Enterprise AI Is Answering the Wrong Question

Businesses are investing billions in artificial intelligence, yet its true potential often remains untapped because the technology isn't deeply integrated into existing workflows. Bhavin Shah, SVP & GM at Moveworks, argues that companies are frequently focusing on the wrong problem: developing the technology itself rather than addressing the challenges of its effective implementation within organizational structures.

The issue stems from differing priorities between AI providers and enterprise CIOs. Providers aim to make models more capable and generalizable, while businesses need AI that aligns with their approval chains, compliance requirements, and organizational setup. This disconnect means AI applications may appear compelling in demos but fail in production if not designed for specific enterprise needs.

Today's enterprise AI solutions, such as copilots or chatbots, primarily enhance user interface usability without altering underlying processes. They can assist with information retrieval or reporting, but actual execution, like securing approvals, still requires human intervention. Fully automated workflows would necessitate AI agents capable of understanding next steps, approval policies, and exceptions, and efficiently handing off tasks to humans.

Shah emphasizes that true delegation requires embedding AI directly into the workflow systems, not just layering it on top. When AI operates within the organization's own rules and structures, it can solve core problems more effectively. For instance, a distribution center general manager could leverage an AI assistant to anticipate parts shortages, manage orders, and secure approvals automatically, freeing the manager for strategic decision-making.

An AI system that understands the organization's structure and operations enables effective governance. Embedding governance models into the AI architecture, rather than treating them as post-deployment constraints, allows AI to act more autonomously and securely. This integrated approach is key to transforming AI from a mere information provider into a true work completer.