Research
We investigate how AI can accelerate work while staying aligned with how teams already operate.
Workflow-native assistance
We study where AI helps most inside existing tools — drafting, review, and handoff — instead of forcing teams onto new platforms.
Trust and reliability
Research on grounding, evaluation, and failure modes so assistance stays useful when stakes are high.
Human–AI collaboration
Understanding how people edit, accept, and reject suggestions so products stay in service of judgment, not the other way around.