Krishna C. Mukherjee reflects on early AI assistants and lessons for today’s copilots

Mukherjee cites his later projects—such as the Intelligent Filing Manager for Wolters Kluwer, Bloomberg’s BVAL pricing service, and the integration of Microsoft Nuance conversational AI at FIS—to illustrate design patterns that have endured. He stresses separating knowledge (rule bases) from execution engines, delivering services centrally like early cloud models, and structuring interactions through intent routing and slot filling. For modern deployments, he recommends hybrid systems that combine large language models with rule‑engine governance, explicit permission checks, audit logs, and human‑in‑the‑loop controls to prevent silent, high‑speed errors.

The interview also touches on measurement discipline, urging leaders to record baselines before automation and to define ownership of ongoing metrics. Mukherjee’s view of “quiet force” AI underscores the need for visible controls as intelligence becomes embedded infrastructure.