Enterprise AI expectations shifted from team collaboration to individual assistants between 2023 and 2025

Enterprise AI in workplace software development has taken a fundamentally different shape than what practitioners anticipated two years ago, according to a new research study from arXiv. The mismatch highlights how the actual trajectory of emerging technologies can diverge sharply from user expectations—with significant implications for future product design.

What the researchers found#

Researchers conducted interviews with 15 software developers and project team members at an enterprise organization in 2023, then re-interviewed 10 of those same individuals in 2025. The contrast between the two time periods reveals a striking reversal in how AI tools actually arrived versus what participants hoped for.

In 2023, team members envisioned enterprise AI systems that would operate at a collective level—tools designed to integrate signals across various workplace applications, identify coordination gaps within teams, and intervene directly in group processes. The focus was on solving shared organizational challenges.

What actually happened#

By 2025, no such team-level systems had materialized in their workplace. Instead, personal AI assistants dominated the landscape. Individual employees gained access to standalone AI tools designed for single-user productivity rather than collaborative workflows.

How expectations diverged#

The researchers identified three distinct patterns in how early hopes evolved over the two-year span:

• Some initial expectations were deferred—postponed rather than abandoned, suggesting ongoing belief in eventual development
• Others were actively rejected as participants came to view them as either undesirable or technically infeasible given real-world constraints
• A third category saw expectations fade from conversation entirely—ideas that practitioners no longer felt compelled to articulate as priorities

Why this matters#

The study underscores how design possibilities are neither fixed nor predetermined. As technologies develop and market forces shape product decisions, the range of feasible futures can expand, narrow, or close entirely. For organizations planning enterprise AI initiatives, this research suggests that the gap between what users say they need and what actually gets built can be substantial—and that understanding this dynamic is essential for responsible technology design.