A Capability That Arrived Before the Rules Did
Every previous wave of business technology — from quality management to cloud computing — followed a recognisable sequence: the capability emerged, organisations experimented, and management frameworks caught up before the technology became truly embedded. Clyde Calhoun, founder of AI governance firm Root Idea, argues that artificial intelligence has broken that pattern entirely. According to Calhoun, AI is "perhaps the first time a business capability has become mainstream before management has figured out how to manage it."
The reason, he contends, is structural. AI didn't enter organisations through steering committees and phased rollout plans. It arrived one employee at a time — people drafting communications, summarising contracts, analysing data, writing code — driven by individual curiosity rather than executive direction.
What the Data Reveals About the Gap
Calhoun anchors his argument in two data points from Deloitte research. In Deloitte's most recent CFO Signals survey, 87% of CFOs said AI will be extremely or very important to their finance department's operations in 2026, with only 2% dismissing its relevance. Yet Deloitte's State of AI in the Enterprise research finds that just one in five organisations has a mature governance model for the autonomous AI agents now entering their workflows.
That asymmetry — near-universal conviction about AI's importance, combined with minimal governance infrastructure — is the specific problem Calhoun is diagnosing. The numbers suggest that executive acknowledgement of AI's significance has not translated into the institutional discipline required to manage it responsibly.
Looking back a decade from now, I suspect we won't remember this period as the time organizations learned how to use AI. We'll remember it as the time they learned how to manage it.
Why This Is a Management Problem, Not a Technology Problem
Calhoun draws a sharp distinction that shapes the entire piece: most technologies automated existing work, whereas AI contributes to the thinking behind the work — generating recommendations, analysing alternatives, and influencing decisions that were previously made exclusively by people. This, he argues, is what makes AI categorically different, and what makes the governance lag so consequential.
His analogy is deliberately uncomfortable: implementing a new financial reporting system before designing internal controls, or deploying a cybersecurity platform before defining an incident response process. No executive team, he notes, would intentionally sequence the work that way. Yet that is effectively what has happened with AI across the enterprise.
The Questions Leaders Haven't Asked Yet
Calhoun frames the critical managerial questions that most organisations have not yet answered: How will AI-assisted decisions be governed? Where must human judgment remain essential? How will AI-generated outputs be validated? And — perhaps most pressingly — who is accountable when AI contributes to a material business mistake?
His reading of the executive psychology here is perceptive. He suggests that the unease many leaders feel isn't rooted in scepticism about AI's potential. It stems instead from the absence of an equivalent playbook: when inflation accelerates, a CFO has decades of tools and proven practice to draw on. When AI raises a hard question, no comparable institutional knowledge exists yet.
Writing the Playbook Mid-Game
Calhoun's closing argument is a reframe of the historical moment. Looking back a decade from now, he predicts, organisations won't remember this period as the time they learned how to use AI — that learning has largely already happened, informally and at scale. They will remember it as the time they learned how to manage it. Whether most leadership teams are moving fast enough to make that transition is, by his own account, still an open question.