The arrival of large language models did more than ignite another software cycle. For Okta Ventures, the corporate VC arm of identity authentication software producer Okta, it fundamentally altered the nature of identity itself.
The inflection point came in early 2024, when the firm’s investment team concluded that AI agents were becoming “first-class citizens” — entities capable of acting independently across enterprise systems and the consumer internet rather than simply serving as software tools. That seemingly technical distinction carried far-reaching implications. If agents were to operate with their own identities, permissions and interactions, they would require an entirely new layer of infrastructure to govern them.
That insight reshaped the venture arm’s investment thesis.
“We realised an entire ecosystem of tooling would emerge to manage the influx of agents operating both inside the enterprise and outside in the consumer internet,” says Austin Arensberg, vice-president and head of Okta Ventures. “Agents would need an entirely new infrastructure built up around them.”
Rather than viewing generative AI as another application layer, Okta Ventures began investing in what sits beneath it: the security, orchestration, governance and infrastructure required to make agentic systems viable at scale.
The shift demanded changes on two fronts. The first was external. The fund expanded its search for AI-native founders, embedding itself within developer communities through hacker houses, hackathons and technical events in an effort to identify emerging talent before conventional venture networks did. The second was internal, helping portfolio companies founded before the ChatGPT era rethink their products around large language models and agentic workflows.
Several early investments validated the approach. Backing companies including Browserbase, Prompt Security and Variance positioned the fund at the centre of rapidly emerging infrastructure categories, while subsequent investment priorities expanded into cloud infrastructure, memory architectures, agent orchestration, AI governance, web security and agentic commerce.
“The demand for agentic tooling has outstripped even our wildest assumptions.”
What surprised the team was not simply the breadth of demand, but its speed.
“The demand for agentic tooling has outstripped even our wildest assumptions,” says Arensberg.
The acceleration has blurred traditional venture-stage distinctions. Companies founded only two or three years earlier are now surpassing $100m in annual recurring revenue, creating opportunities that would once have been considered late-stage despite their youth. As a result, Okta Ventures has become increasingly willing to invest beyond its early-stage roots, participating in series D and E financings where strategic partnerships can accelerate commercial growth.
The experience has reinforced a principle that increasingly defines venture investing during the AI cycle: flexibility matters as much as conviction. “If there’s one lesson we’ve embedded over the past two years, it’s to maintain strategic flexibility with a bias to action.”
The race against time
That same compression is changing how corporate venture investors create value after writing a cheque.
Historically, portfolio support followed a relatively predictable cadence. Introductions to customers, product collaboration and strategic milestones often unfolded over the first year after investment. AI has largely eliminated that luxury.
Prompt Security became the defining example.