AI on the board: Who decides, who is liable?

11.05.2026

4 min read

In 2026, BCG measured that in two out of three executive teams, the CEO makes the final call on AI budgets. In the same quarter, HBR explained why these setups regularly collapse. The real question isn’t which C-level title ultimately owns AI. The real question is which decision-making model a board builds to turn AI into a sustainable investment.

Key Takeaways

CEO dominance is the norm, yet rarely the optimum. BCG 2026 data: in two-thirds of boards, the CEO decides on AI investments. HBR shows why these setups fail: the gap between decision-making distance and operational depth can’t be bridged.
The model matters more than the title. Whether a Chief AI Officer, CIO or CDO formally owns AI depends on the company. What counts is how investment, risk and impact assessments are structurally separated in the boardroom.
Ownership without veto costs two years. If you own AI but lack veto power over business units, pilot projects stall. The bill arrives when pilots end and nothing scales.

What is AI ownership in the boardroom?

What is AI ownership in the boardroom? AI ownership is the formal assignment of investment, impact and risk accountability for AI initiatives to one or more board members. The role covers not only budget approval but also the authority to reject, reprioritize or condition AI projects across business units. In most DACH companies in 2026, this responsibility remains implicit with the CEO; formalized ownership models are still rare.

What the data reveals right now

In spring 2026, BCG analyzed 400 C-level responses. In 66 % of cases, the CEO made the final AI investment decision, with the CIO a distant second. Yet a second finding stood out: companies where AI responsibility sat with the CIO achieved a one-third higher success rate for selected projects than CEO-centric models.

HBR, covering the same period, highlighted a different variable. The decisive factor isn’t the title. It’s the distance between the decision-maker and the operational risk. A CEO far removed from the AI stack decides on decks and pitches. A CIO, closer to implementation, brings deeper operational skin in the game.

Both findings are compatible. They describe two layers of the same problem.

Three Ownership Models I See in 2026

Model one: CEO-centric with advisory anchor. The CEO makes the call, while a strategy consultancy provides the rationale. Advantage: fast decisions, clear external communication. Disadvantage: decisions often align with whatever the consultancy is currently pushing. We’ve seen mandates where a single pitch flipped an entire AI portfolio.

Model two: CIO or CDO formally accountable, CEO as sponsor. Common setup in larger DACH conglomerates. Advantage: investment and risk assessment sit closer to the stack. Disadvantage: without a clear escalation rule to the CEO, conflicts with business units aren’t resolved in time. Risk responsibility then lands with the CIO, without political cover.

Model three: collective model with CFO veto power. Rare, but where it exists, it works surprisingly well. The CIO owns operations; the CFO owns veto power over investments lacking clear outcome metrics. The CEO remains sponsor and tie-breaker. Advantage: investment discipline rises sharply. Disadvantage: requires a CFO who understands AI methodology well enough to justify vetoes.

What Boards Should Examine in 2026

Four questions cut through most debates. They’re not new, yet they’re often skipped in AI discussions.

First: who bears the risk if an AI initiative shows no impact after eighteen months? If the answer is the business unit that ordered it, ownership formally sits with the unit instead of the board. That’s a deliberate choice rather than an accident.

Second: who can halt an AI project against a business unit’s wishes? Ownership without veto power is symbolic.

Third: which metric defines success? If the metric is pilot projects launched, the board builds an activity machine. If the metric is productive scaling with measured business contribution, the board builds an impact machine.

Fourth: how does escalation work upward? An ownership structure without a clear escalation rule unravels at the first real crisis.

The Honest Recommendation

Whoever decides in 2026 lives with the model for at least three years. AI shifts a company’s risk profile faster than most other investments. Concentrating it with the CEO delivers speed. Concentrating it with the CIO delivers depth. Distributing it across the team with clear veto power delivers discipline.

The choice tests the character of the board far more than its tools. That’s the least comfortable truth in this entire debate.

Frequently Asked Questions

Who formally bears AI responsibility in most DACH boardrooms by 2026?

According to BCG data for 2026, two-thirds of companies have the CEO making the final investment decision. Explicitly designated AI leaders at C-level remain rare. In large corporations, the CIO ranks second, while in mid-sized firms responsibility often stays with the CEO.

Should every company appoint a Chief AI Officer?

Not automatically. A CAIO makes sense only when AI is core to the business model or the complexity of AI initiatives exceeds the scope of a single C-level role. For most DACH mid-sized companies, a clearly mandated CIO or CDO with a defined escalation path suffices.

How does the CFO veto model work in practice?

The veto applies to AI initiatives above a predefined investment threshold. The CFO doesn’t evaluate the use case itself but checks for a robust key performance indicator. If it’s missing or unmeasurable, the veto blocks approval. This forces the CIO and business units to define clearly what success means before the investment decision.

What are the warning signs of a misaligned ownership model?

Four red flags stand out. First: many pilots but little productive scaling. Second: AI spending follows consultant pitches rather than strategy. Third: no clear answer on who can halt a project. Fourth: no unified KPI across departments. Anyone who sees three of these four should revise the ownership model itself instead of swapping the vendor.