ORLANDO, Fla. — Today’s CIOs are expected to become experts in and managers of a technology that’s evolving faster than humans can review it — and the pace will only increase.
AI systems are being developed in an architectural vacuum, Gene Alvarez, distinguished VP analyst at Gartner, said Tuesday during a session at the company’s IT Symposium/Xpo. The rapid pace of development has created unmanaged credential paths for AI systems and is angling companies into single-source vendor monopolies, he said.
CIOs need to become “active architects” of their organizations instead of passive observers to avoid these pitfalls, Alvarez said. He outlined a list of top technology trends CIOs can focus on to become active architects over the next several years, based on a review of venture capital investments in hundreds of emerging markets, patent submissions and analyst assessments.
Although the advisory firm included tentative time frames when these trends mature, each organization’s handling of the strategies will be different, Alvarez said.
“You need to match the maturity of your organization to the timing of when you’re bringing this technology in,” he said.
Establish AI agent identity
AI agent use — and agent autonomy — is growing in the enterprise. Agents are tackling some tasks that employees have traditionally handled, and increasingly, acting independently.
But with added autonomy comes risk. Gartner predicted that by 2028, one-quarter of enterprise security breaches will trace back to AI agent abuse due to unmanaged machine privileges and legacy credential practices. For security reasons, CIOs should consider developing identities and boundaries for agents in the next year or so, Alvarez said.
Dynamic identity controls enable organizations to create policy-controlled access, prompts and credentials to agents.
“It’s about securing agents as enterprise workers,” Alvarez said.
Introduce AI coding harnesses
Coding harnesses, or closed loop-control systems that govern autonomous developer behavior across the software development lifecycle, should follow in a similar timeframe, Alvarez said.
Without dedicated harnesses, developers spend time manually verifying machine output, which costs organizations time and money, often erasing productivity gains.
“The AI coding harness determines, ‘Is it the right order? Are we using the models efficiently?’” Alvarez said. “It also is going to enable agents to take on more complex workflows when it comes to coding.”
Assemble AI context platforms
These dedicated context layers provide AI agents with the contextual knowledge they need to interpret inputs, make decisions and act according to business objectives. CIOs can face operational risks, compliance failures, low trust and excessive token consumption when they don’t provide agents the context to their organizations.
“These are ontologies based on your data, your processes, your policies that say, ‘How do we do business?’ And the agents are able to find how we do business in our organization,” Alvarez said.
Gartner predicted that by 2027, organizations that prioritize context in AI-ready data will increase agentic AI accuracy by up to 80% and reduce costs by up to 60%.
Alvarez cautioned tech leaders against one-size-fits-all vendor products in this area, and to instead prioritize open, interoperable devices.
Build for sovereignty
As enterprises increase their tech budgets, sovereignty over their data, software and hardware is becoming a priority. Creating architecture that allows for realistic exit strategies from providers will become increasingly important to tech executives, Alvarez said.
The European Union unveiled the European Technological Sovereignty Package in June in a bid to curb reliance on U.S. tech companies; Gartner also predicted that by 2030, more than 75% of non-U.S. based organizations will adopt digital sovereignty strategies.
“You have to figure out how to have a realistic vendor exit from everything in your entire stack so that you can be perpetually autonomous,” Alvarez said. “Residency is where your data sleeps, but sovereignty is the ability to wake it up at any time.”
Preparing for large quantitative models
Although the trend is farther out — Gartner estimated between five and eight years — transitioning from general purpose large language models to large quantitative models will shift CIO decision-making from reactive troubleshooting to proactive simulation, Alvarez said.
Enterprise use of large quantitative models will enable complex data-driven decisions, require fewer resources than LLMs and improve model accuracy and reliability.
“This rigorous approach drives unprecedented business agility, enabling organizations to safely deploy autonomous forecasting in strictly regulated industries, such as finance and healthcare,” Alvarez said.
