AI integration leads to tech friction—a continuing challenge for IT pros

There’s an old saying, “a gem cannot be polished without friction,” and many in the IT sector hope that the same applies to AI.

What IT pros know as tech friction—the increase in toil and noise that comes from added complexity of the tech stack—is increasing with AI deployment. AI is adding a new layer of difficulty, and it’s one that’s hard to fully understand.

Get it done. In a May report from Harness, an agentic platform provider, researchers found that AI is impacting tech investment and ROI by adding friction to established processes. That raises questions for organizations grappling with appropriate deployment, Trevor Stuart, Harness SVP and GM, told IT Brew.

“What is the overall impact or productivity impact of AI-generated code? Are teams able to effectively monitor this? Are they able to effectively measure that ROI? That’s at the core of the report,” Stuart said.

For many in the IT sector, the disconnect between promised gains from AI and the reality has led to frustration. In part, said Matt Graney, CPO at intelligent automation platform Celigo, that’s because expected gains in productivity haven’t materialized in the way many expected. Meanwhile, executives continue to demand results, contributing to the tension around AI operationalization and tool deployment.

“Some describe this challenge of almost like the telephone game, where you take a kernel of the reality of a situation, and if it goes through a couple of AI cycles, it may to a casual observer look like it’s describing the same thing, but it begins to morph almost like a funhouse mirror reflection of what the original idea was,” Graney said.

Path forward. The problem of AI implementation-related tech friction is twofold, according to Alex Levin, CEO at voice AI agent platform Regal. On the one hand, this new technology has sped up engineering and development; on the other, it’s led to bureaucratic difficulties within organizations as people try to figure out levels of responsibility and how their tasks have changed.

“It’s not that AI is causing the problem, it’s that organizations are being pushed to finally think about this in a new way, and look at leverage-per-employee, revenue-per-employee as the new North Star,” Levin said.

For Stuart, the results from the Harness research show a path forward that makes sense—and should make everyone happy—by getting past friction-related troubles to develop better strategies. Measuring the impact of AI tools on the stack, for example, can help determine how the organization is deploying the technology and the effectiveness of different plans.

“Make sure that you’re approaching it from, how do we optimize the engine and how do we continue to improve our craft and improve our performance?” Stuart said. “That will deliver the inevitable ROI that organizations are looking for.”