Kirin Holdings and Tokyo-based AI firm GenerativeX have built a research environment that puts AI agents directly inside the daily workflow of scientists at select Kirin laboratories. The system, launched this year, is designed to support the full research process - hypothesis generation, information searches, brainstorming, documentation, and knowledge sharing - without requiring researchers to stop and switch between separate tools.
The target problem is what the companies call "fragmented thinking": researchers constantly interrupting their own line of reasoning to search for literature, dig up past materials, or organize notes. Each interruption breaks focus, and creative work suffers. The new setup keeps AI present in the research flow at all times, rather than having researchers call it up only when they need a specific answer.
How the system works
Kirin is responsible for designing the research concept and applying it in real research settings, while generative AI business GenerativeX provides the technology and system implementation. The system is in use at several Kirin labs, with future expansion across the group under consideration.
Perhaps the most relevant distinction for technical readers: the system does not try to automate around the researcher. It is designed to be a collaborative layer that extends their thinking. The companies say this goes beyond typical AI adoption in corporations, which has largely focused on operational efficiency - document creation, search, and similar tasks.
The AI agents are built to stay constantly present within the research workflow, an architecture choice that differs from the more common pattern of a text box, where AI is invoked only when a user decides to ask it something.
Development approach
The system was built using an agile process, with direct feedback from researchers shaping which features and functions made it into the actual product. Researchers involved in the development process recognized the system as a tool suited to their own research activities. The development team aligned what researchers actually wanted to use against what the AI technology could deliver.
Knowledge sharing as a design goal
One of the more interesting pieces of the system is that it doesn't just support the individual researcher. Hypotheses, discussions, and exploration histories are accumulated and made available for use across the organization. This moves knowledge from an individual asset to a shared organizational one, which could have practical value for large R&D organizations, with their researchers spending significant time tracking down what others have already explored.
Why this matters for IT and development professionals
The idea of AI agents as persistent collaborators inside a domain-specific workflow - not as a chatbot you tab over to - is a design pattern worth watching. Most internal AI deployments still follow the "ask and receive" model, but Kirin and GenerativeX have chosen to embed the AI agent directly in the research process, creating a different kind of interaction entirely. The system is still being rolled out and expanded across the company, with the companies planning to add functionality that recognizes an individual researcher's expertise across their research themes. For professionals building AI systems, this is an interesting case study in how careful AI integration goes beyond process automation and into the actual intellectual work of research, a place where AI has been slow to penetrate.
Your membership also unlocks:
