AI speed gains are outrunning changes in decision rights

A senior people or transformation leader should be careful not to mistake faster execution for operating model change. My read on Deloitte's latest survey is narrower than the headline promise around transformation: many organizations may be adding AI into existing work, but far fewer appear to be changing authority, governance and the cost and delay built into the process in the places that would alter how the business actually runs.

Deloitte found in October 2026 that 81% of companies say AI made decisions and execution faster, yet only 25% use it to deeply transform the business. It also reports that 60% have redesigned jobs around AI to some degree, while only 15% report a significant shift in decision-making authority.

Work can get faster before approvals, accountability and control points change with it. In that situation, people may produce more or move faster, but manager review, escalation paths and final approvals still run the old way.

For a CHRO, CPO or CLO, the mistake to avoid is treating this as a learning adoption problem first. If people can already produce analysis, drafts or recommendations faster with AI, but approvals, role boundaries and accountability remain unchanged, the organization often gets local efficiency without much movement in throughput, manager load or time-to-decision.

Sometimes that is the right choice, especially where separation of duties or expert review protects the business. Sometimes it leaves value stranded because the job changed and the approval model did not.

I would start in the workflow, not with a broad training response. Look for a place where AI is already cutting effort. Then test whether the gain only shortened the task or also changed the handoff or the decision. If the work changed, check whether the current approver is still adding judgment or mainly preserving an old control point. The accountability issue follows from that. If an AI-assisted output is wrong, who owns the consequence and what level of human review is still warranted? Speed without that diagnosis scales waste.

Deloitte's framework is useful because it looks at who gets to decide, how much freedom people have in the work, and whether AI reduces cost, delay or rework in the process, not just whether people have access to tools. But this is still a framework and survey evidence, not proof that broad redesign is working at enterprise scale. The same data can support a more conservative reading: many organizations may be limiting authority shifts while they learn where AI is reliable enough, where controls must stay, and where managers are prepared to supervise different kinds of work.

I would not read this as a general market turn. I would read it as a test of whether organizations that adopt AI seriously will also change approvals, accountability and governance when the work changes. If they do not, most enterprises should expect uneven productivity gains and more manager burden than the headline numbers suggest. The better decision now is to set an evidence threshold: do not count AI-enabled speed as transformation unless authority, accountability and the cost and delay in the same workflow have been deliberately reworked.