Executives are ordering AI adoption from the top. Employees are routing around it from the bottom. The gap between the memo and the workflow is an engineering problem, and it is fixable.
Fifty-four percent.
That is the share of workers who bypassed their company’s AI tools in the past 30 days and did the work by hand instead, according to WalkMe’s fifth annual State of Digital Adoption report, which surveyed 3,750 executives and employees across 14 countries. A third had not used the tools at all.
Those numbers land in the middle of a very different conversation happening in the C-suite. Writer’s 2026 enterprise AI survey of 2,400 knowledge workers and executives found that 60% of executives are considering laying off employees who refuse to adopt AI, and 77% say employees who do not become proficient will not be considered for promotion. On the other side of the same survey, 29% of employees admitted to actively undermining their company’s AI strategy. Among Gen Z workers, it was 44%.
Read those side by side and you see the shape of 2026. Leadership is mandating. The workforce is resisting. And the AI investment sitting between them is producing far less than anyone planned.
How we got to the memo era
The modern AI mandate has a recognizable origin. In April 2025, Shopify CEO Tobi Lütke published an internal memo stating that reflexive AI use was now a baseline expectation, and that teams asking for more headcount needed to show why AI could not do the work. Within weeks, Duolingo announced it was going “AI-first.” Coinbase CEO Brian Armstrong later said publicly that he had let go of engineers who did not get on board with the company’s AI coding tools.
Those memos traveled. Boards read them. CEOs at mid-market companies forwarded them with a note that said some version of “we should be doing this.” By 2026, the top-down AI mandate had become a standard management move.
The intent is reasonable. Leaders see real productivity gains in the research, they see competitors moving, and they worry about falling behind. A mandate feels like the fastest way to close the gap.
The problem is what happens after the memo goes out.
What breaks under a mandate
A mandate changes what people are told to do. It does not change the tools, the data, or the workflow. That is where most rollouts quietly fail.
The tool lives outside the work. The most common AI rollout is a chat window in a browser tab. To use it, an employee has to leave the system where the work happens, copy context over, get an answer, and paste it back. Every handoff adds friction. Under deadline, people skip the extra steps, which is exactly what that 54% figure describes.
The tool does not know the business. A general-purpose assistant has no access to your pricing, your job history, your contracts, or your process documentation. Employees get generic answers to specific questions, then conclude the tool is not useful. They are often right.
The trust gap is real. The same WalkMe research found that 61% of executives trust AI for complex, business-critical decisions. Among workers, it was 9%. That is not a training problem. It is a signal that the people closest to the work have seen the tool get things wrong in ways leadership has not.
Usage becomes the metric. When the mandate comes from the top, the easiest thing to measure is logins and prompt counts. So that is what gets reported. Teams learn to generate activity, and leadership learns very little about whether anything improved.
The approved path is unclear. WalkMe also found that 34% of workers do not know which AI tools their employer has approved. A mandate that says “use AI” without an approved, well-built path pushes people toward personal accounts and pasted data, the shadow AI problem we wrote about in August, now with an executive endorsement.
None of these are attitude problems. Every one of them is a design decision someone did or did not make.
There is also a quieter cost. When a mandate pushes people to use a tool that produces unreliable output, someone downstream has to catch the errors. Estimators double-check AI-drafted quotes. Project managers reread AI-generated summaries. Customer service leads clean up AI-written responses. The work does not disappear. It moves to whoever has to verify it, usually your most experienced people, and that cost almost never shows up in the rollout report.
Wisconsin is not immune, just earlier
Wisconsin companies are coming to this from a slightly different starting point. Better Business Wisconsin’s 2026 Accredited Business Survey found that half of its member businesses do not use AI yet, and most of the other half use it for basic tasks like drafting emails and summarizing documents.
Manufacturing is moving faster. The Wisconsin Center for Manufacturing and Productivity’s Wisconsin Manufacturing Report shows AI adoption among state manufacturers rising from 23% in 2023 to 55% in 2025, much of it aimed at workforce shortages.
That puts a lot of Wisconsin leadership teams right at the decision point. The mandate playbook is available and widely copied. The question is whether to copy it or to skip the expensive part of the learning curve that larger companies are paying for now.
What works better than a mandate
The companies getting real returns from AI share a pattern, and it looks more like product development than policy.
They start with a workflow, not a tool. Pick one process with clear volume and a measurable output: quote preparation, service ticket triage, submittal review, invoice matching. Map how it runs today, where the time goes, and where errors creep in.
They build AI into the system people already use. The model should come to the work, not the other way around. That means integration with the ERP, the CRM, the ticketing system, or the document repository, so the AI step happens inside the existing screen with the existing context.
They ground it in company data. Retrieval over approved internal documents, with permissions respected, turns generic answers into specific ones. This is where most of the engineering effort lives, and it is the difference between a tool people tolerate and a tool people want.
They measure outcomes, not usage. Cycle time, error rate, rework, throughput. When the numbers move, adoption follows on its own, because people use what makes their day easier.
They let early users pull, then expand. A small group that sees real gains becomes the internal proof. That proof travels across a company faster than any memo.
Forrester’s 2027 Budget Planning Guides make the same point from the budget side, recommending that companies cut AI pilots that lack governance, ownership, success criteria, or a path to scale. In other words, stop funding activity and start funding outcomes.
The leadership role does not go away
None of this means executives should step back. Leadership still sets direction, funds the work, and makes it clear that AI is a priority. The difference is in what gets mandated.
Mandate the outcome, not the behavior. “We want quote turnaround cut in half by Q2, and we will invest in the tools to get there” lands very differently than “everyone must use AI daily.”
Mandate the approved path. Make it obvious which tools are sanctioned, what data can go where, and who to ask.
Mandate measurement. Require every AI initiative to name its success metric before it gets budget.
That is a mandate employees can work with, because it asks them to solve a problem instead of performing compliance.
Where Earthling fits
Earthling Interactive helps mid-market companies in Madison, across Wisconsin, and throughout the continental U.S. move AI from memo to workflow. That means picking the right first process, integrating AI into the systems your team already uses, grounding it in your data with the right permissions, and measuring what changes. When a mandate is not working, we can usually show you why within a few conversations.
Reach out and let’s look at yours.
