Aaron DeNu – Principal We’re seeing a common pattern play out across organizations: AI tools get rolled out and training is delivered. Some people lean in fast, others hang back, and leadership starts tracking adoption metrics as if they’re the finish line. But underneath all of this, the work itself hasn’t changed.
Data from Gallup’s February 2026 survey of nearly 24,000 U.S. employees tells a similar story:
This gap between individual efficiency and transformative operational impact signals that most organizations haven’t fundamentally redesigned workflows, roles, or processes around AI. Without clarity on how work should change in the age of AI, employees invent their own version of AI-enabled processes, which fuels inconsistency.
After surveying 1,150 full-time U.S. desk workers in September 2025, researchers at BetterUp Labs and Stanford Social Media Lab found that 40% of workers have encountered AI-generated content that looks good but lacks the substance to meaningfully advance a task. Their research estimates that such instances cost employees roughly two hours of rework, which equates to roughly $9 million per year in lost productivity for a 10,000-person company.
To avoid lost productivity and stalled AI adoption, leaders need to define what “good” looks like when AI is in the workflow and reset expectations around performance and accountability. A practical place to start is by reclassifying work. Every activity can be viewed through one of three lenses: what’s automatable, hybrid, or human-critical. Classifying tasks across these three categories requires leaders to think critically about who—or what—should perform each piece of work, and what needs to change structurally to that activity as a result.
These are the rule-based, repeatable, and process-driven activities (e.g., reporting, monitoring, compliance checks, workflow execution), which are often easy to identify. What’s more difficult is designing the automatable tasks out of human roles. When a task moves to an automated system, the role around that task must also change. Otherwise, organizations risk adding a second worker to the same job.
That’s because without removing the automatable task out of the activity, people may continue doing it anyway, whether out of habit, distrust of the system, or because no one told them to stop. This creates duplication, inefficiency, and confusion about where the value in having a human involved actually lives.
Hybrid here means a deliberate partnership between human judgment and AI capability (not a location policy), and it’s the category that requires the most intentional design. It’s an explicit decision about who frames the work, who engages with the AI, who validates the output, and who makes the final call. Without that clarity, ownership becomes ambiguous, consistent habits don’t form, and quality of work suffers. According to the 2025 research by BetterUp and Stanford, more than half of the employees surveyed admitted to sending low-quality AI-generated content to their colleagues.
The same researchers found that this behavior was more common among people who had high trust in AI, low agency over how to use it, organizational encouragement to use it, and low psychological safety at work. In other words, when leaders push adoption without redesigning the work, people produce lower-quality work.
Designing hybrid activities well means partnering with the people who actually do the work to deconstruct and reassemble it. Research has long shown that this kind of participatory design (when workers are co-designers) results in tools that better fit real workflows and foster higher adoption.
This is where humans stay in the driver’s seat because the outcome requires high-stakes decision-making, ethical judgment, relationship ownership, or final accountability that cannot be delegated. This is the kind of work where “Because the AI suggested it” is not an acceptable answer.
When employees use that phrase, accountability diffuses and trust breaks. Roughly half of people surveyed by BetterUp and Stanford viewed colleagues who sent a low-quality AI-generated output as less creative, capable, and reliable than before, and 42% saw them as less trustworthy. AI can inform these activities, but humans must own them.
Reclassifying work is just the beginning of the structural conversation leaders need to have about roles, decision rights, and how AI should show up in the operating model as something with a defined role, not just a tool that employees can access.
BCG’s June 2025 research found that employees who receive strong leadership support for AI adoption feel more positive about generative AI, with training, in-person coaching, and visible leadership modeling as the variables that move the needle. If leaders aren’t using AI or being explicit about how they expect teams to use it, adoption will stall, and the people doing the work will create their own assumptions about how and when to use AI.
Leaders can get started by asking three questions about the work in front of them:
Increasing adoption is only the starting line. Redesigning work is the deep work that will enable organizations to realize business impact from their AI investments. Communication scholars have long observed that every new interface, from the printing press to the digital screen, reshapes how people think, not just how they work. The disruption is less about the technology itself and more about the gap between what the tool makes possible and how people are enabled to use it. Closing that gap means being deliberate about what gets automated, what gets redesigned as a collaboration between humans and AI, and what stays firmly in human hands. The future of work is human and AI together, but only if you design it that way on purpose.
Looking for a partner to identify where automation or AI integration will have the greatest impact so your people can focus on human-critical work? Schedule an Automation Opportunity Assessment to gain clarity and an actionable roadmap for how to get started. Or drop a line in the form below, and a TiER1 AI Strategist with be in touch.