Building a Venture Scale Cyber Security Startup in the Age of AI: Alon Jackson, CEO of Astrix Security (Part 5)

Sramana Mitra: So, two questions on AI. First one is, as you said, there is a tailwind, clearly, because agentic AI this year in 2025 has started to take off, and we are doing this interview in June 2025. We have started hearing about agents quite aggressively all through this year.

So, the vulnerabilities caused by agentic AI are already becoming a bigger opportunity than what you had originally set out to work on. Now, my question is very specific. In your enterprise customers, is agentic AI an issue yet?

Alon Jackson: For the majority of them, yes.

Sramana Mitra: They already have agents running?

Alon Jackson: They have, or are building, or are adopting, or have an AI center of excellence or a program around it. In the modern world, if you’re not building, equipped, or already using not only agentic but AI infrastructure and tools, you’re going to lose the market. You’re going to lose the race.

Sramana Mitra: There is definitely a lot of AI labs inside enterprises. But your product kicks in when there’s actual agentic automation implemented. That’s when the vulnerabilities open up, and that’s when you need a cyber security strategy to plug those vulnerabilities.

That’s my question—where in the continuum of time cycles are we, in terms of vulnerabilities opening up inside enterprises because of agentic AI? Because that’s really your opportunity.

Alon Jackson: There are two types of AI trends. One is AI apps—like you said, different labs are building LLMs and applications on top of them. The second is the agentic wave, which adds a dimension of autonomy and connectivity to these AI apps. Basically, agents are AI-enabled software that can make decisions and operate either on your behalf or autonomously.

Both trends are a huge tailwind in terms of non-human identities and managing access and connections that are non-users. Some might be apps that are now AI-enabled, and others might be agentic software with more autonomy. In both cases, you’re enabling these non-human software agents access to different resources in the organization. As we said, there’s more to gain and more to lose from exposure.

Sramana Mitra: More integrations.

Alon Jackson: More integrations, exactly. And it’s not just about data leakage or exposure. There’s also operational risk when you think of a service that can transfer money, send an email, or control production on your behalf.

These risks are about operational continuity and business continuity. As we delegate and build more capabilities for these services, we’re actually giving them more trust. That means we need to understand who or what is getting that trust and how we’re managing the lifecycle of that entity, similar to how we manage high-permission employees like admins.

Sramana Mitra: So, my second question—remember I said I had two very specific questions—is on the R&D side. What are you doing to leverage AI on the product side? On the software side?

Alon Jackson: A bunch of things, in two major layers.

One is AI-enabled features and capabilities that enhance our product and our offering to our customers. The second is leveraging AI to build the product in the first place.

Starting with the first—both in personal life and enterprise, we’re becoming more used to chatting with our apps and interacting in a natural language, rather than navigating complex UIs. You can ask for something high-level and expect a reasonable answer. That’s one obvious AI feature.

Another one—which we today call AI but used to just call machine learning(ML)—is an ML engine that detects anomalies in non-human identity behavior. This has been training for years on thousands of API connections daily and can detect anomalies across the enterprise. That’s a major piece of intellectual property. So, that’s the capabilities and features that our customers enjoy.

The second part is leveraging AI to build a lot of product components and make our R&D team more efficient. This translates into value for customers in terms of speed in releasing new features and responsiveness.

Sramana Mitra: Double-click on that and tell me a bit more about where you are seeing the productivity gains.

Alon Jackson: It’s really across the board. Even HR is using AI. Instead of going to HR for questions, we have an internal HR chatbot that can guide you from onboarding through ongoing processes. That project was built by our HR department.

Marketing and sales also have an abundance of tools today. But going back to R&D, the interface between product and engineering is shifting. Instead of a product manager writing a detailed product definition document to hand over to an engineering team, they can now interact with an AI tool that builds it for them, shows them what they asked for, and allows them to iterate.

So, the boundary between product manager and engineering team is shifting. Engineers now receive a more “baked” product—code generated by an AI tool, maybe a UI wireframe that reflects the product manager’s intent. This saves a lot of human cycles, which used to require back-and-forth.

On the engineering side, it applies to scaling systems, testing, implementing support for new features, and replicating solutions across environments. We’re training AI tools to understand what our company does and how we implement it, then applying that to new scenarios. That saves orders of magnitude in development cycles.

This segment is part 5 in the series : Building a Venture Scale Cyber Security Startup in the Age of AI: Alon Jackson, CEO of Astrix Security
1 2 3 4 5