When everyone speaks the same language: Boehringer Ingelheim speeds up knowledge sharing with Azure OpenAI Service

How can expert knowledge be used efficiently? Typically, someone comes up with a question and searches until they’ve found their answer. In the pharmaceutical industry, coming up with the answer quickly can save many lives. Companies in this industry store their expert knowledge in numerous internal repositories and databases as well as drawing on external sources. Trawling through these stores of information to gain insights for researching, developing, and providing medications could be extremely time-consuming. Now, with the combination of its iQNow platform and Azure OpenAI Service, Boehringer Ingelheim has made it possible for researchers to access information many times faster than before, leaving them with more time and capacity to develop drugs.

The challenge: The sheer number of documents makes it hard to find the right information

There’s no time to lose when it comes to developing medications—it’s imperative that lifesaving drugs are made available to those who need them as quickly as possible. This is why time is the key metric in many process steps: development, clinical trials, and approval by the responsible authorities. As pharmaceutical companies must observe a clear set of rules and requirements, they have only a limited influence on the efficiency of many process steps. During development, however, there are many workflows that can be speeded up by sharing knowledge more efficiently. But that’s easier said than done.

“Over the lifecycle of any one of our products, a single department will produce more than 60,000 documents of every conceivable format—and that can involve anywhere up to 5,000 experts. These documents contain knowledge that is extremely valuable to other projects as well,” says Michael Schorpp, Senior Expert Knowledge and Learning Architecture & Strategy at Boehringer Ingelheim. “Moreover, there are a great many external information sources that are of vital importance to our research and development activities, including scientific and clinical studies, patents, and conferences. Many of our own experts possess a wealth of knowledge that’s stored only in their head.” In the past, it took ages to sift through this massive trove of collective knowledge to find those insights relevant to a particular drug. “It wasn’t merely a question of finding details of people or documents in the various databases, but rather tracking down exactly the right person and the right document. The next challenge was to actually understand and apply the knowledge once it had been found,” Schorpp says. This was because different departments might use different vocabulary, abbreviations, or process terms: for example, HCP can stand for healthcare provider in one document and for host cell protein in another. Without context, there’s no way to know which meaning is intended.

In other words, Boehringer Ingelheim faced the following challenges: make the company’s immense store of knowledge searchable, give it a uniform language, and make it easier for researchers to extract this knowledge. The mission became to develop a knowledge management platform that would overcome all these challenges.

The solution: Knowledge management aided by Azure OpenAI Service

Boehringer Ingelheim began by laying the foundation for a consistent, long-term vision. “Our aim is to establish interfaces to connect all the relevant information sources—both external and internal repositories—together, and then add an expert search function,” says Aleksandar Kapisoda, Senior Knowledge & Learning Scientist at Boehringer Ingelheim. This is how the company established a semantic search engine: the iQNow knowledge management platform. In contrast to a conventional keyword-based search engine, iQNow also takes into account the context and intentions of the inquiry. This alone proved a massive help, because it meant that employees found the right experts and documents considerably faster. The solution also relies on SharePoint and Azure Active Directory to issue and manage permissions.

“Thanks to Microsoft solutions for data management and identity and access administration like SharePoint and Azure Active Directory, we’ve been able to incorporate highly secure access management into iQNow. That was a fundamental part of our solution’s success—after all, we’re working with very sensitive, often confidential data.”

Aleksandar Kapisoda, Senior Knowledge & Learning Scientist, Boehringer Ingelheim

But finding knowledge isn’t the same as using it. “We wanted to go even further because we found ourselves faced with the typical search engine dilemma,” says Matthias Negri, Senior Data Scientist at Boehringer Ingelheim. “While the platform did suggest lots of appropriate documents, I still had to choose which one I wanted to read. Then I had to comb through it to find the knowledge I was looking for. This is where artificial intelligence (AI) comes in.” Through Azure OpenAI Service, Boehringer Ingelheim opened up a dialogue between knowledge and employees: by way of a secure interface with Azure OpenAI Service, employees can direct their questions to documents they have selected or that have been suggested to them following an AI-assisted search. In this way, employees receive precise, fact-based search results. Especially considering the dozens of internal and external knowledge databases, some of which house documents that are several hundred pages long, this means even complex searches can now be performed with a minimum of effort. Documents are summarized and insights are drawn from the information available. Boehringer Ingelheim’s solution has speeded up searches for knowledge significantly. The efficiency gains were such that within 70 business days, the global corporate group managed to save almost 150,000 work hours. This time flows directly into the development of medications that are then available sooner to the people who need them.

Boehringer Ingelheim is also benefiting from the high compatibility of Microsoft solutions. iQNow is integrated directly into Word and PowerPoint—the two programs in which the lion’s share of drug development documents are written. “We’ve named this plug-in iQMe,” Kapisoda says. “For each document, it provides all metadata, including the author’s name, topic areas, and a brief summary. It also offers users feedback relating to how easy it was to find the document. Users immediately receive suggestions for how to improve the document’s traceability. What’s more, our iQMe plug-in can be integrated directly into these Office solutions like a small, virtual assistant—similar to Microsoft Copilot only highly specialized in our internal rules and processes. This all means that our iQNow platform is always personalized to each user and integrated into the documents.” Indeed, Schorpp says that iQNow’s high level of integration into existing business processes is the key to the platform’s success:

“As iQNow is so compatible with the Microsoft Office world, weaving our solution into our existing workflows was a very organic process. So instead of feeling like it doesn’t belong, it’s easy to learn and use. That’s why it’s been so well received.”

Michael Schorpp, Senior Expert Knowledge and Learning Architecture & Strategy, Boehringer Ingelheim

iQNow continues to grow: the next step will be to add more items from the Azure Cognitive Services portfolio as plug-ins to enable users to mine knowledge from tables, images, and recorded speech. The mobile version of iQNow launched in October 2023. In the future, it will provide Boehringer Ingelheim colleagues in the field, for example, with straightforward access to the company’s combined knowledge in real time.

The vision for tomorrow is a clear one: instead of dealing with individual documents, employees will enjoy a dialogue with the accumulated medical knowledge contained in all connected documents. Much further down the line, this knowledge sharing will be even more immediate thanks to a speech model. Then the question for iQNow will be: “How can we harness the full store of medical knowledge that society has amassed?” And iQNow will answer: “A good solution is to enlist the help of artificial intelligence.”

“We want to enable straightforward communication between our scientists and research documents. They pose a question and instead of receiving a list of documents, they get a fact-based answer. What makes this natural, human way of communicating possible is Azure OpenAI Service.”

Michael Schorpp, Senior Expert Knowledge and Learning Architecture & Strategy, Boehringer Ingelheim