Verana Health ships Agent Claire for life-sciences R&D on real-world data

Verana Health announced Agent Claire for Life Sciences, disease-specific AI agents that let researchers ask conversational questions against Verana real-world datasets in oncology, ophthalmology, and urology. The agent builds execution plans, checks clinical guardrails, returns patient distributions and outcomes with SQL-ready logic, and can plug into Claude, Gemini, Copilot and others via Model Context Protocol (MCP).

HEALTH desk — when life-sciences teams can chat their way into curated specialty datasets with guardrails and SQL receipts, R&D hypotheses move from weeks of analyst queues to minutes of on-demand evidence.

What the lede says the product is. Verana Health calls itself a digital health company dedicated to revolutionizing patient care and clinical research through real-world data. It says it released new artificial intelligence agents to aid researchers. Agent Claire for Life Sciences comes in disease-specific versions. It is meant to help researchers generate analytic insights quickly. A researcher asks a conversational question. The questions can be simple or sophisticated. The agent writes an execution plan. It then checks that plan against embedded clinical guardrails. The release says researchers get immediate, direct access to the data, without programming skills or an analyst in the queue. Real-world data, in plain words, is de-identified patient records from ordinary care. It is not the notebook from a clinical-trial lab. A clinical guardrail, here, is a check inside the agent. A plan that does not fit the data, or a question the clinic would not stand behind, should not come back as a clean answer. These lines are the release’s. This desk did not run a query.

Where it runs, and who it is for. Agent Claire is available across Verana real-world datasets in oncology, ophthalmology, and urology. Oncology is cancer. Ophthalmology is eye care. Urology covers the urinary tract and male reproductive organs. Those three specialties are the availability sentence. The release says brand managers, medical affairs teams, market access teams, and HEOR professionals can ask a question and get an answer in minutes. The answer can include patient distributions, treatment patterns, and outcomes. HEOR is health economics and outcomes research. That is the work of asking what a treatment costs and what it does for patients outside one trial. The release says hypotheses that used to take days or weeks are now on demand. Minutes, and days or weeks, are the release’s clocks. This desk did not time a question.

What the release says a team can do with it, and the SQL that comes back. Access to Agent Claire is supposed to compress R&D cycles. Researchers can identify a patient population, check that enough patients and enough data quality exist for a cohort analysis, and ask an outcomes question directly. A cohort is a defined group of patients. The release says the answers come back in minutes instead of weeks, with SQL output a data scientist can validate. SQL is the query language that names which rows were counted. The receipt is that query. Someone who writes code can rerun it. These lines are the release’s. This desk did not open a SQL editor.

The chief executive, as a quote. Sujay Jadhav, chief executive of Verana Health, said AI agents are a shift away from tools that simply follow instructions, toward systems that help integrate information and guide next steps. He said that as healthcare grows more complex, the company is proud to broaden its capabilities for the life sciences community. The agents, he said, help teams navigate and interpret clinical information, and they offer context and guidance when teams need it, so researchers can gain speed with confidence. That is his sentence on the wire. This desk did not interview him.

Three example prompts the release prints, and only those. A researcher looking at therapies for multiple myeloma might ask how many patients received at least one line of therapy. Then how many of those received CAR-T therapy. Then the names of those CAR-T therapies and how often each was used. Multiple myeloma is a blood cancer. A line of therapy is one course of treatment, first, then second, and so on. CAR-T is a treatment that engineers a patient’s own immune cells to attack the cancer. The release also gives a higher-level objective: map the treatment progression for multiple myeloma patients moving from first-line therapy to CAR-T. The agent is then supposed to build a step-by-step query strategy across each phase, so a person can review it, refine it, or run it. These are examples on the page. This desk did not count patients.

What the release says about rigor, and the quote with no name. For organizations that use real-world evidence for commercial strategy, medical planning, and launch decisions, the release says more people can ask better questions and get faster answers. Every answer includes the underlying query logic. People who do not write code can iterate: refine a hypothesis, filter a cohort, or switch disease areas, without waiting on a technical bottleneck. Verana’s evaluation framework, the release says, lines the answers up with validated queries and adds caution when the analysis is more complex. The agent is said to plug into the clinical systems an organization already uses. A second quote is attributed only to “Chief Digital & Technology officer, top-5 global pharma.” The page does not name the person. It does not name the company. The quote says the shift is from digital tools that assist people to AI agents built into core operations, and that those agents can speed hypothesis testing and development timelines. Those sentences are the release’s. This desk did not identify the speaker.

How it connects to other agents. Agent Claire is built to work with other AI agents. The release names Claude by Anthropic, Google Gemini, Microsoft Copilot, and others. The plug is the Model Context Protocol, MCP. MCP is an open standard so one agent can call another. The release says a person can stay inside the organization’s own agent and call Agent Claire, and its clinical training, in the background. These lines are the release’s. This desk did not connect Claude, Gemini, or Copilot.

The McKinsey sentence, and whose number it is. The release says that according to McKinsey & Company, life sciences agents will boost clinical development productivity by 35 to 45 percent by 2030, and unlock $60 billion to $110 billion or more in annual revenue across pharma operations. The page prints 35-45% and $60-110B+. That is a McKinsey forecast the wire cites. It is not a result Verana measured for Agent Claire. This desk did not open the McKinsey paper. 2030 is the forecast’s year. Do not file those figures as this product’s productivity.

The meeting the release flags, and what the about box adds. The release points readers to veranahealth.com, and to Verana Health booth 2301 at the AAO 2026 Annual Meeting in New Orleans, Louisiana, October 9–12. AAO is the American Academy of Ophthalmology. That expansion is this desk’s. The release prints the letters and the city and the dates. It does not spell the name out. The about box says Verana Health, Inc. uses physician expertise and artificial intelligence on real-world data. It claims exclusive access to the world’s largest patient data sources in oncology, ophthalmology, neurology, and urology. Neurology is on that list. It is not in the sentence that says Agent Claire is available in oncology, ophthalmology, and urology. World’s largest is the company’s claim. The site it names is www.veranahealth.com. SOURCE Verana Health is the last line on the wire. These lines are the about box, except the AAO expansion.

What the still shows. The card is Verana’s press graphic for this launch. On a white field it reads “NEW AI AGENTS” and “to Expedite and Streamline R&D Tasks.” An inset shows an AI AGENTS screen titled Agent Claire for Life Sciences, with the line “Conversational access to real-world datasets,” and a sample prompt: “How many patients received at least one line of therapy?” The Verana Health mark is on the graphic. The still does not print a date. That sample prompt is also one of the release’s multiple-myeloma examples. The inset does not show an answer, a patient count, or a SQL block. Do not treat the picture as a query this desk ran.

Plain English for the rest of the card. Real-world data means de-identified records from care, not a trial lab. HEOR means health economics and outcomes research. MCP is the open plug that lets Claude, Gemini, or Copilot call Agent Claire. SQL is the query a data scientist can rerun. A cohort is a defined group of patients. Multiple myeloma is a blood cancer. CAR-T engineers a patient’s immune cells to attack that cancer. A line of therapy is one course of treatment. 9:00 a.m. Eastern is the wire’s stamp. 1:00 p.m. UTC is the same minute. Forty-three seconds is the gap between datePublished and dateModified. 35 to 45 percent, and $60 billion to $110 billion, are McKinsey’s forecast for 2030, cited by the wire. They are not a Verana result. Booth 2301, and October 9–12 in New Orleans, are the meeting lines. Neurology is an about-box data area. The agent’s availability sentence names oncology, ophthalmology, and urology.

PRIMARY here: Verana Health’s 24 Sep 2026 PR Newswire release, stamped Sep 24, 2026, 09:00 ET and datelined San Francisco — Tier A PRIMARY, the company’s own wire. The Agent Claire announcement, the disease-specific versions, the conversational questions, the execution plans, the clinical guardrails, the no-programming-skills line, the oncology, ophthalmology, and urology availability, the brand, medical affairs, market access, and HEOR audience, the minutes and the days-or-weeks clocks, the patient distributions, treatment patterns, and outcomes, the cohort and SQL lines, the Jadhav quote, the three multiple-myeloma prompts, the CAR-T mapping example, the query-logic and evaluation-framework sentences, the unnamed top-5 pharma quote, the Claude, Gemini, and Copilot names, the MCP line, the McKinsey citation, booth 2301, the October 9–12 New Orleans meeting, and the about box including neurology are that release’s. The words on the still are the graphic’s. NOT claimed: that this desk asked the agent a question, counted patients, or opened a SQL file, that neurology is in the availability sentence, that world’s largest was surveyed, a name or company for the second quote, that the McKinsey 35 to 45 percent or the $60 billion to $110 billion figure is an Agent Claire result, that this desk read the McKinsey paper, a patient count on the still, or a date printed on the card. Distinct from the already-filed veeva-study-builder-agent, grail-galleri-fda-panel, oracle-health-clinical-ai-agent-nurses, and dun-bradstreet-db-ai.