At Sierra Ventures, conversations with enterprise technology leaders help us understand how AI is moving into production inside complex organizations. Mark Fernandes recently spoke with Dave Williams, Chief Information and Digital Officer at Merck, about where AI is already creating measurable value inside the business and why scaling enterprise-wide transformation remains a journey.
Dave’s message was clear. AI will deliver value, but most organizations are underestimating the amount of process, data, and organizational change required to get there.
Drug discovery is already seeing measurable gains
One of the clearest areas of impact at Merck is inside drug discovery, where AI is helping researchers accelerate how new medicines are identified and developed through a combination of wet lab experimentation and in silico modeling.
As Dave explained, Merck has increased productivity in parts of its preclinical candidate workflow over the last several years. In one recent example, a program reached clinical trials roughly a full year earlier than it would have historically.
“The quicker we can get things through into clinical and out of clinical trials and into manufacturing, the quicker our patients get these new medicines.”
This mirrors what we continue to hear across enterprise environments. AI is creating the strongest near-term value inside operational workflows where speed and iteration directly impact outcomes.
Production deployment matters more than pilots
Merck is also applying generative AI to regulatory and commercialization workflows that have historically required significant manual effort. One example Dave highlighted was the company’s Health Technology Assessment process, where teams prepare highly technical dossiers required for market access in different countries.
“These are hundreds of pages. They’re heavily technical, a lot of research.”
By applying generative AI to that process, Merck has cut both the time and cost required to produce those dossiers in half. More importantly, those systems are already being used in live production environments across multiple markets.
This shift from experimentation into operational deployment continues to separate real enterprise adoption from isolated AI pilots.
Organizational change is the real bottleneck
One of the strongest signals from the conversation was that the hardest part of AI adoption is not selecting models or infrastructure. It is an organizational change.
“We’re asking our employees to make sure you run the business every day, stay current with technology that’s changing faster than it ever has, and also rethink how you work.”
He also pointed to systems and workflows as major factors in transformation efforts, particularly between research, clinical development, and manufacturing, where there’s an opportunity for information and processes to work together.
This mirrors what we continue to hear across large organizations: AI adoption becomes significantly harder when workflows and data are optimized around departments instead of the full enterprise process.
The hype cycle is ahead of enterprise reality
Dave was also candid about the expectations surrounding AI adoption.
“I have zero doubt GenAI and agentic AI are going to transform how every company and every industry works,” he said. “It’s just going to take a little longer.”
According to Dave, many organizations are still working through disconnected data environments, legacy operational structures, and the challenge of simultaneously running and transforming the business.
“The reality is transformation in the enterprise is a heavy lift,” he explained.
Build where you differentiate and simplify everything else
Merck’s approach to build versus buy is straightforward. Capabilities tied directly to competitive differentiation, particularly in drug discovery, are often built internally alongside scientists and domain experts. More generalized enterprise infrastructure continues to come from external vendors.
“I don’t have any desire to vibe code my ERP or my CRM.”
For startups selling into large enterprises, Dave’s advice was equally direct: solve clear problems, integrate into existing environments, and reduce complexity rather than adding to it.
Organizations are not looking for more disconnected AI tooling. They are looking for systems that simplify operations and fit cleanly into existing workflows.
Watch the video:
