It was no surprise that conversations at last week’s Gartner® 2025 IT Symposium/Xpo™ in Barcelona were all about AI—but the discussion has shifted noticeably. Last year was all about learning the tools and running small experiments; this year, the focus was on scaling, taking those first pilots to their next stages and transforming the whole enterprise.

Reflecting on the keynotes, breakouts, and client conversations, five major themes stand out.

The AI-ready tech stack is coming into focus

Coexistence. Agents will coexist with enterprise resource planning (ERP) and other systems of record. For most companies, a structured system of record (likely ERP) will remain the backbone of their back-office processes. Full AI transformations won’t be possible without upgrading those underlying systems. These transformations will need to rework future processes for AI—not just adding automation to existing ways of working, but fully rethinking and redesigning processes to capture the full benefits of working in a more flexible, “headless” ERP world.

Build/buy/partner. Companies will need to make clear choices about whether and where to buy, build, or partner in the design of their future agent workforces. Industry-leading software providers are embedding AI in their offerings, providing the opportunity to use native agents for in-application processes. But to fully scale, the tech stack will need to support multiple agents. This can only be done with Model Context Protocol (MCP) services to ensure that agentic systems of intelligence dock effectively with systems of record.

Security. The future stack will also need to be secure by design. In prioritizing speed and agility, initial experiments with generative AI may have overlooked security protocols and exposed enterprise data in uncoordinated ways. As they scale use cases, tech leaders need to ensure they comply with their organizations’ security frameworks. They should also follow strict Know Your Agent (KYA) protocols to make sure the agent actually represents the person or company it’s supposed to represent and that this person or company has authorized the agent to do the thing it’s trying to do.

Data remains the biggest challenge and opportunity

Most organizations still have a journey ahead to clean, organize, and govern their data so it’s readily discoverable and reusable by autonomous agents. To be trusted, data quality needs to be managed at the source:

• Data contracts ensure information is complete, accurate, and regularly refreshed.
• Clear and transparent lineage and provenance provide confidence in the data’s origins.

Companies must educate their workforces

Will AI create or kill jobs? We heard both views.

Competition. In either case, demographic changes will intensify the competition for the best AI talent. Companies should anticipate this shift by implementing ambitious training and education programs for their workforces.

Training. Tech departments are excited by the use cases, and the technology is ready to be deployed. But scaling AI will require business users to understand and adopt this technology. Companies could spend much more on training and education than they spend on the tech—twice as much, according to one conference speaker. 

Change management. Since change management hasn’t proven to be a strength of technology providers and system integrators, companies will need to rethink their deployment programs and build up their change management muscles.

Repatriation of technology

As with physical supply chains, companies are increasingly wary of relying too much on foreign suppliers to store sensitive data and perform critical processes.

To increase resiliency to geopolitical shocks, companies are more interested in sovereign cloud and local solution providers. Hyperscalers understand this trend and are expanding across geographies.

For AI in particular, more stringent data privacy requirements and regional regulations will drive the need for data to be processed locally, at the network edge. This will shape AI training and inferencing, creating a risk of bias.

CIOs will play a crucial role in the AI transformation

As technology democratizes, as tech providers develop use cases, and as vibe coding unleashes agents across the enterprise, is it time for CIOs to slip quietly into the background?

Quite the opposite: This is a critical moment for CIOs to set the ambition and own this space. Our conversations at the symposium suggest that less than 30% of AI initiatives are breaking even or creating value. (MIT’s widely cited study puts it at 5%.)

As leaders of the business, CIOs need to work systematically across domains to understand the right level of automation for each, including deterministic automation, agentic AI, and transformation with AI copilots, while remaining fundamentally human. With that clear understanding, they can paint an integrated picture of full potential ambition.

Finally, they should reinforce the importance of failing fast to abandon unpromising ideas and concentrate efforts on scaling the most valuable initiatives.

It’s exciting to see AI maturing, promising that 2026 will be another year of thrilling developments.