion diagnostics | Partnership

Imagene AI, an AI precision-oncology company, announced on August 18 that it joined the Proscia Ready partner alliance, integrating its H&E screening and immunohistochemistry (IHC) image-analysis tools with Proscia's Concentriq platform. The integration supports using H&E-based patient screening to inform molecular testing, plus AI-powered image analysis for developing IHC-based companion diagnostic programs. It targets pharmaceutical companies, diagnostic assay and platform providers, and CROs looking to operationalize companion-diagnostic programs with centralized image and data management.

‍

‍

AI in Healthcare  

1 | Epic unveils a wave of AI at UGM 2026, headlined by Ergo, Agent Factory, and Cosmos Curiosity | New Product

At its Users Group Meeting (August 17-20 in Verona, WI), Epic detailed a broad push to build AI into the EHR itself, led by Ergo, an AI-native clinical interface arriving in November that fuses the EHR, MyChart, and the Cosmos research database into clinician-ready takeaways. New tools include Agent Factory, a low-code platform for deploying and governing 120+ prebuilt AI agents (broad availability in 2027); Cosmos Curiosity, which answers population-health questions and predicts individual risks like readmission using ~320M de-identified records; autonomous coding (Penny) for radiology and emergency medicine; and real-time prior-authorization checks. The moves signal Epic absorbing capabilities once sold by third-party bolt-on vendors into its core platform.

2 | R1 to acquire Humata Health to bring agentic AI prior authorization into Phare OS | M&A

Revenue-cycle leader R1 agreed on August 18 to acquire Humata Health, an AI-powered "touchless" prior-authorization company, folding its technology into R1's Phare OS revenue-cycle platform and its R37 agentic-AI innovation lab. Humata's platform matches payer policies and connects to payers in real time to automate prior authorization end to end; R1 cites Humata metrics including a 96% first-pass approval rate, 30% fewer write-offs, and a 45% reduction in staff touches. Financial terms were undisclosed, and the deal is expected to close by the end of the third quarter of 2026.

3 | Nature Medicine study finds ED physicians' use of an LLM decision-support tool fell from 68% to 30% over four weeks | New Research

A prospective DECIDE-AI stage 1 evaluation published August 19 in Nature Medicine tested SHAKED, a clinical decision support system built on multiple large language models, in a tertiary emergency department, analyzing 1,138 patients over four weeks across one wing using SHAKED and one following routine care. The system was safe (no adverse events, with expert reviewers rating 99 of 100 sampled outputs clinically appropriate), but clinician adoption fell from 68% to 30% as physicians disengaged during busier shifts, and emergency-department length of stay was unchanged. The authors conclude that sustained clinician engagement, not algorithmic accuracy, may be the main barrier to effective clinical AI, and that the results inform randomized trial design but do not yet justify deployment.

4 | Ambience Healthcare launches 'The Ambience Standard,' tying AI fees to clinical and financial outcomes | Commercial

On August 19, Ambience Healthcare introduced The Ambience Standard, an outcome-based partnership model that puts its fees at risk against each health system's own targets rather than charging per seat, token, or feature. The model pairs performance-based contracts with forward-deployed teams of clinicians, engineers, and value specialists embedded on-site, and runs on Ambience Chorus, the company's healthcare AI agent platform spanning 200+ medical and surgical subspecialties. Ambience cited customer results including Cleveland Clinic (4,000 clinicians onboarded in four months at 70% utilization) and a peer-reviewed St. Luke's study in which physician burnout fell from 45% to 31%.

5 | MiiHealth AI raises seed round for agentic patient-intake assistant DAINA | Financing

Phoenix-based MiiHealth AI closed a seed round in August (reported at ~$2.8M, though the company's release did not disclose the figure) led by Russell Glass, former Headspace CEO and founder of Arteria Capital, with physician angels and digital-health operators participating. The capital will scale DAINA (Dynamic AI Intake and Navigation Agent), which autonomously calls patients before visits, conducts a multi-turn clinical intake in the patient's preferred language, and writes a structured note into the EHR, reportedly giving clinicians back more than two hours a day. MiiHealth plans to expand its specialty-protocol catalogue, deepen clinical-reasoning and safety checks, and accelerate integrations with major EHRs.