Why Your Team Resists AI Tools — and How to Win Adoption

You bought the tool. The demo was impressive, the case for it obvious. Then adoption stalled — logins trailed off, people quietly went back to the old way, and the rollout you championed became the thing nobody talks about in the standup. It's tempting to read that as laziness or change-fatigue. It's usually neither. Resistance to AI is a specific, well-studied behavior, and understanding it is the difference between a tool that gets used and one that gathers dust.

Algorithm aversion is real — and a little irrational

Researchers have a name for it: algorithm aversion. People will abandon an algorithm after seeing it make even a small mistake — even when that algorithm demonstrably outperforms the human alternative (Dietvorst et al., 2015). A human who errs gets a second chance; an algorithm that errs gets switched off. The aversion is also task-dependent: people accept AI readily for tasks they see as objective or technical, but resist it for tasks they consider human or judgment-laden (Castelo et al., 2019) — which is exactly the territory of sales conversations, coaching, and assessment.

So when your team pushes back on an AI tool, they're not being uniquely difficult. They're exhibiting a predictable pattern. The good news: because it's predictable, it's addressable.

Why people actually resist

Three things drive most of the resistance.

The black box. People distrust what they can't see into. When an AI produces a score or a recommendation with no visible reasoning, users assume the worst and disengage (Rai, 2020).

A threat to expertise. Skilled people read "let the AI decide" as "your judgment doesn't matter." For anyone whose identity is built on being good at the job, that stings — and they resist to protect their sense of competence and autonomy (Castelo et al., 2019).

Fear of replacement. If the tool is framed (or imagined) as a step toward doing without them, resistance is rational self-protection, not stubbornness (Monod et al., 2023).

What actually wins adoption

The research points to a consistent set of moves.

Give people control. This is the single most robust finding: people will use an imperfect algorithm far more readily if they can adjust its output even slightly (Dietvorst et al., 2018). A tool that lets users override, annotate, or tune beats one that hands down verdicts.

Make it explainable. Show the evidence behind a score or suggestion. Moving from black box to "glass box" converts suspicion into trust (Rai, 2020).

Frame it as an assistant, not a replacement. Position AI as augmenting people — handling the repetitive, surfacing the useful — while humans keep the judgment (Davenport et al., 2020; Monod et al., 2023).

Build the skill of working with it. Effective human-AI collaboration is itself a competence — knowing when to trust the tool and when to override it. Teaching that "AI literacy" measurably improves how well people use these systems (Sidra & Mason, 2025).

For sales and client-facing teams specifically

The aversion bites hardest exactly where these tools are most useful: scoring conversations, coaching, assessing people. A rep who'd happily let AI clean their CRM will bristle at AI judging their call. The way through is the same — transparency and control. When the criteria are visible, the feedback points to the actual moment in the transcript, and the rep can see why — not just what — the score lands very differently. (That's a large part of why AI conversation scoring works best against clear, observable criteria rather than opaque overall verdicts.) Practice in a private, low-stakes setting helps too: people trust a tool they've used on themselves, with no one watching, before it ever touches their real performance.

Adoption isn't won by mandating logins. It's won by removing the reasons people don't trust the thing — and giving them a hand on the wheel.

Sources for the research cited above: The Research Behind Our Guides.