Executive summary
LHH and McKinsey both surface the same structural gap in 2026: as agentic AI takes on more operational decisions, the accountability structures that assign human responsibility have not kept pace. The defining leadership question is no longer whether to trust AI with decisions - it is who is answerable when AI gets one wrong.
Top leadership insight
Stable at the top, unprepared below
LHH's 2026 C-Suite Research tracked a striking shift: executive turnover dropped from 43% to 19% in a single year, signaling unusual stability at the top. But the same study found a growing skills gap beneath that stability. AI rose from outside the top ten to become the number one perceived gap in executive capability - with 49% of senior leaders identifying AI and technology as their top development priority.
The structural tension is significant. One in four senior leaders says current decision-making processes are inadequate for today's environment, and 28% cite lack of strategic clarity as a primary limiting factor. The executives who remained in seat are more stable than their predecessors - but many are operating in a landscape that is moving faster than their decision frameworks.
For senior leaders, the implication is direct: stability at the executive level is only an asset if it comes with the adaptive capacity to match the pace of change. Longevity without learning is a liability.
Source: LHH, 2026 C-Suite Research | lhh.com
AI and leadership update
Accountability structures outperform technology investments
McKinsey's State of AI Trust in 2026 finds that responsible AI maturity is improving, with the average score rising from 2.0 to 2.3 on a five-point scale. But the progress is uneven - and the differentiator is not technology investment. It is organizational structure. Organizations with explicit AI accountability structures averaged a 2.6 maturity score, compared to 1.8 for those without designated ownership. That 44% differential holds across every dimension McKinsey measures.
The barriers are telling. 64% of respondents cite security and risk concerns as the top obstacle to scaling agentic AI, and 60% point to knowledge and training gaps. Only around 30% of organizations have reached maturity level three or above - meaning most are still in early stages of building AI governance, even as they deploy AI across operations.
The finding is clear: a human name on an AI outcome is not bureaucracy. It is a performance variable. Organizations that have not assigned explicit accountability are not just slower on governance - they are measurably less effective at AI overall.
Source: McKinsey, State of AI Trust in 2026: Shifting to the Agentic Era | mckinsey.com
Leadership tip
At your next leadership team review, take one decision your organization relies on AI to support - and map it: who designed the parameters, who reviews the output, and who is accountable if it is wrong. If those three answers point to different people, or to no one, you have found your most consequential governance gap.
Framework of the week
The OODA loop
Colonel John Boyd, U.S. Air Force
The OODA Loop - Observe, Orient, Decide, Act - is a four-stage decision cycle developed by Colonel John Boyd for high-uncertainty, high-speed environments. Originally applied in aerial combat, it has since become a foundational model in competitive strategy, crisis leadership, and organizational agility. The core principle: the leader who can cycle through the loop faster than the environment changes holds a decisive advantage.
Executive application: As AI compresses the Observe and Orient phases - gathering data and generating analysis faster than any human team - the risk is that leaders skip the Decide phase entirely, letting AI's orientation become the decision by default. Map your most consequential recurring decision against the OODA loop. If AI handles Observe and Orient, where does human judgment enter? If the answer is unclear, you do not have a technology gap. You have an accountability gap.
Statistic of the week
44% higher responsible AI maturity in organizations with explicit AI accountability structures, compared to those without designated ownership.
The gap is not about investment - it is about clarity. Organizations that assign a human name to AI accountability outperform those that do not, across every dimension McKinsey measures.
Source: McKinsey, State of AI Trust in 2026: Shifting to the Agentic Era | mckinsey.com
Middle East lens
UAE elevates AI governance to Cabinet level
The UAE Cabinet has approved the formation of the Artificial Intelligence and Data Authority, consolidating three previously separate entities - the Office of AI, the Digital Economy and Remote Work Authority, and the UAE Data Office - into a single body reporting directly to Cabinet. It is chaired by Omar Sultan Al Olama, who holds the distinction of being the world's first Minister of State for Artificial Intelligence.
The authority's target: migrate 50% of federal government operations to agentic AI within two years. The governance implication for executives operating in the region is significant. The UAE is not debating whether to govern agentic AI - it is building the institutional architecture to do so at scale, faster than most private-sector organizations have begun the same conversation.
For boards and senior leaders in the GCC, this signals an accelerating regulatory environment. Organizations that have not yet established internal AI governance frameworks will find the gap between government expectations and organizational readiness narrowing - quickly.
Source: MIT Sloan Management Review Middle East; Gulf News | mitsloanme.com
Closing insight
When AI makes more decisions, the answer is not to pull back from adoption. It is to build the governance architecture that keeps a human name next to every outcome.
The OODA loop offers a useful frame. AI can accelerate observation and orientation. Deciding still belongs to a leader. Wherever that line is unclear, accountability goes missing - and the organization learns this at the worst possible moment.
The McKinsey data makes the case as plainly as data can: assigning a name to an AI decision is not a formality. It is a performance variable. Build the architecture now, before the decisions that test it have been made.
Zahi Abdein