What if Your Executives and Managers Are Seeing Two Different AI Realities?
Executives are becoming more optimistic about AI. Middle managers are becoming more cautious. And both may be responding completely rationally to what they see.
That tension sits at the center of a recent Harvard Business Review study by Jeremy Korst, Stefano Puntoni, and Prasanna Tambe. The data exposes an extraordinary perception gap between leadership tiers:
- Delivering Significantly Positive ROI: 45% of executives agree vs. only 27% of middle managers.
- Adopting Faster Than Competitors: 56% of executives agree vs. only 28% of middle managers.
These leaders work inside the exact same organizations. How can two levels of leadership look at the same deployment and see radically different outcomes?
They are not disagreeing on facts. They are experiencing different parts of the operating system.
They Experience Different Operating Realities
Senior leaders encounter AI where it performs at its best: high-level synthesis, drafting strategic memos, scenario planning, and executive decision support. Middle managers experience AI where work gets messy and operational friction peaks.
The Executive View: Possibility + Selected Wins + Aggregate Adoption Metrics + Capital Investment
The Manager View: Operational Exceptions + Defect Rework + Workflow Friction + Integration Bottlenecks
When an AI tool succeeds, both tiers celebrate the upside. When it hallucinates, breaks downstream processes, or outputs inconsistent work, middle managers bear the full burden of remediation.
AI Governance Depends on What Travels Upward
This gap is far more dangerous than simple corporate misalignment—it is a broken sensing loop. Senior executives decide capital allocation, workforce restructuring, operating speed, and portfolio scaling. Yet the crucial operational telemetry needed to calibrate those decisions lives on the frontline:
- Where model outputs repeatedly break down
- Where manual rework cancels out initial time savings
- Where teams lack confidence in deterministic accuracy
- Where review capacity is failing to keep pace with output
- Where leadership assumptions clash with day-to-day workflow realities
If those operational signals remain trapped locally, executive decisions are based on an incomplete, artificially optimistic picture.
Broken Feedback Creates Cascading Failures
When leadership misinterprets frontline hesitation, it triggers a self-reinforcing cycle of bad strategy:
Optimistic Executive Assumptions → Accelerated Rollout → Increased Operational Friction → Deepened Managerial Caution → Wider Perception Gap → Distorted Capital Allocation
Believing ROI is proven, leadership scales spend prematurely. Assuming adoption is smooth, they starve middle management of review and verification tooling. Mistaking operational caution for stubborn cultural resistance, they double down on mandate pressure instead of diagnosing root-cause bottlenecks.
Managerial Caution Is Diagnostic Information
In a transformation, cautious managers are frequently labeled as obstacles. In reality, managerial caution is vital diagnostic data.
As the HBR authors observe, managers operate close to AI’s current boundary constraints: hallucinations, brittle integrations, workflow mismatch, and quality assurance overhead. Executives, conversely, are incentivized to imagine broad strategic horizons.
Executives are paid to see possibility. Managers are paid to make reality work. The objective is not to eliminate that tension—it is to build an architectural bridge between them.
Middle Managers Are Local Sensors of Decision Architecture
Middle managers do more than locally execute enterprise strategy into permissions, review gates, and workflows—they act as the primary sensing network of the enterprise. They observe exactly where broad AI strategy collides with real human constraints, data quality issues, and unforeseen error costs.
Executive Strategy → Scaled Rollout → Operational Deployment → Manager Observation & Diagnostic Feedback → Executive Learning → Calibrated Capital Allocation
If the loop terminates at frontline observation without traveling upward, enterprise learning halts. Local workarounds multiply, failed assumptions persist, and the organization scales fragility rather than capability.
Usage vs. Readiness vs. Value
Tracking logins, licenses, and prompt volumes confuses tool activity with enterprise maturity. A comprehensive framework must distinguish between three distinct operating stages:
Usage: Tells you whether employees are interacting with the tool.
Readiness: Tells you whether the organization has the governance, verification capacity, and workflow design to use it responsibly.
Value: Tells you whether executing the work with AI produced an economically superior business outcome.
A company can report 90% software adoption while maintaining near-zero readiness and negative economic value.
You Cannot Govern What You Cannot See
When operational reality fails to reach the boardroom, executive confidence expands precisely as ground-truth information quality degrades. Leadership retains the spending authority to scale deployments, but lacks the visibility required to scale them effectively.
Enterprise AI transformation does not fail for lack of executive vision. It fails for lack of a mechanism that turns frontline operational experience into executive learning.
True decision governance ensures that the real costs, friction points, and failure modes observed by managers flow directly into the systems governing future AI investments.