Executives see AI delivering impressive results. The teams responsible for making it work are experiencing something very different.
For many executives, AI is remarkably straightforward. They ask a question, generate a presentation, summarize a report, or produce a strategic recommendation in seconds. The results are impressive, and the potential for productivity gains seems obvious. From that perspective, it's reasonable to wonder why the rest of the organization isn't moving faster.
But implementing AI across a business is fundamentally different. Employees deal with inaccurate outputs, extra verification, broken workflows, unclear responsibilities, and systems that don't communicate reliably. Managers are trying to accommodate new demands while keeping existing operations running. Much of the effort required to make AI useful happens outside the dashboards executives rely on to assess progress.
The people making decisions about AI and the people responsible for making it work are experiencing two different realities.
This isn't a new problem. In 1983, researcher Lisanne Bainbridge published Ironies of Automation, describing how automation could make routine work easier while leaving employees responsible for increasingly complicated problems. As machines took over predictable activities, people became responsible for monitoring systems, identifying failures, and intervening when something went wrong. Yet automation could also remove the everyday experience employees needed to develop those capabilities.
Four decades later, AI is creating a similar challenge. Generating a financial analysis might take minutes, but someone still needs to verify the numbers, investigate inconsistencies, and determine whether the recommendations are appropriate. Automating customer service might reduce routine interactions while leaving employees to resolve more complicated exceptions. The technology accelerates one part of the process, while the organization absorbs the additional work elsewhere.
Workday's 2026 research suggests that approximately 40% of reported AI time savings are offset by checking, correcting, and reworking outputs. Although this is a survey-based estimate rather than a universal measure, it highlights an important distinction: producing something faster doesn't necessarily mean the business can deliver results faster.
The danger for transformation leaders is that these operational realities are difficult to communicate to executives whose direct experience with AI has been overwhelmingly positive. Reports showing increased adoption, faster task completion, and greater output reinforce the belief that transformation is progressing. Meanwhile, the people responsible for implementation are struggling with the organizational consequences of changes that were never fully designed.
We have understood this problem for decades. Sociotechnical researchers, including Enid Mumford, demonstrated that technology cannot be designed independently of the people, responsibilities, decisions, and relationships that make an organization function. Introducing new capabilities without reconsidering how work is organized often transfers complexity rather than eliminating it.
For leaders caught between executive expectations and operational reality, the answer isn't to argue that AI doesn't work. It's to make the full cost and consequences of AI-enabled work visible.
Start with one workflow where AI is already delivering measurable time savings. Follow the work beyond the initial output. Identify who verifies it, who corrects mistakes, where decisions get delayed, and which responsibilities have changed. Involve the employees doing that work in determining what needs to be redesigned.
This creates a different conversation with executives. Instead of reporting that AI adoption is increasing while teams struggle to keep up, transformation leaders can demonstrate where productivity gains are being lost and what organizational changes are necessary to recover them.
The challenge isn't convincing executives that AI has limitations. It's helping them understand that successful AI transformation requires more than a successful AI experience.
Forty years ago, researchers warned that automation could make the technology look more capable while making the human work surrounding it more difficult. We have the opportunity to avoid repeating that mistake, but only if we measure what's happening across the organization—not just what the technology can produce.