Scaling agentic AI pilots across the enterprise

Details

Enterprise agentic AI adoption is wide but shallow: while 80% of Fortune 500 companies have experimented with AI agents, most organizations are still running isolated pilots rather than coordinated, enterprise-wide deployments. According to Arun Chandra, COO at NiCE, the core problem is that most organizations layer AI onto existing workflows without first examining whether those workflows are worth preserving. "The last thing you want to do is to apply AI on an outdated or an inefficient workflow," he says. The prerequisites for real scale are data architecture and system integration, not just model selection. Chandra argues that agent effectiveness is directly proportional to the quality and completeness of the context they can access — fragmented knowledge bases and disconnected back-end systems produce fragmented agent behavior. Organizations need to define explicit business objectives (revenue growth, cost reduction, or other measurable outcomes) before deploying agents, rather than experimenting for its own sake. Organizational fragmentation is a secondary risk: teams building isolated agent systems without a coordinated architecture create new silos that mirror the data silos they were supposed to solve. Governance, privacy, and change management become load-bearing concerns as agents move from advisory to action-taking roles. Note: this piece is sponsored content produced by MIT Technology Review's custom content arm in partnership with NiCE, and promotes a NiCE webcast. The practical guidance is sound but lacks independent verification or original research.