Amgen builds agent in 6 weeks with Copilot Studio to help accelerate R&D

Identifying a need

Like many companies, Amgen struggled to make its vast corpus of knowledge and resources easily accessible to the teams that work together to develop new therapies. “A drug developer might search our intranet for past trial designs or resources about a specific molecule,” says Bryan Yee, Executive Director of R&D Knowledge and Learning at Amgen. “They’d get a set of links to hunt down and sort through—PowerPoints, PDFs, articles, reports, SharePoint folders. This can take hours, days, even. We wanted to make this a more intelligent, intuitive process, the same way you’d ask a colleague a question.” Yee smiles. “Yes, it’s complex. But we knew we needed to figure it out.”

Building a solution

Amgen is an early adopter of Microsoft 365 Copilot. This made Copilot Studio, a low-code platform that uses generative AI and agentic workflows to build intelligent agents, an obvious place to start. Working with the Amgen technology team, Yee and his knowledge management crew built an agent for the R&D team that could reason over large volumes of information. The entire process took about six weeks. The goal was to use AI to sort, surface, and analyze Amgen’s existing thoughtfully curated resources, with a natural language interface. This led to the development of Catalyst Copilot, an agent embedded with Catalyst, Amgen’s knowledge management platform.

“As with all AI systems, the most important step for us was first having a carefully curated collection of information,” Yee says. “Adding Copilot to that made our ever-growing body of knowledge instantly more accessible and is a key strategy in how we are advancing the way we learn in real-time.”

Accelerating the path to discovery

With Catalyst Copilot, drug developers are tapping into the collective intelligence of the organization faster than ever. They are shortening the cycle of discovery by more easily finding the information and people needed to answer key questions that advance potential new medicines forward. Catalyst Copilot is helping to remove barriers to information-gathering, including the age-old problem of not knowing the right question to ask. Catalyst Copilot’s ability to reason and infer from multiple fact sets enables faster, more accurate conclusions. The Q&A–style chat inputs also help the knowledge management team understand intent, so they more precisely curate content.

Advancing for the future

Having seen the benefits of custom agentic workflows, Amgen looks to incorporate more in other areas of the business. “With Copilot Studio, we have a common platform to rapidly deploy solutions from the low-code and no-code environments, and we can also extend with more complex configurations as needed,” says Skellenger. “We see a future where there’s a single user experience that’s routing people to the agents that they need.”

"One of the key metrics for measuring AI success at Amgen is enabling better science at scale. This includes automating parts of the scientific process and distilling data that leads researchers to deep insights,” says Bruich. “Since Microsoft is a trusted technology collaborator, we can experiment quickly to see where AI can have a big impact.”

Discover more about Amgen on Facebook, LinkedIn, X/Twitter, and YouTube.

Bryan Yee, Executive Director of R&D Knowledge and Learning, shares his perspective on how digital agents boost innovation

Every organization struggles with knowledge and resources scattered around digitally. In Research and Development, this could be a serious impediment to innovation.

Removing this impediment is my team’s job. We manage information architecture for R&D. How can I quickly and intuitively connect drug developers to information that leads to breakthrough insights? The time it takes someone to find the right information, the right person to talk to, the right resource, is white space. We needed to close those gaps to help get therapies to patients faster.

We used Copilot Studio to build an agent in just six weeks that’s embedded in our knowledge platform, Catalyst. We’d already done the critical hard work of curating and organizing our knowledge—lessons learned, articles, presentations, and the like. Now we added AI to help analyze, classify, and apply reason over that information. With our agent, Catalyst Copilot, any one of the many roles that work together to bring potential new medicines to patients can now ask a question in natural language and get an answer. It’s like asking the right colleague a question at the right time, when they have all the right information at their fingertips. This approach has the potential to fundamentally change how R&D approaches knowledge management and how we learn as an organization. Now instead of starting at “Where is the information I need?” our team can go straight to “What do we already know?”

Copilot is helping remove obstacles and is bringing to life new ways to tap into our collective intelligence across Research and Development. This is the future of how we learn. That’s innovation.