The AI Triathlete and the Problem Nobody Wants to Name in the Boardroom
There is a phrase that recurs in nearly every executive committee meeting where artificial intelligence projects are reviewed: "the pilot was successful." And then, silence. Nobody asks why the pilot never became anything else. The organization celebrates the experiment, files away the learnings, and three months later launches another pilot. The cycle resets without anyone having resolved the underlying question: who is responsible for making this scale.
That is the real problem identified in an article recently published in Forbes Technology Council under the byline of Anna Drobakha, global director of digital transformation and AI at Groupe SEB. The central argument does not revolve around technology or budget. It revolves around leadership architecture. And that distinction matters more than most boards of directors are willing to accept.
Drobakha's proposal is concrete: organizations that fail at AI transformation do not do so for lack of strategy, nor for lack of technical talent, nor for lack of investment. They fail because they distribute the three disciplines that transformation requires — strategic clarity, capability integration, and execution accountability — across different people, functions, and organizational layers, without anyone owning what happens in the spaces between them. The metaphor she uses is precise: a triathlon is not three separate races. It is a continuous effort in which the transitions between disciplines are as demanding as the disciplines themselves.
Why Pilots Don't Scale
Drobakha's diagnosis is not new, but most organizations treat it as though it were. Every so often a new nomenclature emerges for the same problem: implementation gap, change debt, adoption void. The name changes, the fracture persists.
What the article puts on the table with unusual clarity is that this fracture is neither technical nor budgetary. It is a fracture of accountability. In most large organizations, the strategist who designs the AI roadmap does not control data resources. The capability architect who builds the platform does not govern operational workflows. The operator who attempts to implement change has no authority over the agenda of the committee that approves what gets scaled and what gets discontinued. Each does their part with rigor. Nobody owns what happens between the parts.
This is not a minor dysfunction. It is the exact mechanism by which most corporate AI initiatives die a slow and dignified death, without failing with enough stridency to generate urgency for correction. The pilot "was successful." Adoption "is in progress." Scale "requires further alignment." And the organization keeps investing in technology while the real bottleneck — the coherence of leadership across strategy, capability, and execution — remains without intervention.
What Drobakha calls "the AI triathlete" is not a hiring profile or a new title for the organizational chart. It is a description of the capacity that organizations need to develop in their full executive teams: the ability to sustain all three disciplines in simultaneous motion, read signals in one and make operational decisions in another without losing systemic coherence. That capacity is not delegated to a Chief AI Officer and considered resolved. It is built — or not built — within the collective leadership. There is no structural shortcut.
The Transition as Unit of Measurement
There is a detail in Drobakha's argument that deserves more attention than it typically receives in transformation analyses: the idea that the transitions between disciplines are where initiatives are won or lost. Not in the initial strategic sprint. Not in the execution phase. In the step from one to the other.
This has concrete implications for how organizations should measure the maturity of their AI transformation. Most current frameworks measure capabilities: do they have quality data? Do they have deployed models? Do they have data science talent? These are legitimate questions, but incomplete ones. What they do not measure is the quality of the transition between strategic diagnosis and operational redesign, or between operational redesign and adoption at scale. That is precisely where accumulated work either dissipates or consolidates.
An executive team can have impeccable strategic clarity about where AI generates value for their business, build a solid technical platform, and still watch adoption stall because nobody rigorously designed the step between the logic of construction and the logic of organizational mobilization. They are distinct disciplines. They require distinct attention. And in most organizations, that step is taken for granted or delegated to change management as if it were an administrative process separate from the strategic core.
Drobakha frames it more rigorously: leaders who sustain transformation do not react to each transition. They manage the system continuously, anticipating where energy will dissipate before it happens and redirecting resources accordingly. That is not project management. It is systems thinking applied to leadership architecture.
The difference between these two capacities is not trivial. A project manager executes the plan. A systems thinker modifies the plan when they detect that the conditions that justified it have changed, and does so without waiting for failure to compel them. In practice, organizations that scale AI sustainably have at least some leaders of this second type operating with sufficient visibility and authority to adjust the system as it moves forward. Those that do not accumulate successful pilots.
The Structural Problem the Metaphor Does Not Resolve
Drobakha's article is rigorous in its diagnosis and honest about the complexity of the problem. However, there is a point where the argument requires greater tension to be useful as an instrument of executive decision-making.
Asserting that organizations must develop the capacity of the "AI triathlete" across the entire executive team — that every functional leader must operate with coherence across strategy, capability, and execution in AI — is correct as a description of the target state. But it omits the question of how that capacity is financed without dissolving the functional focus that makes each executive position valuable in the first place.
A CFO who dedicates significant cognitive energy to modeling the impact of AI on cost structure while managing a complex debt cycle and a fiscal close process is not being strategically irresponsible when prioritizing. They are making a choice. And that choice carries a visible opportunity cost. The triathlete proposal assumes that the capacity to operate across three disciplines simultaneously is additive without an equivalent cost in functional depth. That assumption deserves to be challenged before it becomes an organizational expectation.
What distinguishes organizations that are resolving this in a more durable way is not that every leader is equally competent in all three disciplines. It is that they have sufficient leadership density at the transition points — people with real authority, not just visibility — so that none of the gaps between disciplines goes without an owner. This can be achieved with individual triathletes. It can also be achieved with executive teams where deliberate overlapping of responsibilities covers the spaces between functions.
The distinction matters because it defines what an organization is looking for when it hires or develops leaders. Searching for complete triathletes is a concentrated bet on rare profiles. Designing teams with deliberate strategic overlap at transition points is an organizational architecture problem that is more accessible and, in many contexts, more robust against talent turnover.
The Concession the Article Does Not Name
There is something that Drobakha's argument leaves implicit but that deserves to be made explicit, because it is where most organizations never arrive: scaling AI sustainably requires that the executive team accept discontinuing initiatives with the same rigor with which it launches them.
The problem of pilots that do not scale is not only that nobody owns the transition. It is also that organizations rarely have the discipline to close what is not working before launching the next thing. The result is an accumulation of active initiatives competing for the same data resources, the same technical talent, and the same leadership attention capacity, without any of them having the critical mass to reach scale.
Drobakha mentions in passing that execution requires "disciplined governance over what to test, what to stop, and what to scale." That sentence deserves to be the center of the analysis, not a subordinate clause. Because the capacity to stop is, in practice, the scarcest of all. Stopping a visible pilot that was launched with the political backing of the CEO carries a real organizational cost. It requires someone with sufficient authority to execute it, justify it, and absorb the relational cost of doing so. That decision is, strictly speaking, the most demanding of the triathlete's three disciplines. Not the most technical. Not the most strategic. The most human.
The organizations that are gaining ground in AI are not necessarily those that invest the most or those that have the most sophisticated models. They are those that have developed the institutional capacity to commit to fewer things and sustain that commitment when pressure to demonstrate breadth of agenda is felt from the board. That is the concession that defines whether the leadership architecture described in the article is an operational framework or a well-worded aspiration.
The AI triathlete, in the most useful version of the concept, is not the leader who knows everything. It is the one who knows what to let go of so that everything else actually gets somewhere.
