Measuring what matters: GenAI Effectiveness 2Q metrics

31% of employees now use GenAI at work. Only about 5% are using it effectively enough to create meaningful productivity gains. Our new research report digs into why — and what to do about it.

"We are encouraging our employees to use GenAI apps — but are we more productive?" It is one of the most common questions we hear from leadership teams, and one of the hardest to answer.

We sampled 4,800 business users and 139,000 GenAI interactions across a typical enterprise mix of knowledge-worker profiles during 2Q26, and the results are now published in our latest research report.

The key finding: GenAI usage is way ahead of effectiveness

Active GenAI usage nearly doubled in a single quarter, from 19% to 31% of eligible employees, and the number of app families in use more than doubled, from 19 to 53. By any adoption measure, GenAI has crossed from a curious minority into the mainstream of daily work.

Effectiveness tells a different story. Segmenting employees by how often they use GenAI and how skillfully they prompt shows that only about 5% qualify as truly effective users — daily or weekly users prompting at a high skill level. The rest are either not using GenAI at all or using it in ways unlikely to move any productivity needle. Overall, the prompting skill-level is polarizing rather than improving evenly: 89% of prompts are still simple, Google-search-style questions, even as the share of employees reaching expert-level skill jumped from 28% to 48%.

Governance has not kept pace with any of this. More than half of active users are still on free or personal app plans the organization cannot see into, and 23% of uploaded attachments — increasingly screenshots and documents, not just text prompts — are going to those same free and consumer apps.

Why this matters

The report's key finding is that frequency and prompting skill, not access to AI, are what actually drive productivity. An organization can roll out enterprise licenses to everyone and still see almost no effectiveness gain if usage stays occasional and prompts stay basic. That reframes the governance question: instead of asking "how do we get more people onto GenAI," the better question is "how do we get the people already using it to use it well, safely, and often."

The report lays out a practical framework for closing that gap — productivity-first governance — built around allowing broad access while maintaining visibility, using guardrails as a safety net rather than a barrier, surfacing what top users already do well, and building prompting skill team by team.

Get the full report

The full research report breaks down all of the 2Q26 benchmarks (in comparison to 1Q 2026) in detail — adoption and weekly usage, the app landscape (including how Microsoft Copilot, ChatGPT, and Claude are actually being used), the prompting-skill level, data-exposure trends, and the GenAI Effectiveness Index methodology — along with the productivity-first governance framework for closing the gap.

Download the full research report to see:

• The latest enterprise GenAI usage and effectiveness benchmarks, including 31% active vs. 5% effective usage
• Why frequency and prompting skill — not GenAI access — are the biggest drivers of productivity
• A practical framework for measuring and improving GenAI effectiveness across your workforce, following best practices with productivity-first governance

This is part 3 of the Measuring What Matters blog series. Read Part 1 and Part 2.

‍