No Priors

Ofir Ehrlich & Gonen Stein on why data is the only moat

Ofir Ehrlich and Gonen Stein· Co-founders of Eon at Eon
·~35 min·English·No Priors
AI InfrastructureAgentsAI SafetyBusiness Strategy
TL;DR

Eon's founders argue that in the AI era models and compute are commodities with near-zero switching cost, so the durable moat is the data a company already owns — if it can find, classify, protect, and activate that data before agents turn it into a security problem.

01Core Mental Model

Models Are Ephemeral, Data Is the Moat

<strong>Models, compute and tools are ephemeral — the data you own is the only moat.</strong> Switching between models costs almost nothing, so no model is a lasting advantage.

The most valuable thing that you have is actually your data. And you see more and more companies finding this out.

— Ofir Ehrlich & Gonen Stein, No Priors
Key Insight
The claim isn't that models don't matter — it's that they don't lock in. When moving from one model to the next costs almost nothing, no model is a durable advantage; the only asset with real switching cost is the data a company has accumulated, which is why Eon reframes a backup product as a data moat.
02Market Signal

They Didn't Buy Airplanes. They Bought the Data.

<strong>Google paid $10M for a bankrupt airline's data, not its planes.</strong> When AI buyers bid against each other for enterprise records, data has become a tradeable, training-grade asset.

They didn't buy airplanes. They bought the data. They bought the data for $10 million because they think it's very important in that perspective. They're using that to train models.

— Ofir Ehrlich & Gonen Stein, No Priors
Key Insight
The tell isn't the price, it's the bidding. Two AI buyers competing over a dead airline's records means the market now prices raw enterprise data as training-grade supply — and the scarcity is real-world behavioral data (why the public Enron corpus is still mined), which synthetic data cannot fully replace.
03The Real Bottleneck

The Data Exists — It's Just Locked

<strong>The data already exists inside the enterprise — it's just locked.</strong> The real problem isn't missing tools; it's silos, conflicting incentives, and servers no one dares turn off.

You were tasked with doing that. I'm tasked with making sure my systems work and I'm tasked with making sure that data is intact.

— Ofir Ehrlich & Gonen Stein, No Priors
Key Insight
The blocker was never a missing tool — the data was already sitting in the enterprise. It's locked by org structure, not technology: the person who owns the data and the person who needs it answer to different mandates, so the friction is political before it is technical, which is what Eon's map-classify-mask layer is really unblocking.
04Security

The Same Threat, Now From Your Own Agents

<strong>The same kind of threat now comes from agents with legitimate permissions.</strong> Detection methods barely change, but the velocity of an insider agent is extreme.

the same type of threat is coming from non-human actors, agents that essentially have legitimate access to the environment with legitimate permissions.

— Ofir Ehrlich & Gonen Stein, No Priors
Key Insight
The defense barely changes — the same anomaly detection that caught ransomware can catch an agent suddenly dropping a table. What changes is the clock: an approved agent acts with valid credentials at extreme velocity, so 'assume breach, malicious or not' stops being paranoia and becomes the default posture.
05Democratization

Everyone Is a Builder Now

<strong>Everyone in the org is now a builder — and that's a good and a bad thing.</strong> Non-technical staff and agents-spawning-agents handle sensitive data outside the org's rules.

It's a good thing and bad thing that everyone inside organization can become builders.

— Ofir Ehrlich & Gonen Stein, No Priors
Key Insight
Democratized building quietly dissolves the security perimeter. Every non-technical builder and every agent-spawning-agent is a new identity handling company data outside the org's rules; the governance problem isn't the tools people adopt, it's that no one can even enumerate the actors anymore — which is why non-human-identity is now a top-tier security category.
06Unlocking Value

Burgers, Pizza, and the Answer You Couldn't Compute

<strong>Siloed data hides answers you could never compute.</strong> Unify it and add context, and questions that were impossible across separate teams suddenly become trivial.

Let's say that there's one person in organization who have all the list of all the people in New York who love burgers and another person in the organization have who has a database of all the people in New York who love pizza.

— Ofir Ehrlich & Gonen Stein, No Priors
Key Insight
Siloing doesn't just hide data — it hides the questions. The valuable answer (who loves both) never existed in either database; it only appears once the data is unified and given context. Integration therefore creates new analytical possibilities, not merely cleaner versions of the old queries.
07Cloud vs AI

Cloud Was Abstract. AI Is Visceral.

<strong>Cloud was abstract; AI is visceral — so adoption moves far faster.</strong> Everyone felt the ChatGPT moment, and boards now push companies to adopt AI immediately.

Cloud is basically just someone else's computer, but who knows what it is. It's hard to explain to my grandmother about the cloud AI. Everyone understands AI.

— Ofir Ehrlich & Gonen Stein, No Priors
Key Insight
Adoption speed tracks how tangible a technology feels, not how powerful it is. Cloud stayed abstract and demanded heavy technical and human effort; the ChatGPT moment made AI legible to boards and shareholders at once, accelerating adoption 'on steroids' — even as the fear of losing control makes customers pause.
08Go-To-Market

A New Way to Consume Software

<strong>Companies now consume software in brand-new ways — PLG, forward-deployed engineers, and buy-and-transform.</strong> Some acquirers just buy a company and turn it into an AI company outright.

instead of you adopting AI I know how to do it more efficiently. If I can buy the company and transform that into an AI company we can all win.

— Ofir Ehrlich & Gonen Stein, No Priors
Key Insight
The go-to-market is mutating with the technology. The same founders who argued PLG doesn't work for dev tools now see PLG, forward-deployed engineers, and outright acquisition ('buy the company, make it an AI company') as parallel routes — a sign the constraint has shifted from selling software to installing transformation.