HBR: only 6% of companies fully trust AI agents to handle core business processes

A Harvard Business Review Analytic Services survey of 603 business and technology leaders (July 2025) finds a wide gap between rapid experimentation with agentic AI and confidence in letting it run core processes—only 6% fully trust AI agents for critical workflows even as 9% have fully deployed them and half are piloting or exploring use cases. Adoption momentum is high (86% expect investment to rise), but readiness is lagging: just 20% say infrastructure is fully ready, 15% for data, 12% for governance, and only 27% qualify as leaders on a composite readiness index; top barriers are cybersecurity/privacy (31%), data output quality (23%) and immature processes (22%). Firms are responding by prioritizing enterprise orchestration (8% implemented, 74% working/planning), training (44%) and governance guardrails (39%), and 72% say benefits outweigh risks—suggesting that improved connectivity, data quality and change management will determine whether AI agents move from limited pilots to trusted operators of mission‑critical workflows.

Analysis

Harvard Business Review Analytic Services surveyed 603 business and technology leaders in July 2025 and found only 6% of companies fully trust agentic AI to run core processes, even as 9% report full deployment and 50% are piloting or exploring use cases. Eighty-six percent expect investment in agentic AI to increase over the next two years while only 10% have decided not to proceed, indicating strong adoption momentum despite limited willingness to delegate mission‑critical work.

Readiness shortfalls are the central constraint: 20% say infrastructure is fully ready, 15% say data and systems are ready, and 12% report risk and governance controls are in place, producing a composite of 27% leaders, 50% followers, and 24% laggards. Security and privacy (31%), data output quality (23%) and immature processes or tech limits (22%) are the top barriers, and organizations report realized productivity, cost and CX gains are measurable but generally below expectations.

Responses concentrate on enterprise orchestration (8% implemented, 74% working/planning), training/upskilling (44%) and guardrails (39%), with 72% saying benefits outweigh risks. The report warns of "garbage in, garbage out" outcomes if data, governance and architecture lag, which would erode trust and slow conversion from pilots to core automation.

Key Decisions for Investors

• Favor exposure to vendors and integrators that provide enterprise orchestration, connectivity and governance capabilities because 74% of firms are working or planning such projects and leaders are likely to capture early enterprise budgets
• Monitor objective readiness metrics—infrastructure (20% fully ready), data systems (15%), and governance controls (12%)—and reduce conviction in end‑user firms showing material deficits in these areas given their correlation with stalled deployments
• Allocate to companies demonstrating concrete upskilling and responsible‑AI programs (44% and 39% prioritizing these), since workforce readiness and guardrails materially increase the odds of safe, scalable adoption
• Hedge or limit exposure to firms promising immediate core‑process automation without demonstrable data architecture and governance, because realized benefits are currently below expectations and security/privacy risks (31%) could trigger setbacks