Agentic AI integration remains a difficult—and costly—problem for IT teams

Agentic AI is the way of the future—at least, that’s what tech companies are constantly telling us. However, many of the employees tasked with deploying it within organizations aren’t sure they’re ready to meet the moment.

A new report from Harvard Business Review, sponsored by telemetry provider Cribl, is the latest evidence of this disconnect. According to the research, a near-universal 96% of respondents—all of whom were in the HBR audience and “from organizations considering, piloting, or actively using agentic AI and who are both involved in their organization’s agentic AI decisions and familiar with telemetry”—report that agentic AI is expected to be important to their organization going forward.

But a mere 23% actually have a strategy for utilizing the technology.

Arranging your tech stack to address the needs of agentic requirements is essential, The AI Lab CEO Ryan Kurt said in comment in the report.

“Without the right infrastructure, you’ll hit a ceiling,” Kurt said. “There is absolutely no way to break through it unless you have the data scaffolding, the governance, and the integrated workflows that you need.”

Moving forward. We’re over a year into agentic AI’s widespread usage, and the technology’s adoption has continued to present challenges for IT teams. The new HBR research shows that implementing agentic AI isn’t just difficult, but also expensive—only 7% of respondents said that costs are coming in under projections, while 47% claim the costs are higher than expected.

Unsurprisingly, Cribl points to a telemetry gap as part of the problem. Data is often a point of concern for IT leaders, who regularly tell IT Brew that they are struggling to manage the demands of agentic integration with the security requirements to keep things safe.

Last year, telemetry was cited as a motivating factor behind Google’s acquisition of cloud security company Wiz. Anetac security platform co-founder and CEO Timothy Eades told IT Brew at the time that the telemetry capabilities were an unsung part of the deal that could lead to better application development.

“You now have incredible telemetry sources about how applications interact with cloud infrastructure,” Eades said. “Roll that into an AI model and then you can transform your cloud.”

Taking off. Telemetry aside, managing workflows and systems integration is a continuing challenge as agentic AI becomes a must-have for the tech stack.

“Think of it like an airplane engine,” Kurt said. “The more important that engine is for your company, the more you need to monitor it on a frequent basis so you can know when it starts to behave a little differently and intervene when necessary.”