The onset of the enterprise AI race was marked by urgency. Boards demanded faster implementation, while market expectations focused on productivity gains through AI. In these early stages, leaders prioritized quick integration with AI providers perceived as dominant, often overlooking the long-term goals of such initiatives.
Changing Perspectives in AI Investment
Pressure remains, but its driving forces have evolved. Research conducted by AI platform Dataiku and The Harris Poll reveals that 65% of CEOs express greater concern over excessive investment rather than insufficient investment in AI. Unlike the early race, revenue growth has overtaken productivity as the primary measure of AI success, indicating a shift away from mere experimentation.
A significant 77% of CEOs anticipate that a peer may be removed due to failures in AI strategies or crises prompted by AI. According to Florian Douetteau, CEO of Dataiku, increased investment has paradoxically led to diminished confidence in AI deployment. As companies invest more, they realize the extent of their lack of control over new systems.
Unanticipated Structural Risks
Many organizations have entrenched themselves in vendor relationships that are now challenging to alter. Unpredictable consumption, opaque pricing, and shifting capabilities characterize these partnerships.
Companies dedicated to a single provider have constructed workflows dependent on their offerings, assuming stability in these relationships. However, changes such as contract renewals or competitive advancements can upend these assumptions. Douetteau likens this situation to cementing furniture, then discovering the need for frequent relocation before finalizing construction.
The risks extend beyond commercial terms. AI infrastructure is increasingly influenced by geopolitical factors. Regulatory measures, export controls, or governmental actions can impact access, transcending the vendor relationship itself. Contracts can secure aspects such as price and usage rights, but cannot fully shield enterprises from policy shifts impacting access and usage conditions.
Seventy-six percent of CEOs feel excessively exposed to operational or strategic risk due to reliance on limited AI vendors. Moreover, 67% have reconsidered or challenged decisions made by AI vendors and their internal teams within the past year. The integration burden is rising, as highlighted by a Dataiku/Morning Consult survey showing 74% of IT decision-makers struggle with fragmented AI tools.
Centralizing AI Strategy and Ownership
Decisions shaping AI implementation are dispersed across teams, vendors, and systems, yet accountability rests with CEOs. An intriguing gap highlighted by Douetteau’s survey indicates that 70% of CEOs claim ownership of AI strategies, while only 6% are involved in daily decision-making.
This division fosters dependency, as the comprehensive picture is never monitored by those responsible for its formation.
The Solution: Embracing Flexible and Default AI Systems
Rather than seeking the ideal vendor, CEOs are increasingly favoring adaptability in response to changing vendors, models, and economic factors. This strategy ensures retention of acquired knowledge rather than solely licensed capabilities.
A vendor relationship provides access to capabilities for as long as the vendor permits. Meanwhile, an orchestration layer above specific providers allows for model swapping without necessitating system rebuilds. It preserves the logic, governance, and internal knowledge within enterprise control.
Florian Douetteau emphasizes, “The layer above models and systems permits adding vendors, swapping models, connecting new data sources, while maintaining governance and previous work.” Dataiku offers a governed AI environment that facilitates AI development, deployment, and adaptation across existing vendors, models, and systems. It ensures control and traceability.
A successful AI strategy impacts not only operational efficiency but also long-term legacy. Eighty-one percent of CEOs assert that AI decisions are shaping or securing their legacy.
The companies thriving in this era are those who build systems flexible enough to adapt to market shifts, while retaining clarity over their operations a year down the line.
Douetteau poses a critical question for CEOs: What judgment elements have they embedded within controlled systems? Can these systems be clearly explained to regulators or themselves?
The winners of this period will be those who maintain flexibility and ensure that their AI reliance continues to be under their control, encompassing judgment, governance, and workflows.
For further insights on the survey data, explore the Global AI Confessions Report: CEO Edition.
*The Harris Poll conducted online research on behalf of Dataiku (February–March 2026), surveying 900 CEOs from companies with annual revenues of $500M or more across the US, UK, France, Germany, UAE, Japan, South Korea, and Singapore.