Fran Rosch is the CEO of Imprivata, a digital identity company for life- and mission-critical industries.
When my partner and I visited Rhode Island several years ago, we came across an interesting house for sale. It was built in 2013, in an eclectic architectural style that did not really match our taste for a traditional, New England shingled home. But the bones of the home were strong, and the views of the Pawcatuck River were terrific. We decided to take a risk to purchase the property, believing we could transform it into our dream home. We’ve now spent almost two years renovating the space inside and out.
Our weekends disappeared into examining floor plans, walking the property and working with experts to balance our architectural vision with engineering constraints. In the process, I realized the hardest questions that came up weren’t about what we wanted to build or our favorite finishes, but how the space needed to work. How it flowed, how light entered and where existing walls or structures disrupted movement. Only after we understood those patterns could we make good decisions about what to preserve, what to change and where new systems belonged.
It was a familiar experience—a version of what I often saw in my work as a CEO helping organizations navigate technological change. The instinct in both settings is to start with the new thing: the wall you want to move, the technology you want to introduce and the capability you want to unlock. This instinct is particularly enticing for organizations today rushing to adopt AI. But renovating the house reinforced something I had come to believe about transformation: The goal isn’t to layer the future on top of the past. It’s to understand what you have well enough to thoughtfully redesign it for what comes next.
Ask the right questions.
Many business leaders often start their AI adoption conversations with capabilities. Their instinct is to ask about which models, copilots or potential vendors they should consider. Architects, however, don’t start by asking about materials. They first want to know how people will actually use and move through a space—where they’ll naturally gather, the access points they’ll most frequently traffic and how the environment can support those behaviors.
The same principle applies to AI adoption. Automation promises efficiency gains, but filling gaps requires first knowing which gaps to fill. The first instinct is often to ask what AI can do. Architects ask a different question: How will people live in the space? That’s a more uncomfortable place to begin because it forces you to observe before you build. Where have people adapted to unnecessary friction? Which tasks feel repetitive because the workflow itself is poorly designed?
AI magnifies workflow quality.
It’s often assumed AI is inherently transformative—that simply deploying it will produce better outcomes. In reality, AI is an amplifier, not a correction mechanism. It doesn’t fix poorly designed workflows; it accelerates whatever already exists.
If work gets duplicated, AI duplicates it faster. If approvals bounce unnecessarily between teams, AI accelerates those handoffs without making them any smarter. If employees spend hours hunting for information across disconnected systems, AI won’t magically create coherence.
The quality of the outcome still depends on the quality of the underlying system. Renovating taught me the same lesson. Updating finishes without rethinking the floor plan doesn’t improve the experience of living in a house. It just makes an inefficient layout look newer.
This philosophy has shaped how I’ve approached AI inside my own organization. Rather than asking where we could tack automation onto existing work, we first looked at how software development teams operated and where work most often slowed. Only after redesigning those processes did we introduce AI, resulting in accelerated feature development. AI could only transform the work after we redesigned how the work got done.
Form follows function.
More than a century ago, architect Louis Sullivan popularized the principle that “form follows function.” Good design begins not with how impressive a structure can look, but with what people need it to do. It’s an architectural idea, but it’s just as relevant to organizational design.
That distinction matters as organizations move from experimentation to embedding AI into everyday work. The success of that transition will be determined in large part by the less visible systems surrounding new capabilities: how information moves between applications, how access is granted, how decisions are routed and how much friction people encounter along the way. Each interruption may take only seconds, but together they shape whether technology feels intuitive or obstructive.
Buildings work much the same way. The experience of a space is often determined by design decisions that visitors barely notice. A doorway placed a few feet in the wrong direction can make movement awkward. A staircase can technically connect two floors while still discouraging people from using part of the building. Together, these decisions determine whether moving through a space feels natural or requires constant adjustment.
Looking back, the renovation succeeded because we resisted the temptation to start with what was visible. We spent more time understanding the structure than decorating it. Organizations should approach AI the same way. The strongest AI strategies are built on a clear understanding of how people work and a willingness to redesign those workflows and systems before asking AI to accelerate them. Only then can people spend less time navigating technology and more time applying judgment, creativity and expertise—the qualities AI is meant to amplify, not replace.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?