Verified
Merck & Co.
Scaling- Pharma
- ~75,000 employees
- GPTeal, built in-house
- 80%+ of staff use its AI platform
- Google Cloud AI deal up to $1B (2026)
Merck & Co. deploys AI through GPTeal, an in-house platform that gives employees secure access to OpenAI, Anthropic and Meta models, and through a multi-year Google Cloud deal worth up to $1 billion that brings Gemini Enterprise agents to research, manufacturing, commercial and corporate work. Merck said in Jan 2026 that more than 80% of its workforce uses its proprietary AI platform, and Gaurav Gupta took over IT, data and AI strategy as chief information and digital officer on Sep 1, 2026.
merck.comTheir approach to AI
- The platform
- GPTeal, a platform Merck & Co. built to give staff secure access to outside large language models, carries everyday writing work; a Google Cloud deal worth up to $1 billion adds Gemini Enterprise agents (Apr 2026).
- The models
- GPTeal offers OpenAI, Anthropic and Meta Llama models (Mar 2025); Merck adds Google's Gemini, Anthropic models on AWS Bedrock, and its own KERMT and TEDDY research models.
- The reach
- Merck said in Jan 2026 that more than 80% of its workforce uses its proprietary AI platform; Ron Kim, then chief technology officer, put regular GPTeal users above 50,000 in Mar 2025.
- The catch
- Merck has quantified one AI return, reimbursement dossiers compiled in half the time and cost (Apr 2026), but no firm-level figure, and Merck's chief information and digital officer and Chief AI Officer both changed in Aug and Sep 2026.
Snapshot
- Employees
- ~75,000
- Dec 31, 2025, about 30,000 of them in the U.S. (2025 Form 10-K)
- Technology budget
- Not disclosed
- Google Cloud AI deal worth up to $1B over several years (Apr 2026); R&D expense $15.8B in 2025
- AI/ML headcount
- Not disclosed
- More than 80 experts built the clinical study report platform (Jun 2025)
- Who owns AI
- Gaurav Gupta, EVP and CIDO
- Since Sep 1, 2026, on the Executive Team; succeeded Dave Williams, CIDO from Aug 2020
- Platform
- GPTeal, built in-house
- Plus Gemini Enterprise from Google Cloud (Apr 2026)
- Models
- OpenAI, Anthropic, Meta Llama, Gemini
- GPTeal models as of Mar 2025, Gemini since Apr 2026, plus Merck's own KERMT and TEDDY research models
- Reach
- 80%+ of the workforce
- Jan 2026; more than 50,000 regular GPTeal users in Mar 2025
- Use cases
- No count published
- Production use in discovery, clinical documents, trials, manufacturing and commercial work (Jan 2026)
Key findings
Merck & Co. said on Jan 9, 2026 that more than 80% of its workforce uses its proprietary AI platform, and Ron Kim, then chief technology officer, told Business Insider in Mar 2025 that more than 50,000 employees regularly used GPTeal, the in-house platform that offers OpenAI, Anthropic and Meta models.
Merck signed a multi-year agreement worth up to $1 billion with Google Cloud on Apr 22, 2026 to deploy Gemini Enterprise and an agentic AI platform across research, manufacturing, commercial and corporate functions, with Google engineers working inside Merck teams.
Dave Williams, then Merck's chief information and digital officer, told Reuters on Apr 22, 2026 that Google technology had already cut the time and cost of compiling country reimbursement dossiers by half, the only AI return Merck has quantified; Rob Davis said on Sep 14, 2026 that savings from AI-driven molecular design are built into budgets, without an amount.
Merck said in Jun 2025 that its generative AI platform for clinical study reports cut first drafts from two to three weeks to three to four days and the time for a fully human-reviewed first draft from an average of 180 hours to 80, with errors halved.
Gaurav Gupta, a former EY senior partner who joined Merck in Jul 2026, became executive vice president and chief information and digital officer on Sep 1, 2026, succeeding Dave Williams; CDO Magazine reported in Aug 2026 that Merck's Chief AI Officer, Anton Groom, had left to become GE Aerospace's chief data and AI officer.
Where Merck sits
- Stage
- Scaling
Merck & Co. sits at Scaling: Merck said in Jan 2026 that more than 80% of its workforce uses its proprietary AI platform, AI runs in production across drug discovery, clinical documents, trial operations, market access, manufacturing, supply chain and commercial work, and chief information and digital officer Gaurav Gupta, an Executive Team member, owns the AI strategy. Merck & Co. falls short of Embedded because it has not published how often employees use the platform or how many use cases run in production, and its only quantified AI return covers one process, reimbursement dossiers compiled in half the time and cost (Apr 2026), with no firm-level financial effect.
How the stages are defined - Ownership
- EVP and CIDO Gaurav Gupta, on the Executive Team (since Sep 2026)
- Platform
- Hybrid: GPTeal built in-house; Gemini Enterprise bought from Google Cloud
- Models
- OpenAI, Anthropic and Meta Llama in GPTeal; Google Gemini; own KERMT and TEDDY
- Hosting
- AWS (Bedrock, SageMaker) and Google Cloud; GPTeal's hosting not disclosed; company data kept from external exposure
- Reach
- More than 80% of the workforce (Jan 2026)
How they deploy
Build the platform, bring in the models
Merck built GPTeal as its own generative AI platform so employees can use outside large language models, including OpenAI's ChatGPT, Meta's Llama and Anthropic's Claude, while company data stays protected from external exposure (Business Insider, Mar 2025). Ron Kim, then chief technology officer, said Merck wanted to stop scientists spending their time as copyeditors, and early use covered emails, memos and regulatory documents that people review before submission.
Google Cloud as the agentic AI partner
Merck signed a multi-year agreement worth up to $1 billion with Google Cloud on Apr 22, 2026 to build an agentic AI platform across research, manufacturing, commercial and corporate functions on Gemini Enterprise, with Google Cloud engineers working alongside Merck teams. R&D World reported that the deal covers infrastructure, engineers and Gemini Enterprise licenses, and Dave Williams told Reuters he expects at least a decade-long partnership. TechTarget reported that forward-deployed engineers from Google Cloud and Google DeepMind will help train foundation models for drug discovery, rework workflows around agents and train Merck's 75,000 employees, while Merck keeps using AWS, Microsoft and ServiceNow. Rob Davis told investors on Apr 30, 2026 that the Google Cloud deal, an expanded collaboration with Tempus AI and an agreement with Mayo Clinic are steps to accelerate Merck's AI transformation.
Own models for discovery, with the wet lab
Merck builds its own AI models for drug discovery, including KERMT, a small-molecule property model developed with NVIDIA and released as open source in Dec 2025, and TEDDY, a family of single-cell foundation models built with BCG X's AI Science Institute (Jul 2025). Rob Davis said in Sep 2026 that Merck does not expect to go straight from computer models to human trials and pairs the wet lab with the dry lab, and that AI's clearest payoff so far is in molecular design and optimization.
Outside data and discovery partners
Merck adds outside data and AI tools through research deals: Protillion Biosciences' lab-in-the-loop AI for antibody design on several targets, with up to $510 million in milestone payments (Jun 2026), Tempus AI's de-identified data, Lens platform and Workspaces computing environment for oncology biomarker work (Mar 2026), Mayo Clinic's clinical, imaging and molecular data for AI-enabled target discovery in inflammatory bowel disease, atopic dermatitis and multiple sclerosis (Feb 2026), Illumina's Billion Cell Atlas of CRISPR perturbation data for training AI models, as a founding participant (Jan 2026), Variational AI's Enki generative chemistry platform for two undisclosed targets in a deal worth up to $349 million (Sep 2025), and Turbine's simulated cell models for hard-to-treat cancers (Aug 2025).
Self-funded transformation
Merck has paid for its transformation projects out of annual savings from process efficiency work and uses Celonis process mining to decide where AI is applied, according to Constellation Research (Apr 2026). Merck's multiyear optimization initiative targets $3.0 billion in annual cost savings by the end of 2027, all reinvested in growth areas, but Merck's filings do not credit AI with any share of it.
- Deployment model
- Merck runs AI from a central information technology, digital, data and AI organization led by the chief information and digital officer, an Executive Team member, with delivery inside the divisions. Merck Research Laboratories has its own data science and AI groups that build discovery models such as KERMT, the manufacturing division runs computer-vision inspection, and commercial teams use generative AI for content and field support. Merck opened a Hyderabad Technology Hub in 2025 with cloud, cybersecurity, AI and machine learning, automation, data science and digital product staff, and under the Apr 2026 Google Cloud deal Google engineers work alongside Merck teams. Constellation Research reported in Apr 2026 that Merck moved its Celonis center of excellence into a Chief AI Office; CDO Magazine reported in Aug 2026 that Anton Groom, Merck's Chief AI Officer, had left for GE Aerospace, and Merck has not named a successor.
- Model providers
- GPTeal gives Merck employees access to OpenAI's ChatGPT, Meta's Llama and Anthropic's Claude (Mar 2025). Merck licenses Google's Gemini Enterprise under the Apr 2026 Google Cloud agreement, and Constellation Research reported that Merck also runs text-to-SQL on Anthropic models through AWS Bedrock. For discovery, Merck built KERMT with NVIDIA, with its code available through NVIDIA BioNeMo and Clara Open Models, and TEDDY with BCG X; the clinical study report platform, co-developed with McKinsey, uses large language models Merck has not named.
- Data and infrastructure
- Merck began moving a substantial part of its IT infrastructure to AWS with Accenture by Nov 2023, including SAP, machine learning workloads, data warehouses and high-performance computing for drug discovery such as AlphaFold on AWS HealthOmics; Dave Williams said then that the move would let Merck use technologies like generative AI. HawkAVI, Merck's manufacturing image analytics platform, runs on AWS services including Amazon SageMaker, S3 and Redshift (Nov 2022). Google Cloud added AI infrastructure and Gemini Enterprise in Apr 2026, Constellation Research describes Merck's setup as multi-cloud, and TechTarget reported that Merck also works with Microsoft and ServiceNow. Merck says GPTeal keeps company data secure from external exposures; Merck has not said where GPTeal itself is hosted.
Platforms and internal tools
GPTealBy Mar 2025
GPTeal is Merck's proprietary generative AI platform that gives employees access to large language models while keeping company data secure from external exposure. Merck staff use GPTeal to draft emails, memos and other documents, and scientists use generative AI to draft regulatory documents that people review before submission. Merck said in Jan 2026 that more than 80% of its workforce uses its proprietary AI platform, which applies large language models, without naming it in that story.
- Users
- More than 50,000 regular users (Mar 2025)
- Models
- OpenAI ChatGPT, Meta Llama, Anthropic Claude
Clinical study report authoring platformBy Jun 2025
Merck's clinical study report platform combines table pre-processing with generative AI authoring to map tables, extract data, style and validate drafts under the oversight of medical writers. A team of more than 80 data science, AI, life science technology and medical experts on three continents produced and submitted the first live reports in 2025, and Merck said in Jun 2025 that it was scaling the platform across its late-phase pipeline by the end of the year. Merck co-developed the platform with McKinsey.
- Models
- Large language models, not named
Gemini Enterprise agentic platformAnnounced Apr 2026
Merck is deploying Google Cloud's Gemini Enterprise and other Google Cloud AI services as an agentic platform across research, manufacturing, commercial and corporate functions under a multi-year agreement worth up to $1 billion. Forward-deployed engineers from Google Cloud and Google DeepMind are embedded with Merck teams to deploy the platform, train foundation models for drug discovery and train Merck's employees, TechTarget reported in Apr 2026.
- Models
- Google Gemini
KERMTDec 2025
KERMT (Kinetic GROVER Multi-Task) is a deep-learning model Merck built with NVIDIA to predict how small molecules behave in the body during lead optimization, pretrained on more than 11 million molecules and fine-tuned for industrial drug discovery tasks. Alan Cheng, Merck's senior director of data science, told Healthcare Brew in Dec 2025 that versions of KERMT run on scientists' desktops inside real-time design cycles, and Merck released the code as open source.
- Models
- In-house, built on the GROVER architecture
TEDDYJul 2025
TEDDY is a family of single-cell biology foundation models that Merck & Co. and MSD (UK) scientists built with BCG X's AI Science Institute, with code and configurations published on Hugging Face.
- Models
- In-house foundation models
HawkAVIBy 2021
HawkAVI is a Merck manufacturing platform on AWS that applies analytics, AI and machine learning to images of rejected drug products to find the root causes of false rejects and tune automated visual inspection machines. Merck engineers reported in Nov 2022 that HawkAVI, a 2021 Manufacturing Leadership Award winner, cut false reject rates by about half across product lines; Merck has not reported on it since.
Vendors and partners
Google CloudApr 2026
Agentic AI platform partner; Gemini Enterprise, cloud services and embedded engineers · Merck's multi-year agreement with Google Cloud is valued at up to $1 billion; Dave Williams told Reuters he expects at least a decade-long partnership, and TechTarget reported that Google DeepMind engineers are embedded too.
Amazon Web ServicesBy 2021
Primary cloud for Merck's IT migration; Bedrock for Anthropic models; hosts HawkAVI · Merck began moving a substantial part of its IT infrastructure, including SAP and machine learning workloads, to AWS with Accenture (Nov 2023).
AccentureBy Nov 2023
Systems integrator for Merck's AWS migration
McKinseyBy Jun 2025
Co-developer of the clinical study report authoring platform · Merck's Jun 2025 release calls the platform co-developed and links to McKinsey's account of the work; Constellation Research also reported that Merck built it with McKinsey.
NVIDIADec 2025
Co-developer of the KERMT small-molecule model; BioNeMo and Clara Open Models distribution
BCG XMay 2024
Generative AI for drug target discovery from omics data; co-developer of TEDDY
OpenAI, Anthropic and MetaBy Mar 2025
Model providers inside GPTeal · Constellation Research reported in Apr 2026 that Merck also uses Anthropic models through AWS Bedrock for text-to-SQL.
Mayo ClinicFeb 2026
AI-enabled drug discovery and precision medicine collaboration through Mayo Clinic Platform_Orchestrate
Tempus AIExpanded Mar 2026
De-identified multimodal data, the Lens platform and the Workspaces computing environment for precision medicine biomarkers, starting in oncology · Tempus announced the expanded multi-year collaboration on Mar 3, 2026, and Rob Davis cited it on the Apr 30, 2026 earnings call; terms have not been disclosed.
Protillion BiosciencesJun 2026
Lab-in-the-loop AI for therapeutic antibody design, under a multi-target discovery collaboration and license agreement · Protillion is eligible for up to $510 million in milestone payments plus an undisclosed upfront payment.
IlluminaJan 2026
Billion Cell Atlas, a CRISPR perturbation dataset for training AI models; Merck is a founding participant
Variational AISep 2025
Enki generative chemistry platform, fine-tuned on Merck data, for two small-molecule targets · Variational AI is eligible for up to $349 million in upfront and milestone payments.
TurbineAug 2025
Simulated cell and virtual tumor models for hard-to-treat cancers
Celonis
Process mining used to decide where Merck applies AI · Constellation Research reported in Apr 2026 that Merck is a long-time Celonis customer and moved its Celonis center of excellence into the Chief AI Office.
Leadership
AI, data and platform leaders
Executives
AI org map
Central IT, digital, data and AI organization under the CIDO; discovery AI built inside Merck Research Laboratories
Merck & Co. places AI strategy with its chief information and digital officer, an Executive Team role held by Gaurav Gupta since Sep 1, 2026 and by Dave Williams from Aug 2020. Merck Research Laboratories, under Dean Li, runs its own data science, AI and genome sciences groups that build discovery models such as KERMT and TEDDY. Constellation Research reported a Chief AI Office in Apr 2026, whose chief, Anton Groom, left for GE Aerospace in Aug 2026, and Merck opened a Hyderabad Technology Hub for cloud, AI and data work in 2025.
Merck & Co. does not publish an AI or data science headcount. Merck's AI work sits in the information technology, digital, data and AI organization under Gaurav Gupta, which includes the Hyderabad Technology Hub opened in 2025, and in data science groups across Merck Research Laboratories that Merck says span every therapeutic area. More than 80 data science, AI, life science technology and medical experts on three continents built the clinical study report platform (Jun 2025), and Google Cloud and Google DeepMind engineers work inside Merck teams under the Apr 2026 agreement.
Reporting lines
Chairman and CEO
Information technology, digital, data and AI
Gaurav Gupta, EVP and CIDO
Executive Team member; the line to the CEO is inferred from Executive Team membership, not published
Chief Technology Office
Ron Kim, SVP and global CTO from 2020, said he would retire at the end of 2025 (Michigan Ross alumni note); successor and reporting line not published
Chief AI Office
Holds the Celonis process mining center of excellence (Constellation Research, Apr 2026); Anton Groom held the Chief AI Officer title until he left for GE Aerospace (CDO Magazine, Aug 2026); successor and reporting line not published, placement inferred
Hyderabad Technology Hub
Opened in 2025 with cloud, cybersecurity, AI/ML, automation, data science and digital product staff; reporting line not published
Merck Research Laboratories
Dean Li, EVP and President
Discovery Chemistry
Robert Garbaccio, VP and Head
Merck & Co. does not publish an AI org chart. Gaurav Gupta's biography says he is a member of the Executive Team and leads global IT, digital, data and AI strategy; Merck has not published where the Chief AI Office, the chief technology officer, the chief data officer or the Hyderabad Technology Hub report. Merck's 2025 Form 10-K says an interim chief information security officer runs the information security program and reports periodically to the Audit Committee, the board, the CEO and senior management.
Named people
Use cases
Merck & Co. describes AI in production in drug discovery, clinical study report writing, clinical trial operations, market access dossiers, manufacturing inspection, supply chain risk, commercial content and field support, recruiting and counterfeit detection, alongside an employee AI platform used by more than 80% of its workforce, but has not published a count of use cases. Merck & Co. has quantified a business return for one: Dave Williams told Reuters in Apr 2026 that Google technology had cut the time and cost of compiling reimbursement dossiers by half. Merck has published no firm-level AI return: Rob Davis said in Sep 2026 that molecular design savings are built into budgets without giving an amount, and the $3.0 billion annual savings target for 2027 is not credited to AI.
| Use case | Area | Status | ROI |
|---|---|---|---|
| GPTeal employee LLM platform Merck employees use GPTeal to draft emails, memos and other documents with OpenAI, Anthropic and Meta models, and scientists use generative AI to draft regulatory documents that people review before submission. | Firmwide | In production | |
| Clinical study report authoring Merck's generative AI platform maps tables, extracts data and drafts clinical study reports under medical writers' oversight, cutting first drafts from two to three weeks to three to four days and the time for a fully human-reviewed first draft from an average of 180 hours to 80, with errors halved. | Clinical development | In production | |
| AI molecular design and optimization Merck uses AI to design and optimize molecules early, which Rob Davis credits with much of the above-average probability of technical and regulatory success in its Phase 1 pipeline; Davis said in Sep 2026 that the work lowers cost and that Merck builds those savings into its budgets, without giving an amount. | Drug discovery | In production | |
| KERMT property prediction KERMT predicts the absorption, distribution, metabolism, excretion and toxicity of small molecules, and Merck's Alan Cheng said in Dec 2025 that versions run on scientists' desktops inside real-time design cycles. | Drug discovery | In production | |
| Target discovery with Mayo Clinic data Merck applies AI, machine learning and its virtual cell technologies to Mayo Clinic clinical, imaging and molecular data to find targets in inflammatory bowel disease, atopic dermatitis and multiple sclerosis. | Drug discovery | Announced | |
| Trial site selection, patient matching and retention Merck uses AI to improve trial site selection, patient matching and retention, with predictive models flagging participants at higher risk of dropping out (Jan 2026). Merck's Impact Report says AI tools from partner companies cut timelines for several site activations and patient matches from about five to eight months to two weeks (Aug 2026). | Clinical operations | In production | |
| Reimbursement dossier compilation Merck uses Google technology to compile the dossiers many countries require before they reimburse a new medicine; Dave Williams said in Apr 2026 that Merck was submitting dossiers built this way and scaling the capability globally. | Market access | In production | Time and cost to compile reimbursement dossiers cut by half with Google's AI tools (Apr 2026) |
| Computer-vision inspection of vials and syringes Merck uses computer vision to inspect vials and syringes for defects in vaccine manufacturing; HawkAVI, its image analytics platform on AWS, cut false reject rates by about half across product lines in a Nov 2022 report. | Manufacturing | In production | |
| Event-based supply risk assessments Merck's generative AI systems produce risk assessments in under 30 minutes after events such as natural disasters or port delays, showing which products and sites are affected. | Supply chain | In production | |
| Field representative chatbot A generative AI chatbot summarizes relevant insights for Merck field representatives and helps them respond to healthcare provider needs in real time. | Commercial | In production | |
| AI across content creation and review Merck uses AI across the life cycle of content for healthcare providers, from conception through medical, legal and regulatory review, to personalize messages. | Commercial | In production | |
| AI-assisted recruiting Merck uses AI-enabled tools to source, screen and match job candidates, with recruiters keeping oversight at every stage. | Human resources | In production | |
| Counterfeit detection from images An AI platform compares images of suspected counterfeit Merck products against expected artwork; Merck used it in field assessments in 2025 and plans to expand it. | Product integrity | In production | |
| Gemini Enterprise agents across functions Merck plans Gemini Enterprise agents for end-to-end research workflows, predictive manufacturing analytics, personalized commercial and patient engagement, and automation in corporate functions. | Enterprise | Announced |
Adoption
Merck spread AI in layers: data science and machine learning in research for more than a decade, AI in manufacturing quality such as the HawkAVI platform on AWS (by 2021), an employee generative AI platform called GPTeal that more than 50,000 employees used regularly by Mar 2025, and agentic AI with Google Cloud from Apr 2026. Merck said in Jan 2026 that more than 80% of its workforce uses its proprietary AI platform and that an enterprise-wide training program covers generative and agentic AI; Ron Kim described self-serve digital courses, monthly generative AI webcasts and boot camps for software developers in Mar 2025, and Google engineers are to help train Merck's 75,000 employees under the Apr 2026 deal. Merck has not published weekly or daily usage, and the 50,000 regular GPTeal users and the 80% share are different measures.
By the numbers
Rollout, first pilot to today
Aug 2020
Dave Williams became Merck's executive vice president and chief information and digital officer.
Nov 2022
Merck engineers described HawkAVI, a manufacturing image analytics platform on AWS that won a 2021 Manufacturing Leadership Award and cut false reject rates by about half across product lines.
Feb 2023
Iya Khalil joined Merck from Novartis as vice president and head of data, AI and genome sciences.
May 2023
Merck published its principles for developing and using AI ethically and responsibly.
Nov 2023
Merck said it was moving a substantial part of its IT infrastructure to AWS with Accenture, including SAP, machine learning workloads and drug discovery computing, partly to use generative AI.
May 2024
BCG X began working with Merck on generative AI that mines omics data for new drug targets.
Feb 2025
Merck Research Laboratories described using large language model agents for data gathering, hypothesis refinement and repetitive analysis.
Mar 2025
Ron Kim, then chief technology officer, said more than 50,000 Merck employees used GPTeal regularly, with access to OpenAI, Meta and Anthropic models.
Jun 2025
Merck reported that its generative AI clinical study report platform cut first drafts from two to three weeks to three to four days and planned to scale it across the late-phase pipeline by the end of 2025.
Jul 2025
Merck and BCG X's AI Science Institute published TEDDY, a family of single-cell foundation models.
Sep 2025
Merck signed a generative chemistry collaboration with Variational AI worth up to $349 million for two small-molecule targets.
Dec 2025
Merck and NVIDIA released KERMT, an open-source small-molecule property model trained on more than 11 million molecules.
Jan 2026
Merck said more than 80% of its workforce uses its proprietary AI platform and that AI runs in discovery, trials, manufacturing and commercial work.
Apr 2026
Merck signed a multi-year agentic AI agreement with Google Cloud worth up to $1 billion to deploy Gemini Enterprise across the company.
Sep 2026
Gaurav Gupta took over as executive vice president and chief information and digital officer, succeeding Dave Williams.
Economics & controls
Spend and ROI
Technology and R&D budget
Merck & Co. does not disclose a technology or AI budget. Merck spent $15.8 billion on research and development in 2025 on sales of $65.0 billion, and the Google Cloud agreement of Apr 2026 commits up to $1 billion over several years to cloud infrastructure, Google engineers and Gemini Enterprise licenses.
What AI itself costs
Merck has not published what it spends on AI. The Google Cloud deal is Merck's only disclosed enterprise AI commitment, valued at up to $1 billion and described by Merck as multi-year; Dave Williams told Reuters he could easily see Merck investing $1 billion over the next several years and expects at least a decade-long partnership. Merck has not disclosed the terms of its research AI deals beyond the potential payments to Variational AI (up to $349 million) and Protillion Biosciences (up to $510 million in milestones plus an undisclosed upfront).
Reported ROI
Merck has published no firm-level return on AI. The one quantified return is in market access: Dave Williams told Reuters on Apr 22, 2026 that Google technology had already cut the time and cost of compiling country reimbursement dossiers by half, with no dollar figure, and that Merck was scaling the capability globally. Rob Davis said on Sep 14, 2026 that AI in molecular design and optimization drives much of the above-average success rate in Merck's Phase 1 pipeline and lowers cost, and that Merck builds those savings into its budgets, without giving an amount. Merck's other published AI figures measure speed and quality rather than return, such as clinical study report drafts cut from 180 hours of work to 80 with errors halved (Jun 2025).
Cost program and headcount
Merck had about 75,000 employees at Dec 31, 2025, about 73,000 of them full time. Merck's 2025 Restructuring Program, approved in Jul 2025, eliminates positions in sales, administration and R&D, costs about $3.0 billion and is expected to save about $1.7 billion a year by the end of 2027, within a multiyear optimization initiative targeting $3.0 billion in annual savings that Merck will reinvest in growth areas. Merck's chief financial officer said on Jul 29, 2025 that the savings will come from productivity across the enterprise and that the program also uses technological advances to raise productivity, but Merck's filings do not tie the cuts or the savings to AI.
Governance and risk
AI principles
Merck published principles for ethical and responsible AI in May 2023: use AI safely with appropriate human judgment, fairly and inclusively, transparently, with ongoing monitoring and retraining, secured under internal AI policies, with disclosure when people interact with an AI system, and with accountability to user feedback. Merck says the principles align with the OECD AI Principles and IEEE's Ethically Aligned Design.
Data boundary and human review
Merck built GPTeal so employees can use outside large language models while company data stays protected from external exposure, and Merck's AI-drafted regulatory documents and clinical study reports are reviewed by people before submission. Merck says its AI-enabled recruiting tools keep human oversight at every stage (Aug 2026).
Disclosed risks
Merck's 2025 Form 10-K says the growing use of AI to automate processes, analyze data and support decisions carries risks from flawed, biased or malfunctioning systems, from laws such as the EU AI Act and the NIS2 Directive, and from leaks of confidential information if employees or third parties use unauthorized open-source or generative AI tools. Merck's 10-K also names a competitive risk if Merck fails to adopt AI in time.
Privacy oversight
Merck's global privacy program evaluates the privacy risks of AI under a cross-functional Privacy and Data Protection Board and a network of more than 300 privacy stewards, and Merck's Chief Ethics and Compliance Officer, who leads the Global Privacy Office, reports to the General Counsel (Impact Report, Aug 2026).
Board oversight
Merck's board named the company's approach to digital technologies, including AI, as a priority topic after its most recent self-evaluation (2026 proxy statement, Apr 2026). Merck's interim chief information security officer reports periodically on cybersecurity to the Audit Committee and the full board (2025 Form 10-K).
Choosing where AI goes
Merck uses Celonis process mining to decide where AI is applied and moved its Celonis center of excellence into the Chief AI Office, Constellation Research reported in Apr 2026.
Recent signals
Sep 14, 2026
Rob Davis told the Morgan Stanley healthcare conference that AI-driven molecular design explains much of Merck's above-average Phase 1 success rate, lowers cost, and produces savings Merck now builds into its budgets.
Sep 1, 2026
Gaurav Gupta became Merck's executive vice president and chief information and digital officer, leading IT, digital, data and AI strategy, succeeding Dave Williams.
Aug 18, 2026
Merck's 2025/2026 Impact Report said AI tools from partner companies cut timelines for several clinical trial site activations and patient matches from about five to eight months to two weeks, and that Merck opened a Hyderabad Technology Hub for cloud, AI and data work in 2025.
Aug 2026
CDO Magazine reported that GE Aerospace had appointed Anton Groom, previously Merck's Chief AI Officer, as its chief data and AI officer; Merck has not named a successor.
Jun 16, 2026
Merck signed a multi-target discovery collaboration and license agreement with Protillion Biosciences for lab-in-the-loop AI antibody design, with up to $510 million in milestone payments.
Apr 30, 2026
Rob Davis told investors that the Google Cloud deal, an expanded Tempus AI collaboration and the Mayo Clinic agreement are steps to accelerate Merck's AI transformation and support productivity.
Apr 22, 2026
Dave Williams told Reuters that Google technology had already halved the time and cost of compiling country reimbursement dossiers, that Merck was scaling it globally, and that he expects at least a decade-long Google partnership.
Apr 22, 2026
Merck and Google Cloud announced a multi-year agentic AI agreement worth up to $1 billion to deploy Gemini Enterprise across research, manufacturing, commercial and corporate functions.
Apr 8, 2026
Merck's 2026 proxy statement listed the company's approach to digital technologies, including AI, among the board's priority topics.
Mar 3, 2026
Tempus AI announced an expanded multi-year collaboration giving Merck its de-identified data, Lens platform and Workspaces computing environment to find precision medicine biomarkers, starting in oncology.
Feb 24, 2026
Merck's 2025 Form 10-K described its growing use of AI to automate processes and support decisions as a risk, citing biased output, the EU AI Act and leaks through unauthorized generative AI tools.
Feb 18, 2026
Merck and Mayo Clinic announced an AI-enabled drug discovery collaboration, Mayo's first of this scale with a global drugmaker, starting in inflammatory bowel disease, atopic dermatitis and multiple sclerosis.
Jan 13, 2026
Illumina launched its Billion Cell Atlas, a CRISPR perturbation dataset for training AI models, with Merck, AstraZeneca and Eli Lilly as founding participants.
Jan 9, 2026
Merck said more than 80% of its workforce uses its proprietary AI platform and described AI in discovery, trials, manufacturing, supply chain and commercial work.
Dec 11, 2025
Merck and NVIDIA released KERMT, an open-source model that predicts how small molecules behave in the body, trained on more than 11 million molecules.
Quick answers
- How does Merck & Co. deploy AI?
- Merck & Co. deploys AI through GPTeal, an in-house platform that gives employees secure access to OpenAI, Anthropic and Meta models, and through a multi-year Google Cloud deal worth up to $1 billion that brings Gemini Enterprise agents to research, manufacturing, commercial and corporate work. Merck said in Jan 2026 that more than 80% of its workforce uses its proprietary AI platform, and Gaurav Gupta took over IT, data and AI strategy as chief information and digital officer on Sep 1, 2026.
- What does Merck & Co. use AI for?
- Merck & Co. describes AI in production in drug discovery, clinical study report writing, clinical trial operations, market access dossiers, manufacturing inspection, supply chain risk, commercial content and field support, recruiting and counterfeit detection, alongside an employee AI platform used by more than 80% of its workforce, but has not published a count of use cases. Merck & Co. has quantified a business return for one: Dave Williams told Reuters in Apr 2026 that Google technology had cut the time and cost of compiling reimbursement dossiers by half. Merck has published no firm-level AI return: Rob Davis said in Sep 2026 that molecular design savings are built into budgets without giving an amount, and the $3.0 billion annual savings target for 2027 is not credited to AI.
- Who leads AI at Merck & Co.?
- AI at Merck & Co. is led by Gaurav Gupta (Executive Vice President and Chief Information and Digital Officer); Dave Williams (Executive Vice President and Chief Information and Digital Officer, Aug 2020 to Aug 2026); Asheesh Chhabra (Vice President and Chief Data Officer). Ownership of AI on the org chart: EVP and CIDO Gaurav Gupta, on the Executive Team (since Sep 2026).
- What AI platform does Merck & Co. use?
- GPTeal, a platform Merck & Co. built to give staff secure access to outside large language models, carries everyday writing work; a Google Cloud deal worth up to $1 billion adds Gemini Enterprise agents (Apr 2026).
- Which AI models does Merck & Co. use?
- GPTeal offers OpenAI, Anthropic and Meta Llama models (Mar 2025); Merck adds Google's Gemini, Anthropic models on AWS Bedrock, and its own KERMT and TEDDY research models.
- How many Merck employees use AI?
- Merck said in Jan 2026 that more than 80% of its workforce uses its proprietary AI platform; Ron Kim, then chief technology officer, put regular GPTeal users above 50,000 in Mar 2025.
- How far has AI spread at Merck & Co.?
- Merck & Co. sits at Scaling: Merck said in Jan 2026 that more than 80% of its workforce uses its proprietary AI platform, AI runs in production across drug discovery, clinical documents, trial operations, market access, manufacturing, supply chain and commercial work, and chief information and digital officer Gaurav Gupta, an Executive Team member, owns the AI strategy. Merck & Co. falls short of Embedded because it has not published how often employees use the platform or how many use cases run in production, and its only quantified AI return covers one process, reimbursement dossiers compiled in half the time and cost (Apr 2026), with no firm-level financial effect.
Open questions
- Is the proprietary AI platform that more than 80% of Merck's workforce uses (Jan 2026) GPTeal, and how many employees use it each week?
- What has the reimbursement dossier capability saved in dollars as it scales globally, and what does AI return across Merck as a whole?
- How large are the molecular design savings Rob Davis says Merck builds into its budgets?
- Who succeeded Anton Groom as Chief AI Officer and Ron Kim as chief technology officer, and where do the Chief AI Office, the chief data officer and the Hyderabad Technology Hub report under Gaurav Gupta?
- How many AI use cases does Merck run in production, and how many AI, machine learning and data science staff does it employ?
- What does the Google Cloud agreement cost each year, and how will Gemini Enterprise sit alongside GPTeal?
- Which models does GPTeal offer today beyond the OpenAI, Meta and Anthropic models named in Mar 2025, and where is it hosted?
- Did the clinical study report platform scale across the late-phase pipeline by the end of 2025 as planned, and what has it saved in cost or medical writer capacity?
- What business return do Merck's deployed tools with no published figures deliver, such as computer-vision inspection, the field representative chatbot, AI content review, trial site selection and supply risk assessments?
- Does any part of Merck's $3.0 billion optimization target or the 2025 job cuts come from AI?
- How will Merck's AI strategy change under Gaurav Gupta, and is Iya Khalil still in her role?
Sources
Merck & Co.
- Merck and Google Cloud partner to accelerate agentic AI enterprise transformation (Apr 22, 2026)
- Merck expands internal generative AI solutions for clinical study reports (Jun 25, 2025)
- Merck and Mayo Clinic AI-enabled drug discovery collaboration (Feb 18, 2026)
- 5 ways we're transforming artificial intelligence into impact (Jan 9, 2026)
- Our AI model KERMT is helping to advance drug discovery (Mar 19, 2026)
- Our researchers incorporate LLMs to accelerate drug discovery and development (Feb 3, 2025)
- Data science and artificial intelligence at Merck Research Laboratories
- Merck's commitment to developing and using AI ethically and responsibly (May 2023)
- Executive team
- Gaurav Gupta, EVP and chief information and digital officer, biography
- Robert M. Davis, chairman and CEO, biography
- Dean Y. Li, president, Merck Research Laboratories, biography
- Dave Williams, chief information and digital officer, biography (Aug 2020)
- Company fact sheet
- Purpose for Progress: Merck Impact Report 2025/2026 (Aug 2026)
Filings and investor materials
Partners
- Google Cloud press corner: Merck and Google Cloud partnership (Apr 22, 2026)
- AWS for Industries: Merck builds HawkAVI platform on AWS (Nov 2022)
- Accenture: AWS and Accenture help Merck use cloud technology (Nov 29, 2023)
- BCG X: Meet TEDDY, a new family of foundational AI models (Jul 2025)
- BCG: generative AI collaboration with Merck (May 30, 2024)
- Variational AI: collaboration with Merck (Sep 23, 2025)
- Turbine: research collaboration with MSD (Aug 4, 2025)
- McKinsey: with gen AI, Merck and McKinsey transform clinical authoring
- Tempus: strategic collaboration agreement with Merck (Mar 3, 2026)
News
- Business Insider: How pharmaceutical companies are training their workers on AI (Mar 2025)
- NAM: Pharma embraces AI, on the Business Insider report (Mar 27, 2025)
- Reuters via BNN Bloomberg: Merck to partner with Google Cloud on AI initiatives (Apr 22, 2026)
- TechTarget: Merck, Home Depot tap Gemini Enterprise for AI agent development (Apr 23, 2026)
- R&D World: Merck's $1 billion Google pact to provide backbone for AI projects (Apr 22, 2026)
- Constellation Research: Merck inks Google Cloud agentic AI deal worth up to $1 billion (Apr 22, 2026)
- Healthcare Brew: Merck and NVIDIA release new drug discovery model (Dec 11, 2025)
- CXO Digital Pulse: Gaurav Gupta joins Merck as EVP and CIDO (Jul 23, 2026)
- CDO Magazine: Merck names Gaurav Gupta EVP and CIDO (updated Aug 18, 2026)
- Endpoints News: AI leader Iya Khalil switches from Novartis to Merck (Feb 2023)
- MedCity News: For Merck, AI puts humans in the right room (Dec 1, 2025)
- CDO Magazine: GE Aerospace appoints Anton Groom chief data and AI officer (Aug 2026)
- Michigan Ross alumni class note: Ron Kim (2025)
- GEN: Merck and Protillion launch AI drug discovery collaboration (Jun 16, 2026)
- BiopharmaTrend: Illumina launches Billion Cell Atlas with AstraZeneca, Merck and Lilly (Jan 13, 2026)
Data confidence
- Confirmed
- Stated in a primary or reputable public source.
- Estimated
- Derived or approximated from public reporting.
- Unconfirmed
- Reported, but not yet corroborated.
Merck & Co. profile · Enterprise AI diffusion report by Imagine · facts checked against their sources on Nov 1, 2026
