From tool to knowledge broker: How AI can change our work

“The historians of the future will divide history into a time before and after ChatGPT.” I read this sentence in the Süddeutsche Zeitung newspaper around three years ago and it has stayed with me to this day. After all, ChatGPT truly ushered in a new era: We suddenly had access to a program that could answer questions(almost) like a human rather than a robot.

Today, many AI tools, especially those based on large language models (LLMs), are being used in various professions, where they are primarily used to provide support and boost efficiency. For example, the scientist who asks an AI chat assistant to summarize a study in bullet points to help decide whether the study is relevant to their research. Then there’s the journalist who feeds research findings into Claude Sonnet and has it generate a first rough draft. And then there’s the computer programmer who uses an LLM application to check code for errors instead of examining it painstakingly line-by-line.

We also use these tools at our company. Alongside myGPT – our internal version of ChatGPT – we use other generative AI applications that are tailored to our needs. As a science and technology company, we wanted to better understand how these models affect our world of work. After all, to shape developments rather than simply respond to them, you need to understand how processes and behaviors are changing.

To this end, we have built a partnership with the Technical University of Darmstadt. In an initial joint research project, we observed how research and development colleagues from the Electronics business sector at Merck KGaA, Darmstadt, Germany interact with an AI application we developed internally. Our findings indicate that LLMs can influence knowledge brokerage between employees

in various ways. We observed these three possible forms:

• Jog our memory: They help us retrieve specialist knowledge quickly, such as details of previous projects, internal terminology, or regulatory requirements.
• Prepare for dialogue: Before our scientists engage in dialogue, the concise summaries
  drawn up by LLMs can create a shared foundation of knowledge that makes it easier to address in-depth questions. This shared foundation enables deeper discussions and promotes cross-disciplinary collaboration.
• Simulated dialogue partner: In complex projects, employees use LLMs to structure their
  thoughts and generate new ideas on the basis of existing knowledge. 

To put it simply, AI is evolving from a passive technology into a knowledge broker.

This project is not the end – we will continue to conduct research in this field to contribute valuable insights for science, industry and of course for our employees.

AI builds bridges between fields of knowledge

Experts from many different fields work together at Merck KGaA, Darmstadt, Germany: from materials research to biotechnology or high-tech materials production. This diversity is our strength. But for a long time, it also brought with it a major challenge: How can people from completely different disciplines with different ways of thinking reach a common understanding? 

As an example, let us take a chemist who is working on new materials for our customers in the semiconductor industry. To turn this development into a success, the chemist needs input from our application engineers who know how our materials actually work in our customers’ production facilities.

In the past, colleagues with experience in both worlds were often needed to act as a broker and build a bridge between the areas of chemistry and application technology – if they were available.

AI is now opening up new possibilities here. To stay with our example: The chemist could use an AI tool to translate between the two fields. They describe their approach in the familiar language of chemistry; the tool then translates this information into the language of application technology, explains the customer-relevant aspects to the chemist and specifically prepares the right questions for the discussion with the application engineer. As such, the tool acts as a translator and builds a bridge between two worlds of knowledge, enabling the specialists to enter into dialogue on equal terms.

Critical approach to AI: In the end, we decide

To enable everyone at our company to get the most out of the possibilities offered by AI, we are making targeted investments in the AI competence of our entire workforce: With comprehensive training courses, access to the latest AI tools, and an open learning culture, we want to make AI accessible for all – irrespective of department or the level of prior experience.

Our goal is to use AI in a sensible way and to handle its results in a critical manner. This means that AI provides inspiration, analysis and suggestions – but our employees are still in charge and critically assess what the AI tool delivers. In this context, I like to think of AI as being like a car navigation system: It knows a lot of roads, suggests routes and issues warnings about traffic jams. But in the end, we decide where to drive to and whether to follow the proposed route or to find our own way.

 It remains to be seen whether historians will one day divide history into a time before and after ChatGPT, but here at Merck KGaA, Darmstadt, Germany, we certainly do.