Accenture and Wharton Warn: AI Agents Are Spreading Faster Than Anyone Can Govern Them

The boardrooms are buzzing with autonomous agents. AI that books meetings, writes code, handles customer complaints, manages supply chains, and even makes pricing decisions is no longer a future possibility — it's a present reality spreading through corporate America at a pace that is, according to a new landmark report, outrunning the humans supposed to be overseeing it.

The report, titled The Age of Co-Intelligence: How Humans, AI Agents and Robots Are Redefining Value, is a joint effort from Accenture's Global Products practice and Wharton's AI and Analytics Initiative. Released March 26, 2026, it synthesizes task-level data across 18 industries covering more than 120 million workers and delivers a warning that cuts through the hype with unusual clarity.

"Intelligence may be scalable, but accountability is not." — The Age of Co-Intelligence report, Accenture & Wharton

The Scale Is Staggering

The numbers are hard to sit with. Using ONET and Bureau of Labor Statistics data, Accenture researchers found that more than 50% of working hours across the American economy are now in play — subject to reshaping by the approximately 60 digital and physical AI agents studied. In banking and capital markets alone, the share of hours impacted by digital agents exceeds 45%.

Nearly three-quarters of knowledge workers are now using AI — frequently through unsanctioned, bring-your-own tools, a phenomenon the report calls "shadow AI." By 2028, roughly a third of all enterprise applications are expected to embed agentic capabilities.

For a $60 billion company modeled in the report, Accenture and Wharton estimated approximately $6 billion in potential annual revenue growth from deploying agentic AI at full maturity, alongside $1.7 billion in annual productivity gains. The catch: by 2028, roughly one-third of those productivity gains would show up not as direct cost savings, but as "capacity freed" — hours that need to be deliberately redirected toward higher-value work, or they simply evaporate.

The Accountability Gap

The report's most provocative argument is that the smarter AI gets, the more consequential human leadership becomes — not less. This runs counter to a popular narrative that better AI means less need for human judgment.

James Crowley, Accenture's Global Products Industry Practices Chair and co-author, put it plainly: "We like to say humans in the lead, not in the loop." The distinction matters enormously. "In the loop" implies passive oversight — a human who checks in occasionally. "In the lead" means someone is actively setting strategy, making decisions about what matters, and owning outcomes when agents go wrong.

When agents aren't properly overseen, errors compound at scale. The report offers concrete examples: one agent's hallucinated inventory figure causing downstream agents to massively overorder stock; a customer service agent telling a customer a problem is solved when it isn't, with no human catching the error. These aren't hypotheticals — they're the predictable failure modes of autonomous systems running without adequate governance.

"Shadow AI" and the Governance Crisis

Perhaps the most alarming finding is that AI agents are already spreading "rapidly across the enterprise value chain, often ahead of formal strategy and governance." Workers are adopting AI tools independently, outside of IT oversight, outside of approved vendor relationships, outside of any accountability structure whatsoever.

Sales functions are identified as simultaneously the top candidate for early agent deployment and "a governance-critical domain where trust, accountability, and human oversight must be deliberately designed." That's a domain combining massive decision volume, high digital agent suitability, and elevated commercial risk — customer interactions, pricing, commercial judgment. It's exactly where unchecked AI agents could do the most damage.

What Companies Should Do

The report stops short of being alarmist — it's genuinely optimistic about what co-intelligence (humans and agents working together) can achieve. But its prescription is clear: companies cannot simply deploy agents and hope for the best. Governance structures must be built proactively, before the agents are running critical processes.

"In a co-intelligent enterprise, leadership does not diminish as AI improves," the report reads. "It becomes more consequential." The companies that figure out how to redeploy freed human capacity toward genuinely higher-value work — rather than treating productivity gains as pure cost savings — will be the ones that thrive. The ones that don't will find that intelligence was scalable, but the chaos that follows is not.