.svg)
As artificial intelligence becomes the new normal, a familiar story is playing out in enterprise boardrooms and on engineering teams. Instead of driving ROI or achieving scale, most AI projects are stalling out. Without precise orchestration and executive buy-in, organizations risk creating what some experts call "AI sprawl." Now, winning in this new era means resisting the hype and pursuing a disciplined, iterative journey instead.
To learn more, we spoke with Anshu Mishra, Director of Enterprise Technology at Unity, a tool provider for creating, marketing, and growing games and interactive experiences. With decades of experience leading integration and automation architecture at companies like Deloitte, Atlassian, and WeWork, his expertise in enterprise tech runs deep. From Anshu's perspective, a fundamental shift in how leaders approach strategy will be key to finding success with AI.
Solve one problem, create many: Teams today can build faster than ever with low-code and AI co-piloting, Mishra said. But without enough oversight, this can create a chaotic environment ripe for agent sprawl. "Complex systems are getting more difficult to control. As agents and bots take the lead across enterprises, there’s more code and build sets but less governance and review. As a result, enterprises are finding themselves in a situation where they’ve created ten problems in their pursuit of solving just one."
Mishra said, even the most elegant architecture will fail if the people it's meant to serve don't adopt it. Here, he connected the low ROI figures plaguing the industry to a failure of human-centric design.
Build and implement together: "Rather than build in a silo and push a finished product onto a team, implement with them. Get their regular feedback and fail fast. Friction often comes from people, so it's very important to include them on this journey rather than going it alone." Collective learning encourages teams to share failures and successes, Mishra said, creating an environment ripe for improvement and iteration. As business unit boundaries blur, he continued, cross-functional AI implementation will be essential for this very reason.
To make sense of agentic chaos, Mishra compared the recent explosion of AI agents to the API revolution roughly ten years ago. Because every system offered its own API, he explained, central orchestration layers suddenly became critical. Now, Mishra said, the same architectural pattern increasingly applies today.
Avoiding agent sprawl: Without precise orchestration and executive buy-in, Mishra cautioned, organizations risk creating unmanageable AI sprawl, compliance failures, and a wider internal trust gap. To avoid this type of outcome, the first step is to establish a clear-cut strategy. He recommended starting with a simple yet profound question about ultimate purpose: Who is AI helping? Just like any other technology, he said, "the answer is 'humans.' And because humans are the beneficiaries here, they must stay in the loop."
The organizations that balance human-centric processes with disciplined execution will be the ones to emerge triumphant following this transformational period, Mishra concluded. For him, it all comes back to the foundational purpose of technology. Even as agentic orchestration evolves, "a machine is not going to help a machine. The entire system must serve a final, human purpose."
