Nudge Security Research

AI Adoption in Practice 2026

In the second edition of our report, we use aggregated and anonymized telemetry from Nudge Security customer environments, not surveys, to show where AI risk lives now: in the agents, integrations, and tools your employees adopt at the Workforce Edge.

Nudge Security SaaS asset discovery

Finding 1

Many agents are running on your employees' identities.

Among organizations using AI agents, half run 7 or more, and more than 4 in 10 run more than 10. These agents run inside platforms your teams already use, like Atlassian, Zapier, and Cursor.

Many act through the credentials of the employee who built them, so they inherit that person's access to company data. Your logs show the employee, not the agent.

Nudge Security SaaS asset discovery
Nudge Security SaaS asset discovery

Finding 2

AI connects directly to your most sensitive systems.

74% of organizations have at least one integration with Anthropic, and 67% with OpenAI. These connections reach Google Workspace, Microsoft 365, GitHub, and Slack.

These are typically OAuth grants that move data from one cloud service to another without crossing your network, so AI gateways and DLP never see them. Each grant keeps its access until someone revokes it.

Finding 3

AI tools spread faster than any approved list.

The typical organization we observe has had 27 different AI tools show up in its environment. Many overlap, like Otter.ai and Read.ai for meeting notes.

The most popular tools keep spreading, too. In January, five AI tools were found in more than half of organizations. Now 12 are.

Nudge Security SaaS asset discovery
Nudge Security SaaS asset discovery

Govern what AI can reach.

An approved list covers the part of AI adoption that's already settled. Your risk sits in what those tools connect to and what acts on your employees' behalf. Download the full report for all six findings, the data behind them, and five steps you can take now.

Thanks for downloading our report!

Here's your copy.