The Visibility Gap

Two in three of the 467 working professionals surveyed say the changes in how people work because of AI are producing meaningful value. Only one in five reports that their organization has a formal method for measuring the business value of its AI investments. Another 25% say one is in development, 22% say there is none, and 31% do not know.

The gap is not confined to skeptics. Among respondents who agree AI is producing meaningful value, only 28% report a formal method, and 26% do not know whether one exists. Confidence follows the method. Respondents who report a formal method average 3.87 on a five-point confidence scale, against 2.21 where there is none.

  • 67%

    agree AI is producing meaningful value for their organization.

  • 21%

    report a formal method for measuring it.

  • 31%

    do not know whether one exists.

  • 25%

    are very or extremely confident their organization knows whether its AI investments are delivering.

  • What the Study Measured

    Most AI measurement counts what is easy to count: licenses, logins, queries and estimated hours saved. This study asks a different question. Do people believe AI is producing value, does anyone have a way of verifying that belief, and do the people using AI and the people overseeing it see the same thing?

    The survey asked individual contributors and Managers & Executives identical statements about their own work group, fielded online from September 30 to October 3, 2026. The report uses the term behavioral value for the value an organization realizes through changes in how people work, net of the behavioral costs of verification, workarounds and avoidance. The survey does not measure behavioral value directly, and it does not measure financial return. It measures perceived value and the conditions that tend to accompany it.

What the Report Finds

  • Time Saved is Not Time Used

    Eighty-one percent of respondents agree AI helps people in their work group complete tasks faster, but 68% agree the time saved is redirected toward higher-value work. The difference is statistically reliable overall and at both job levels. The report calls this pattern leakage, and it describes perception rather than measured lost hours. The link to perceived value is strong: among those who agree AI saves time, 84% of those who also see it redirected rate AI’s overall value favorably, against 51% of those who do not.

  • Line of Sight Depends on Where You Sit

    Three in four Managers & Executives (75%) agree it is clear how AI use in their work group contributes to outcomes that matter to the organization, against 56% of individual contributors. It is the largest level difference of the ten statements measured. It holds when respondents are split by whether they have been involved in an AI adoption decision, and by whether they recognized a genuine AI governance term, so it is not simply a matter of who has been close to the decisions.

  • Individual Contributors Often Do Not Know Whether Value is Measured

    Forty-two percent of individual contributors do not know whether their organization measures AI value at all, against 17% of Managers & Executives. Even among Managers & Executives, only 28% report a formal method.

  • The Costs of AI Look Alike at Every Level

    The burden of verifying AI outputs and worry that AI threatens job, status or expertise are rated almost identically by individual contributors and Managers & Executives. Only about three in ten at either level disagree that these costs exist. A comparison between levels will not reveal them, and neither will usage data.

  • Experience Helps, But Not Enough

    The share reporting a formal method rises with how long the organization has used AI, from 8% under one year to 20% at one to two years and 39% at three to five years. Even in the longest-tenure group that can be interpreted, three in five do not report one. Tenure is also confounded with organization type, so the pattern should not be read as the effect of time alone.

  • What Closes the Gap

    Four moves separate organizations that can show what AI returns from organizations that rely on belief. None is technically difficult. All require someone to decide that they matter enough to fund.

    Decide what evidence would look like before measuring it. Agree in advance what would count as evidence of AI value, beyond usage and estimated hours saved. This is the step that makes the other three possible.

    Give saved time an owner. Unless someone decides what a recovered hour is for, the saving can be absorbed rather than redeployed. Name who decides, and check where the time goes.

    Ask both levels the same question, and compare the answers. Individual contributors and Managers & Executives answered identical statements and saw different things. A single-vantage measure, whether a leadership dashboard or a staff pulse survey, will miss the difference.

    Measure the behavioral costs directly. Verification burden and worry about status look alike at both levels, so they have to be asked about. They will not surface in usage data.

  • What the Study Can and Cannot Say

    This is a convenience sample of self-reported perceptions, drawn mainly from the United States (83%) and weighted toward healthcare (37%). It is cross-sectional, so it shows associations and cannot show that one condition causes another. The sample size was not fixed in advance, and interim results were reviewed during fielding. Managers & Executives are mostly managers, so the level comparison speaks to staff versus managers more than to staff versus senior executives.

    The study shows that a visibility gap exists across a broad set of organizations and suggests where it tends to appear. It cannot show whether any particular organization has it. That requires measuring behavioral value inside the organization, by level and function, and tied to specific AI uses.

  • Where to Go From Here

    Each finding corresponds to a question an organization can answer about itself.

    Could we produce evidence that a specific AI-assisted decision received meaningful human review? BOLD is a single-decision audit, completed in three to five business days.

    Does our governance architecture produce the oversight behavior it is designed to? B-GRIT is a governance-system diagnostic that tests whether oversight works in practice, not only on paper.

    Is AI adopted the way leadership believes it is across the organization? BEAR is a deployment-wide adoption diagnostic, delivered in 21 days, that compares what leaders see with what the people using AI experience.

    Will pilot-level trust in AI survive as it scales? STAR is a scale-up trust diagnostic for organizations moving from pilot to enterprise-wide use.

  • Research Approach

    Behavieural fielded an original survey, screening for professionals whose organization currently uses AI or automated decision-support tools, across healthcare, finance & insurance, and manufacturing & operations. Nineteen core statements across five categories were measured on a five-point agreement scale, alongside two behavioral outcomes, usage frequency and intent to continue, combined into a single composite measure.

    Regression analysis, which isolates each category's independent contribution once the other four are accounted for, was used throughout rather than simple agreement rates, since respondents tend to rate every category somewhat positively together, a pattern that can mask which factor is actually connected to behavior. The report's methodology section addresses this and several other likely criticisms directly, including the perception-based measurement approach and the iterative nature of the data collection.