Okta Index Tracks Multi-Vendor AI Adoption Across 20,000 Businesses
Okta's Enterprise AI Index analyzed anonymized SSO access data from more than 20,000 organizations between June 2022 and June 2026, finding that most enterprise customers now use more than one AI platform. The report tracks 109 products consolidated into 74 suites and says agentic tools are widening identity risks as credentials, permissions, and audit trails spread across vendors.
Okta released an enterprise AI adoption index on July 16 based on anonymized application-access data from more than 20,000 organizations. The four-year dataset shows companies assembling multi-vendor AI stacks as the market shifts from autocomplete and chat toward tools that can act across business systems.
What the index measured
The analysis covers Okta Workforce Identity activity from June 2022 through June 2026. Researchers tracked 109 AI products, tools, and aliases, then consolidated them into 74 application suites across six categories: foundation models, developer tools, enterprise search, collaboration and productivity, meeting AI, and creative software.
The scope matters. The dataset reflects applications delivered through Okta single sign-on and excludes Auth0, model access through major cloud platforms, AI infrastructure, cybersecurity AI, and AI embedded inside data platforms or systems of record. Okta also cautions that behavior on its own platform may not represent global enterprise adoption.
Multi-vendor stacks become the norm
Okta found that a majority of enterprise customers used more than one AI platform in June 2026. The share using only one provider fell 1.2 percentage points from May as companies added tools for different jobs.
The report classifies vendors whose corporate customer bases grew more than fourfold as AI-native. Within that group, Anthropic passed OpenAI in enterprise accounts in March 2026 and in monthly active users the following month, according to Okta's underlying usage data. Scale still favors established suites: Anthropic had less than half Microsoft 365's account volume and less than one-tenth of its monthly active users.
Agents turn adoption into an identity problem
The index dates the start of an agentic phase to May 2025, when coding tools began executing longer chains of work with less prompting. More autonomous software also means more connections to internal systems, each carrying permissions, tokens, or service credentials.
In reporting tied to the index, Okta researcher Fei Liu told Help Net Security that organizations still authorize agent workflows with static API keys, service accounts, and sometimes shared human logins. That last pattern makes it harder to tell whether a sensitive action came from an employee or an automated system.
For engineering and security teams, the operational takeaway is narrower than choosing a winning model vendor. Every agent needs a distinct owner, scoped access, credential rotation, and an audit trail that survives delegation across tools. Okta's data shows why those controls must work across a portfolio rather than one preferred AI platform.
Key Points
- 1Okta analyzed anonymized SSO activity from more than 20,000 organizations, covering June 2022 through June 2026.
- 2Most enterprise customers used multiple AI platforms by June 2026, while the single-provider share fell 1.2 percentage points from May.
- 3Agent workflows still rely on service accounts, static API keys, and sometimes shared human logins, creating ownership and auditability gaps.
Scoring Rationale
The index provides a large, behavior-based view of enterprise AI adoption and a useful signal on multi-vendor and agentic deployment patterns. Its reach is meaningful for platform and security teams, but the data comes from one identity vendor's customer base and does not measure the entire market.
Sources
Primary source and supporting public references used for this report.
Practice with real Ad Tech data
90 SQL & Python problems · 15 industry datasets
250 free problems · No credit card
See all Ad Tech problems
