Evans Learning Labs
LinkedIn Email
Applied Research Brief

Why AI Adoption Fails on Perception, Not Infrastructure

Organizations investing in AI readiness overwhelmingly measure technical infrastructure: data quality, system integration, computational capacity. The technology acceptance research finds that whether employees actually use the AI tools an organization provides depends considerably more on two specific perceptions, that the tool is useful and that it is easy to use, than on how sophisticated the underlying infrastructure actually is.
Abstract
Davis's (1989) Technology Acceptance Model established that two specific perceptions, perceived usefulness and perceived ease of use, predict individual technology adoption more reliably than the technology's actual objective capability, a finding validated across decades of subsequent research on workplace technology adoption. Venkatesh and Davis's (2000) extension of the model identified the specific organizational and individual factors that shape perceived usefulness, including social influence and job relevance, factors largely independent of technical sophistication. This brief reviews what the technology acceptance research indicates about AI adoption specifically, why organizations investing heavily in AI infrastructure while underinvesting in the perception factors that actually predict use frequently see low adoption despite genuine technical capability, an AI-specific perception dimension the original framework did not directly address, calibrated trust in probabilistic outputs, which can produce nominal usage without the genuine reliance that delivers actual performance benefit, and a genuine limitation: perception-focused readiness investment and technical capability investment are complementary requirements, not substitutes for each other.

Perception Predicts Adoption More Than Capability Does

Investment level and predictive value run in opposite directions
Technical infrastructure
Investment
Predicts adoption
Perceived usefulness & ease of use
Investment
Predicts adoption
Figure 1. A technically sophisticated system with weak perceived usefulness sees lower actual adoption than a less capable one perceived more favorably.
Davis, 1989

Davis's (1989) Technology Acceptance Model, developed initially for workplace information systems and subsequently validated across a wide range of technology categories, established that individual adoption of a given technology is predicted primarily by two specific perceptions: perceived usefulness, whether the individual believes the technology will genuinely improve their job performance, and perceived ease of use, whether the individual believes using it will require reasonable rather than excessive effort. Critically, his research found these perceptions predicted actual usage behavior more reliably than the technology's objective capability, meaning a a genuinely powerful, well-built, sophisticated system with weak perceived usefulness or perceived ease of use would see lower actual adoption than a less capable system employees perceived more favorably on these two dimensions.

This finding has direct and somewhat uncomfortable implications for how most organizations currently approach AI readiness. Organizational AI investment overwhelmingly concentrates on the objective capability layer, data infrastructure, model selection, system integration, precisely the dimension Davis's research found least predictive of whether employees would actually adopt and use the resulting tools. An organization can build genuinely sophisticated AI capability while significantly underinvesting in the perception-shaping work that Davis's model identifies as the actual determinant of adoption, producing a technically capable system with disappointing usage rates that the infrastructure investment alone cannot explain or resolve.

What Shapes Perceived Usefulness and Ease of Use

Venkatesh and Davis's (2000) extension of the original model identified the specific factors that shape perceived usefulness, moving beyond the original model's treatment of these perceptions as fixed individual assessments toward an understanding of what organizations can actually influence. Social influence, whether colleagues and supervisors visibly use and endorse a given technology, emerged as a significant predictor of perceived usefulness independent of the technology's actual features, meaning an identical AI tool introduced with visible, genuine leadership adoption produces measurably different perceived usefulness among employees than the same tool introduced without that visible endorsement. Job relevance, whether an employee can concretely connect the tool to their actual daily work rather than perceiving it as a generic capability disconnected from their specific responsibilities, similarly predicted perceived usefulness independent of the tool's objective sophistication.

Perceived ease of use, while partly shaped by genuine interface design quality, is also shaped by factors organizations can influence directly: the availability of genuine, accessible support during the learning period, and the framing of initial mistakes as a normal part of learning a new tool rather than as performance failures, both of which affect how effortful the adoption experience is actually perceived to be independent of the underlying interface's objective usability. Organizations that introduce AI tools with minimal support and an implicit expectation of immediate proficiency are producing lower perceived ease of use than the same tool introduced with genuine onboarding support, regardless of how well designed the interface itself actually is.

Why Infrastructure-Only Readiness Produces Disappointing Adoption

What organizations can actually influence
1
Visible, genuine leadership and peer adoption
Not top-down mandate alone -- social influence shapes perceived usefulness directly
2
Concrete job-relevance communication
Connected to specific daily responsibilities, not generic capability messaging
3
Genuinely supportive onboarding
Early mistakes framed as normal learning, not performance failure
Figure 2. None of these are infrastructure investments -- they're the specific levers the research identifies as shaping the perceptions that actually predict use.
Venkatesh and Davis, 2000

Organizations measuring AI readiness purely through technical infrastructure assessment, data quality, integration capability, computational resources, are measuring exactly the dimension Davis's and Venkatesh and Davis's research found least predictive of actual employee adoption, while leaving the perception factors that most directly predict adoption largely unmeasured and, in most organizations, largely unaddressed as a deliberate readiness investment. This produces a specific, recurring organizational pattern: substantial infrastructure investment followed by disappointing adoption rates that the infrastructure investment alone cannot explain, because the infrastructure was never the primary constraint on adoption in the first place.

The practical implication is that genuine AI readiness assessment needs to measure perceived usefulness and perceived ease of use directly, at the population level before a tool is introduced and calibrated to specific employee populations rather than assumed to be uniform, alongside the technical infrastructure assessment most organizations already conduct. An organization that discovers through this assessment that a target population has genuinely low perceived job relevance for a planned AI tool has identified a specific, addressable readiness gap that infrastructure investment alone will not close, requiring instead the kind of job-relevance communication and social-influence cultivation Venkatesh and Davis's research identifies as actually shaping that perception.

Building Readiness Around Perception, Not Just Capability

Organizations building genuine AI adoption readiness should invest deliberately in the specific levers this research identifies as shaping perceived usefulness and ease of use: visible, genuine leadership and peer adoption rather than top-down mandate alone, concrete communication connecting the tool to specific employees' actual daily responsibilities rather than generic capability messaging, and accessible, genuinely supportive onboarding that treats early mistakes as normal learning rather than performance failure. These are the levers Davis's and Venkatesh and Davis's research identifies as actually predicting adoption, and they require a different kind of organizational investment than technical infrastructure, one most organizations' current AI readiness planning substantially underweights relative to what the evidence indicates it should receive.

AI-Specific Perception Risks Beyond the Original Model

Davis's original model was developed for conventional workplace software, and AI adoption specifically introduces a perception dimension the original framework did not directly address: trust in the system's outputs, distinct from perceived usefulness and ease of use, given that AI tools, unlike conventional software, frequently produce probabilistic outputs an employee must exercise judgment about rather than deterministic outputs whose correctness is immediately verifiable. An employee can perceive a given AI tool as genuinely useful and genuinely easy to use while still remaining reluctant to actually rely on its outputs without independent verification, a distinct adoption barrier the original technology acceptance framework, built around more deterministic software categories, does not fully capture.

This trust dimension matters specifically because it can produce a pattern of nominal adoption without genuine reliance: an employee technically using the AI tool, satisfying a usage-rate metric, while continuing to independently verify or redo the tool's output before acting on it, meaning the organization records adoption success while realizing little of the actual efficiency benefit the tool was meant to provide. Organizations measuring AI adoption should distinguish nominal usage from genuine reliance specifically, since the perception factors this brief has reviewed predict the former more directly than the latter, and closing the gap between the two requires building calibrated trust, employees developing an accurate sense of when the tool's output can be relied on and when independent verification is still warranted, a distinct readiness dimension from the usefulness and ease-of-use perceptions Davis's original model addresses.

Sequencing Perception-Building Before Broad Rollout

Given the evidence this brief has reviewed, organizations planning AI tool introduction should sequence perception-building deliberately ahead of broad rollout, rather than introducing a tool organization-wide and hoping perception develops favorably afterward. Identifying a genuinely credible early-adopter population, employees whose visible, authentic use of the tool will function as the social influence Venkatesh and Davis's research identifies as shaping perceived usefulness for the broader population, and ensuring that population has a genuinely well-supported, genuinely positive initial experience before broader rollout, builds the social proof and word-of-mouth credibility that a broad, simultaneous rollout, with its inevitable mix of positive and negative initial experiences occurring all at once, cannot realistically replicate.

This sequencing also allows organizations to identify and address job-relevance and ease-of-use gaps specifically before they have been experienced by the full target population, since problems identified during a smaller, deliberately sequenced initial rollout can be addressed before they shape the broader population's first impression, while the same problems discovered only after full organization-wide rollout have already produced exactly the negative early perception this brief's evidence indicates is genuinely difficult to reverse once it has taken hold. Organizations that treat AI rollout sequencing as a genuine strategic decision worth real planning investment, rather than defaulting to simultaneous organization-wide deployment simply because it is administratively simpler to execute, are applying the evidence this brief has reviewed with the deliberateness the underlying research actually supports.

Limitations: Perception Cannot Substitute for Genuine Capability

The case this brief has developed for prioritizing perception factors requires an important qualification the technology acceptance research itself does not contradict but also does not fully address: perceived usefulness and perceived ease of use predict adoption of a given technology, but they do not predict whether that adoption actually produces genuine performance benefit if the underlying technology's objective capability is genuinely inadequate for the task. An organization that successfully drives high perceived usefulness and high adoption of an AI tool that is objectively poorly suited to the tasks employees are using it for has achieved adoption without achieving the actual performance benefit AI investment is meant to produce, a failure mode distinct from but not addressed by this brief's central argument.

The practical resolution is that perception-focused readiness investment and genuine capability investment are complementary requirements, not substitutes for each other: capability investment ensures the tool is objectively worth adopting, while perception-focused investment ensures that objectively worthwhile capability actually gets used. Organizations that invest exclusively in one while neglecting the other risk two distinct and different failure modes, technically excellent tools nobody adopts, or widely adopted tools that do not deliver genuine performance benefit, and genuine AI readiness requires building both capabilities deliberately rather than treating either as sufficient on its own.

Synthesis

The technology acceptance research this brief has reviewed identifies a specific, well-evidenced gap in how most organizations currently approach AI readiness: substantial investment in the technical infrastructure layer, and comparatively minimal deliberate investment in the perception factors, perceived usefulness and perceived ease of use, that Davis's and Venkatesh and Davis's research established as the actual predictors of whether employees adopt and use the resulting capability. Organizations building genuine AI readiness need to measure and invest in both dimensions deliberately, technical capability sufficient to produce genuine performance benefit, and the social influence, job-relevance communication, and supportive onboarding that this research identifies as actually shaping whether employees perceive that capability as worth adopting in the first place.

References
Explore the Evans Learning Labs diagnostic toolkit
Structured diagnostic tools covering leadership, team, and organizational performance.
Browse the toolkitBack to articles
Welcome back
Sign in to access your assessments
No account?
Terms of Use and Disclaimer

Informational and Educational Use Only

The diagnostic tools, assessments, profiles, and indexes offered by Evans Learning Labs are designed for informational and educational purposes only. Results do not constitute professional consulting advice, legal advice, psychological assessment, clinical evaluation, or any form of certified professional guidance.

Self-Reported Results

All results are based entirely on the responses provided by the individual completing the assessment. Evans Learning Labs makes no representation that scores or profiles accurately reflect objective organizational conditions or any other measurable external reality.

No Guarantee of Outcomes

Evans Learning Labs does not guarantee that use of these tools will produce any specific organizational, leadership, or performance outcome. Recommendations are general in nature and may not be appropriate for every individual, team, or organizational context.

Limitation of Liability

To the fullest extent permitted by applicable law, Evans Learning Labs, its principals, employees, and affiliates shall not be liable for any direct, indirect, incidental, consequential, or punitive damages arising from the use of or reliance on these tools or their results.

Governing Law

These terms are governed by the laws of the United States and Commonwealth of Kentucky.