Okta data shows enterprises using multiple AI platforms

Okta has published data showing enterprises are increasingly using multiple AI platforms rather than relying on a single provider. The findings are based on analysis of application access patterns across more than 20,000 organisations on its platform.

Its Enterprise AI Index found that most enterprise customers now use several AI platforms at once. Exclusive single-platform deployments fell by more than 1.3% in June, suggesting businesses are broadening their AI estates as adoption moves beyond early trials.

The analysis covered more than 100 AI products across six workflow categories: foundational AI models, software development, workplace productivity, enterprise search, collaboration and creative tools. It pointed to a mix of AI-native applications such as Anthropic's Claude, OpenAI and Cursor, alongside AI features embedded in workplace software including Microsoft Copilot, Google Gemini, GitHub Copilot, Slack AI, Adobe and Notion.

That pattern suggests companies are choosing different tools for different functions rather than committing to a single technology stack. AI-native applications are seeing rapid account growth, while established software providers are driving adoption through products already used across the workplace.

Shift in use

The report also highlighted what Okta called the "Agentic Paradigm", describing a shift from chatbot-style assistants to AI agents that can carry out tasks across applications and workflows. That change raises questions about how companies monitor and control software agents as they gain access to systems and data.

Mike Reddie, Vice President and General Manager, ANZ at Okta, said organisations are changing how they approach deployment. "We are seeing organisations move beyond experimenting with AI to building AI environments that combine different models, applications and AI agents," he said.

"Different AI tools are being adopted for different business outcomes. The focus is increasingly on maintaining the flexibility to use the right AI capability for each task, rather than building around one ecosystem."

Identity management is emerging as a central issue as these systems spread across workplaces. Companies must decide not only how an AI agent proves its identity when connecting to software, but also what data and actions it should be allowed to access once inside.

Reddie said the challenge becomes more acute as AI tools interact with more systems at once. "As AI agents become connected to multiple applications, systems and data sources, organisations need visibility into what those agents can access, what they can do and how they are governed. Identity provides a consistent way to manage that access and accountability across increasingly diverse AI environments."

Interoperability push

The findings also point to a broader industry push for interoperability in identity and access controls. Okta cited support for Cross App Access, an open protocol intended to help organisations apply consistent controls as AI agents operate across different applications and ecosystems.

According to Okta, more than 25 technology companies support the initiative. The group includes Australian-founded Canva and Atlassian, as well as Anthropic, Cloudflare, Slack and Zoom.

The case for common approaches is gaining ground as businesses adopt software from multiple vendors and begin connecting AI agents to those systems. In such environments, fragmented identity rules can make it harder to see what non-human actors are doing and to limit their permissions across platforms.

Australian gap

Separate audience polling conducted at Okta AI Identity Summits in Australia suggested governance is lagging behind interest in AI agents. Nearly half of Australian technology leaders surveyed, or 48%, described agentic AI as a high organisational priority.

Yet only 13% said they could restrict or shut down an AI agent if required, while just 9% reported having full visibility of the AI agents and non-human identities operating across their environments. The figures suggest that while organisations are moving quickly to explore AI agents, many still lack the controls needed to manage them once deployed.

The combination of wider multi-platform adoption and limited oversight of non-human identities is likely to sharpen scrutiny of access management as AI use expands. For companies rolling out several AI services at once, the practical issue is no longer only which model to use, but how to govern a growing network of tools and agents operating across business systems.

The Enterprise AI Index suggests that trend is already under way, with most enterprise customers in Okta's data using more than one AI platform at the same time.