When AI makes software easier to create, the IT operating model has to change. Circles CIO Kai Chuang explains how CIOs can experiment with AI tools, work with teams to find high-value use cases, and rethink how IT operates.
In August 2023, I took a briefing from AWS's then-fledgling Bedrock product team on a whim, expecting a routine vendor pitch. Instead, I got a tour of the foundation-model ecosystem: the depth of Hugging Face, the pace of open-model development, and the energy of a community already building in public.
I came away with a singular thought: “We're behind.”
That kicked off a concerted AI program at Circles, which provides workplace hospitality services for enterprise employers, handling everything from concierge requests to benefits support. By early 2025, we had built a custom AI system and deployed it in our core contact center operations, reducing request turnaround time by 25 percent. It was a meaningful win. But the code behind it was still, by and large, written by humans.
By late 2025, frontier models crossed another threshold. They were no longer just changing what we built. They were changing the act of building itself.
This is the question CIOs now face. If AI makes building faster and easier, who gets to build? The answer can no longer be “IT, eventually, after intake and prioritization.” The CIO’s job is not simply to add AI tools to the current operating model. It is to rewire how the organization identifies and creates value.
Start with Building Something Yourself
CIOs can't lead an AI transformation from the sidelines. As Steve Jobs said in a 1990 interview, "The doers are the major thinkers... Did Leonardo have a guy off to the side, thinking five years out what he would paint? Of course not. Leonardo was the artist, but he also mixed all his own paints."
It's hard to champion something you’re not doing. AI makes building accessible to just about everyone, including CIOs.
In my spare time, I developed an iOS app to track my daughter’s basketball game stats: a low-stakes way to understand how fast AI coding tools are improving. In early 2024, I uploaded a UI mockup and asked it to write the code. The code was semi-functional and required extensive rework. But even then, it was clear where programming was headed. Over the course of 200+ commits, I had a front-row seat to advancements in state-of-the-art LLMs, where they excelled and where they still fell short. No amount of reading about AI can beat hands-on experience. A low-stakes personal project gave me room to experiment and learn without creating enterprise risk.
The Rewiring Starts Inside IT
Before AI changes who gets to build across the organization, it has to change how IT builds first.
Six months ago, our engineering team was hesitant to use AI tools—not out of ignorance, but from a rational fear of becoming less valuable. I worked with them to identify concrete use cases and showed them that AI didn’t reduce their value, it let them deliver more.
Today at Circles, roughly 80 percent of new code is AI-generated. Headcount stayed flat. We ship two to three times faster. The productivity gain meant more building, not fewer builders. When building gets cheaper, more projects become viable.
MIT Media Lab’s “The GenAI Divide: State of AI in Business 2025" study reinforces this: the barrier to value isn't model quality—it's learning, integration, and adaptation. Tools alone don't get you there.
From Roles to Capabilities
LLMs are inherently interdisciplinary, eliminating some of the handoffs that exist in traditional IT operating models. A good product manager can now design user interfaces. A strong engineer can explore use cases. A skilled designer can build prototypes. This is already happening in many companies.
Business relationship managers can no longer be just the bridge between IT and business units. Similarly, project managers can no longer just manage projects. They need to become AI-fluent and understand the business deeply enough to transform it—or risk getting left behind.
More builders means more code to evaluate. Security shifts from primarily assessing third-party vendors to continuously reviewing first-party code. Code-analysis tooling, integrated into the development process, becomes essential.
What the Rewired Org Looks Like
The end goal is an IT function that operates less like disconnected departments and more like a portfolio of internal product teams. IT is not waiting for fully formed business requirements; it is discovering opportunities together with the business. The organizations that benefit most will be those that deliberately build these hybrid skills.
The CIO's role shifts from gatekeeper of delivery to partner in value creation. For organizations that aren't there yet, this is the change journey.
