The AI tool has been live for eight weeks. The outputs are patchy, the team is polite but disengaged, and the results keep disappointing. You bring in the vendor to check the configuration. Nothing is wrong with the tool. Then someone mentions offhand that the inputs have been inconsistent from day one, and the picture snaps into focus. Inconsistent, and not by accident.
That is what passive resistance to AI looks like. It rarely announces itself.
What does AI resistance actually look like?
Passive resistance to AI rollout takes three main forms. The most visible is the workaround, where people complete the task manually and log the output as though the AI produced it. The second is minimal engagement, using the tool just enough to avoid challenge while giving it too little context to produce useful results. The third is bad-data feeding, deliberate low-quality inputs that generate predictably poor outputs.
None of these shows up in adoption dashboards. Logins look normal. Usage figures hold. The feedback forms are blank or vague. From a distance, the rollout appears to be progressing. The actual failure is happening in the quality of what people put in.
BCG’s 2025 research found a widening gap between AI tool usage rates and measurable business impact across the organisations they studied. Usage is climbing; results are not moving. That gap is not always a configuration problem. In many cases it reflects teams who have adopted the surface behaviour without any genuine engagement with the work the tool is supposed to change.
A delegate who reads this pattern as a technology issue will spend time and budget on vendor calls. The real conversation is a different one.
Why does this matter more than open objection?
Open objection is manageable. When a team member says clearly they don’t trust the tool or fear what it means for their role, you can respond. You can have the conversation, share context, adjust the approach. Passive resistance is harder because it is deniable. Workarounds plausibly look like teething problems. Bad inputs look like learning curves. The delegate gets no clear signal to work with.
Korn Ferry’s research on AI leadership readiness describes a common failure mode: organisations assign AI leadership to strong operators who then manage adoption as a project rather than a people change. The technical rollout looks structured. The human layer is treated as a compliance exercise. Tick the training box, send the comms, monitor the dashboards.
Spencer Stuart’s analysis of AI delegation found that the further leadership distances itself from the actual adoption process, the harder it becomes to read what is happening inside the team. The delegate who misreads passive resistance mistakes a stalling rollout for a technology problem, which compounds it. They go back to the vendor, adjust the tool, push harder on the metrics. The team’s resistance stays hidden, and from their perspective, confirmed. An authority figure is now more invested in defending the tool than in understanding their concerns.
That pattern is expensive to unpick.
Where will you actually meet it?
Passive resistance tends to cluster among people with the longest tenure and the most exposure to the specific tasks the AI is being asked to automate. These are often your most experienced team members. Their resistance is proportional to how much of their professional value they believe sits in the work the tool now handles. Writing that off as obstructiveness misreads the situation.
A senior administrator who has built institutional knowledge over a decade will not welcome an AI that replaces the judgment calls they have spent years developing. A finance manager who curates and interprets data for leadership will not be neutral about a tool framed as doing that curation instead of them. Their response is rational. The threat is plausible to them, and no rollout email has addressed it.
The OECD’s 2025 analysis of AI adoption in owner-managed businesses found workforce readiness and capability concerns cited consistently as the primary adoption barrier, above cost and above technology access. The barrier is human, and it sits in the middle of your most experienced team.
How the tool was introduced matters too. If AI arrived as an announcement rather than a conversation, resistance is the expected response.
What is it telling you about trust?
Passive resistance is best read as a signal about trust and perceived threat rather than about the tool itself. When team members fear that AI will reduce their headcount, expose their weaknesses, or hand their decisions to a system they cannot interrogate, they respond defensively. That response is rational. Address the underlying concern and the behaviour changes. Dismiss it, and the resistance compounds.
HRExecutive’s research on employee AI distrust identifies four fear categories driving passive resistance: fear of job loss, concern about the accuracy of AI outputs, lack of transparency about how AI decisions are made, and anxiety about personal data. Addressing the tool without addressing these fears produces compliance without genuine adoption.
The ESGDive survey found nearly half of executives fear AI-related job loss within a year. The people on your team are not unusual for feeling this. What is unusual is a rollout that takes the fear seriously rather than issuing reassurances.
Schellman’s analysis of AI implementation failures identifies organisational and human factors, including trust gaps and weak change management, as consistent contributors to pilot failure alongside technical ones. The tool usually works. The human layer is where things tend to come apart.
How do you defuse it?
HRExecutive’s research on employee AI distrust identifies five practices that a delegate can act on. Partner employees in AI education rather than delivering it to them. Co-create the processes that will use the tool rather than designing them without input. Communicate the business context behind decisions. Spread accountability for AI outcomes across the team. And address data rights explicitly, rather than leaving them unresolved.
Of these, co-creation is the practice that rollouts most commonly skip. A team that was involved in designing how a tool is used will defend it when results disappoint. A team that was handed a process will blame the tool when results disappoint. The deciding factor is whether people had agency in the decision, not how well the comms were written.
The conversation about job security needs to happen directly. Vague reassurances along the lines of “AI is here to support the team, not replace anyone” read as evasion. Team members who have heard enough corporate language know what it sounds like. Honest framing is more useful: what work this tool is taking over, what it is freeing people up to do instead, and what the plan actually is.
That conversation takes thirty minutes. The damage from not having it takes months to undo.
If the outputs from your AI tool keep disappointing, run a different check before calling the vendor. Look at the quality of what is going in. Ask who designed the process for using the tool, and who was in the room when that happened. The answer usually tells you where the resistance is coming from.
