No PriorsRethinking Legacy Data Infrastructure with Eon Co-Founders Ofir Ehrlich and Gonen Stein
At a glance
WHAT IT’S REALLY ABOUT
Backup becomes AI data foundation as agents reshape enterprise security
- Eon’s founders describe building a cloud “data foundation” that maps, classifies, ingests, and efficiently stores enterprise data to enable both disaster recovery and AI/LLM workflows on top of the same consolidated corpus.
- They argue that as models and compute commoditize, proprietary enterprise datasets—especially long-term historical and operational data—become the enduring moat, citing Google’s reported purchase of Spirit Airlines’ data from bankruptcy proceedings.
- The conversation emphasizes that enterprises struggle less with AI tooling and more with locating scattered data, negotiating cross-business incentives, and enforcing security/compliance while extracting data without impacting production systems.
- They warn that security threats are shifting from human attackers to autonomous agents with legitimate permissions, increasing the speed of destructive events (e.g., dropped tables) and raising the importance of rapid detection and granular recovery.
- They compare the AI era to the cloud migration wave but “on steroids,” with faster transformations, new go-to-market motions (forward-deployed engineering and PLG), and heightened demand for control and observability as agents and non-technical builders proliferate.
IDEAS WORTH REMEMBERING
5 ideasEon reframes backup as the fastest path to an AI-ready enterprise data layer.
Eon positions itself as a cloud-based “data foundation” that discovers data across hyperscalers, classifies sensitivity, ingests structured/unstructured sources, and then supports protection/recovery plus search/query for AI workloads.
In the AI era, enterprise data becomes the primary competitive moat.
They argue that models and compute are increasingly interchangeable, while proprietary enterprise data (including long-tail historical data) becomes the durable differentiator—illustrated by Google’s reported $10M purchase of Spirit Airlines’ data rather than physical assets.
High-fidelity, real-world operational datasets are emerging as premium training fuel for agents.
The Spirit dataset example is used to highlight a scarcity of realistic, high-quality “how a company works” data for training agents; many teams resort to limited public datasets or synthetic data that may not match real-world complexity.
The hardest AI problem in enterprises is not algorithms—it’s finding, understanding, and safely mobilizing data.
They describe a core organizational bottleneck: data teams are tasked with “use AI on our data,” but data is scattered across business units, poorly understood, risky to extract, and constrained by compliance/security and uptime concerns.
Agentic systems create insider-like threats at machine speed, making ‘assume breach’ operational.
Security risk shifts from primarily human-driven attacks (e.g., ransomware) to “non-human actors” (agents) with legitimate permissions that can rapidly delete/alter data; mitigation relies on similar detection/recovery patterns but at much higher velocity.
WORDS WORTH SAVING
5 quotesUp until now, the concerns came from human threats. What we're seeing now on steroids is that the same type of threat is coming from non-human actors, agents that essentially have legitimate access to the environment with legitimate permissions.
— Gonen Stein
If you're a company, whether you're a hotel chain or you're a food chain, a technology company, doesn't matter, the most valuable thing that you have is actually your data.
— Ofir Ehrlich
They didn't buy airplanes. They bought the data. They bought the data for $10 million because they think it's very important.
— Ofir Ehrlich
We need to assume breach, whether it's malicious or not, and we need to be able to handle it and, uh, act accordingly.
— Ofir Ehrlich
So it creates a complete set of actors inside an organization, not bound by the rules of the organization, and not necessarily running within the premises of the organization, but handling sensitive data.
— Ofir Ehrlich
High quality AI-generated summary created from speaker-labeled transcript.
