How to Succeed with Your AI Strategy

Something unusual happens in the boardrooms and executive committees of major financial institutions when the topic of artificial intelligence comes up. Every leader has an opinion — firm, confident, sometimes passionate. Yet a significant proportion of them have never used a single AI tool in their professional practice. And an even larger majority confuses the AI question with digital tools, cloud, or data — as if the current revolution were simply an extension of the previous one.

My academic and professional background has led me to explore this topic in depth. What I have learned is that the core challenge is not technological. It is primarily strategic, organizational, and human. And the mistakes leaders make are predictable, because they are structurally the same in nearly every organization.

First Mistake — Treating AI as an IT Project

The first and most widespread mistake is assigning AI deployment to the IT department. This is not a matter of competence — IT teams are often perfectly capable of deploying tools. It is a matter of positioning and sponsorship by the board, the executive team, and business leaders.

AI is not a tool. It is a new way of making decisions, designing processes, interacting with customers, and managing risks. Effective deployment requires that business teams be at the center of the steering — not in the position of internal clients of the IT department. An AI program run without strong business anchoring produces brilliant technical pilots that never reach production, or that teams never adopt, because no one in the organization truly owns them.

The first condition for success is identifying high‑impact business use cases, assigning their ownership to operational leaders, and building mixed teams around them where technical and business expertise are on equal footing.

The telltale sign: If your AI program is driven exclusively by IT, it is unlikely to transform the organization. If business teams co‑lead it with equal authority and responsibility, the odds of success increase dramatically.

Second Mistake — Confusing GenAI with AI

ChatGPT created a perverse effect in executive committees: many leaders now have direct experience with generative AI — and have concluded that this is where everything happens. Yet generative AI, powerful as it is for certain use cases, is only one component of what can be deployed.

Predictive AI — which learns from historical data to anticipate future behaviors — remains the most value‑creating technology:
credit scoring, fraud detection, risk pricing, churn prediction, early client diagnostics, patient‑journey personalization, predictive maintenance. These applications are mature, proven, and deliver measurable economic returns. They do not need GPT‑4 to work. They need quality data, well‑calibrated models, and integration into operational decision‑making processes.

Organizations that succeed do not deploy the most sophisticated AI — they deploy the AI that matches their organizational maturity.

The question for the executive committee: Do we know how to distinguish, within our AI initiative portfolio, what belongs to predictive AI, generative AI, and agentic systems? And have we calibrated our ambition to our actual level of readiness?

Third Mistake — Neglecting Hearts and Minds

The third mistake is the deepest, and the least addressed: technology is not the limiting factor. Humans are.

Organizations often fall into four possible configurations:

• Teams with the skills but not leadership support — frustrated.
• Teams with leadership support but not the required skills — enthusiastic but ineffective.
• Teams with neither — indifferent or resistant.
• Teams with both (expertise and governance support) — the primary driver of success.

This is not theoretical. It is exactly what we observe in the field. An organization whose teams fear being replaced by AI will never deploy AI effectively — they will sabotage the process, consciously or not. An organization whose teams understand that AI frees them from low‑value tasks so they can focus on what truly requires human judgment — that organization accelerates.

The first challenge for a leader launching an AI strategy is not choosing the right tools. It is crafting the narrative — convincing, honest, grounded in business reality — that explains to teams what AI changes for them, and what it does not. Silence is interpreted as a threat. Transparency is interpreted as respect.

What Leaders Must Decide Now

AI strategy is not a topic for three years from now. Competitive positions are being built today — and will be difficult to catch up once established.

Three decisions are essential for leaders who want to avoid irreversibility.

First decision:

Choose two or three high‑impact business use cases and execute them with industrial‑grade rigor — measurable objectives, dedicated teams, and a scaling plan defined from day one.
AI is deployed through use cases, not platforms.

Second decision:

Invest in data before investing in models.
Data quality remains the most underestimated limiting factor in AI projects, regardless of industry. Sophisticated models fed with poor‑quality data produce unusable — and sometimes dangerous — results from a regulatory or strategic standpoint.

Third decision:

Raise the AI governance question to the board level.
Who in the organization is responsible for decisions made by algorithms?
How are biases detected and corrected?
What is the compliance strategy in a rapidly evolving regulatory and competitive environment?
These questions cannot be left solely to technical teams.

The Uncomfortable Truth

In three years, the organizations that succeed in their AI transformation will not necessarily be those with the best data scientists. They will be those whose leaders were capable of learning enough to ask the right questions — and humble enough to listen to the answers.

AI is not a technological revolution. It is first and foremost a managerial revolution. It requires leaders to learn, unlearn, and drive change under high uncertainty. These skills are not new. What is new is the urgency with which they are required.