AI is everywhere in industry talk, and new research by Foundry for BMC Helix reveals that while many European companies have successfully deployed AI across their IT services and operations functions, they’re lagging behind in agentic AI maturity.

AI is making headway

Across the board, European organizations are putting AI into practice in areas where speed and consistency matter, including:

• Change risk prediction and management (73%)
• Ticket automation and resolution (73%)
• Capacity planning/resource optimization (71%)
• Vulnerability risk remediation (66%)
• Root cause analysis (58%)

The value these deployments afford put vulnerability management and remediation on top with 79%, change risk prediction and management at 70%, and a little more than half (56%) seeing benefits in infrastructure cost and management.

The survey confirms that AI is transforming the role of IT teams by improving decision-making, freeing up resources for transformation, increasing the focus on business outcomes, and boosting innovation by making it easier to roll out projects or initiatives fast and fix issues post-deployment.

Universal, but fixable concerns about agentic AI

Many of these same companies reaping the benefits of general-purpose AI are hesitant to give autonomous, agentic AI free rein on their core workflows. And they have a consistent set of concerns: governance, risk, and compliance, and the data foundations that make autonomous actions safe and repeatable.

Fifty-nine percent mention governance, risk, and compliance as their leading concern. And because autonomy doesn’t work well amid unclear ownership and process boundaries, 43% are worried about how their siloed operations, management, and service teams will adapt.

Data is also a looming issue, with poor data quality a concern for 37%. In France, organizations are taking data so seriously that 70% continuously monitor data quality metrics, 59% have dedicated data quality teams, and 56% are using data quality tools. But that rigor doesn’t equate practical readiness. Less than half of them have followed through with a data governance framework, and only 12% are meeting their data challenges with a holistic approach that combines all of those measures.

Read more: Secure your business with AI governance that works

Fortunately, those things can be addressed with the right guardrails and processes that can explain what the AI did, why it did it, and how enterprise data was handled end to end.

A solid data and governance strategy is a core requirement

Since AI initiatives, most especially autonomous AI, thrive on high-quality data and struggle without it, a robust data strategy is a must-have, and AI can help with the very thing that is its lifeblood. As I comment in the report, you can’t wait until your data is perfect. but you can’t ignore data quality, either. The solution is to treat data improvement as a continuous process that AI itself accelerates.

Even when autonomous, agentic AI proves it can handle individual tasks, many organizations aren’t ready to extend that autonomy across the business. My advice: don’t force it—let trust compound through evidence.

Governance frameworks and user trust haven’t scaled with the technology. You don’t flip a switch from “human approves everything” to “AI acts freely.” You build a graduated system where AI earns expanded autonomy through demonstrated performance in progressively higher-stakes domains.

Five considerations for moving toward autonomous, agentic AI

If your concerns about agentic AI outweigh its promise, the key is to treat it like an operating model change—not a feature rollout. Here are five things to consider before deploying agentic AI, or scaling your pilot.

• Start with governed autonomy: Define what the agentic AI can do, when it must ask for approval, and how decisions are logged and audited. Build guardrails up front so autonomy expands by policy—not by accident.
• Invest in data readiness early: Assign owners, monitor quality metrics, and fix the highest-impact feeds before expanding scope. Be mindful of your company’s regulatory compliance obligations. That means being explicit about what data the AI can access, what it must never see, and how sensitive data is stored and protected.
• Roll out workflow by workflow: Pick high-volume, repeatable ops use cases (ticket resolution, change risk, vuln remediation) and earn trust through measurable wins.
• Align the human operating model: Clarify ownership across ops/service teams and design escalation paths so autonomy doesn’t get trapped in silos.
• Measure what matters: Keep track of time to resolution, change failure rate, risk reduction, and avoided toil, and expand autonomy only when performance holds. Require explainability and traceability at the same time: so that leaders can explain what happened without guesswork.

Conclusion

Implementing autonomous, agentic AI, or scaling your existing deployment to make it part of integrated workflows, doesn’t require a steep learning curve. But it does require the right governance frameworks, platform, and expertise. With those in place, you can put agentic AI to work for your business and start yielding the benefits now.

Click here to access the complete findings and download a complimentary copy of the full report.

Written by Erhan Giral, BMC Helix.