As organisations move deeper into the era of generative and agentic AI, enthusiasm remains high—but the focus is shifting. According to the “Building a high-performance data and AI organization” report from MIT Technology Review Insights, 65% of organisations have already deployed generative AI. The next phase, however, is less about experimentation and more about delivering measurable business outcomes.
Conversations with data and AI leaders across industries point to a common priority: building unified, well-governed data estates that can reliably support AI agents and applications. As enterprises look to scale specialised agents that can reason within their own environments, evaluation, governance and architectural simplicity are becoming as important as model performance.
Looking ahead, several strategic themes are expected to define how enterprises approach data and AI in 2026.
Model choice becomes a strategic requirement
The rapid pace of innovation among large language model (LLM) providers has created both opportunity and uncertainty for enterprises. While competition among AI labs has driven significant improvements in model capability, organisations are increasingly reluctant to commit to a single provider.
Instead, many enterprises want the flexibility to select models based on performance, cost and suitability for specific tasks—treating model choice as an architectural decision rather than a long-term lock-in.
“When innovation is this fluid, IT flexibility and the ability to switch between underlying models become major competitive advantages. Open technologies give companies the control they need to thrive in the new era of constant AI-driven disruption,” said Dael Williamson, Field CTO.
This emphasis on optionality reflects a broader shift towards open, interoperable AI stacks that can evolve as the underlying technology changes.
Governance moves to the centre of agentic AI
Governance is no longer viewed as a secondary concern limited to access controls. As organisations deploy AI agents that can reason, act and interact across systems, governance has become a core layer of the AI architecture itself.
In agentic environments, governance extends to semantics, lineage, dashboards and workload visibility—essentially defining how AI agents understand context, access data and operate within acceptable boundaries.
“Any successful AI strategy has to answer three questions: Can the business identify the data used? Do they understand which LLMs are being called? And can they explain what happened across the entire agentic AI chain? A strong and unified governance is the key to addressing each of these challenges,” said Robin Sutara, Field CDO.
As regulatory scrutiny and internal accountability increase, the ability to explain and trace AI-driven decisions is expected to become a baseline requirement rather than a differentiator.
AI development gravitates toward the data layer
In many enterprises today, AI development is fragmented—spread across multiple tools, teams and environments. This fragmentation often slows time to value and makes it harder to govern AI workloads consistently.
A growing view among data leaders is that AI agents and applications should be built closer to where enterprise data already resides, using open and interoperable formats. This approach reduces architectural complexity and makes it easier to apply consistent governance, lineage and security controls.
Unified, multimodal data—combining structured and unstructured sources—is emerging as a foundation for scalable AI adoption.
“The best, most adaptable businesses are using data to guide them in a fast-changing global marketplace. Simplifying the AI architecture and building new agents and applications where core, multi-modal business data already resides helps a wider number of users get to this important, business-critical intelligence faster,” said Dael Williamson.
By consolidating AI development around the data layer, organisations can also extend access more confidently across business functions, without creating new silos.
From AI ambition to “boring AI” execution
While discussions around AI superintelligence continue, enterprise priorities in 2026 are expected to be more pragmatic. Rather than pursuing highly speculative use cases, many organisations are focusing on applying AI to repetitive, routine and operational tasks—areas where gains can be realised quickly and consistently.
At the same time, there is growing emphasis on pairing AI with human expertise. Highly specialised agents are increasingly being designed to augment domain experts, enabling them to apply decades of institutional knowledge more effectively.
“A people-first approach to AI deployment is key. Organizations can maximize on institutional knowledge by arming veterans and newcomers alike with specialized tools that keep them focused on high-value tasks,” said Robin Sutara.
This shift towards what some leaders describe as “boring AI” underscores a maturing mindset: success is less about breakthrough moments and more about sustained productivity and better decision-making.
A more disciplined phase of enterprise AI
Taken together, these trends suggest that 2026 will mark a more disciplined phase of enterprise AI adoption. Flexibility in model choice, unified governance, architectural simplification and human-centric deployment models are emerging as the pillars of scalable AI strategies.
For data and AI leaders, the challenge is no longer whether AI can work—but how to operationalise it responsibly, transparently and at scale.