Gonen Stein

1. Key Themes

Enterprise Data Is the Last Remaining Competitive Moat in the AI Era

As AI commoditizes models and tooling, proprietary historical data becomes the only durable differentiator. Gonen Stein makes the case that switching costs on models are effectively zero, but data is irreplaceable.

"Models, everything is relatively ephemeral, almost zero switching costs. And those are infrastructure, important parts of the infrastructure for the industry. But if you are a company, whether you are a hotel chain or you are a food chain technology company, it doesn't matter. The most valuable thing that you have is actually your data." 00:03:02

Bankrupt Company Data Sets Are Becoming Acquisition Targets

Real-world enterprise data is so scarce and valuable for training agents that companies are buying it out of bankruptcy proceedings — a trend expected to accelerate significantly.

"Just two days ago, you saw Google buy something from the bankrupt Spirit Airlines. 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 it to train models." 00:03:29

"We've seen it for multiple use cases. That's what's really cool about it. And you see Mercor, you see other companies are continuously trying to already trying to buy data. If you're a tech data CEO today, I can tell that you constantly get the questions: are you willing to sell your data?" 00:04:30

AI Agents Are a New, More Dangerous Security Attack Surface

The threat from AI agents acting inside corporate environments with legitimate credentials is emerging faster than organizations can respond — and it is already causing real incidents at scale.

"What we're seeing now on steroids is that the same type of threat is coming from non-human actors, from AI agents that essentially have legitimate access to the environment with legitimate permissions into such and such databases. And all of a sudden, and this now happens very rapidly, a table is all of a sudden dropped." 00:00:00

"Six months ago, no one would even discuss with me. But a few months ago, pretty much every person I meet, every leader in a company tells me either they are afraid of that happening to them, or it personally happened to that person who was speaking with me." 00:17:01

Legacy Data Infrastructure Is Fundamentally Broken for the Agentic World

Tools like Fivetran and dbt were built for narrow, single-purpose data tasks. The explosion of agent-generated data and cross-organizational data needs has made that plumbing obsolete.

"The plumbing today is very limited and everyone built a solution to their set of problems... It wasn't so important before to have all the context, all the data for an organization because you could only do with the data things you really intended to do to begin with. So you had a single purpose in your mind when acting on the data. Today it's very different." 00:22:30

"Fivetran was an incredible company, dbt and so on and so forth, but very niche, very specific tools for that purpose." 00:25:31

Non-Human Identity (NHI) Has Become a Top-Tier Enterprise Security Problem

The proliferation of agents — including those built by non-technical employees — is creating an uncontrolled population of actors inside organizations that security teams cannot track.

"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 which is the property of the organization — could be exposed to the world, could be incorrect, could be incorrectly used." 00:00:21

"You see the proliferation of cybersecurity companies. How many cybersecurity companies you see in NHI, non-human identity, right now? An infinite amount. And there's a reason for that. It became a number one, number two problem right now." 00:19:23

The AI Adoption Forcing Function Is Fear, Not Just Opportunity

Unlike cloud, which was abstract and slow to permeate, AI is being mandated top-down by boards and CEOs under existential pressure — and that urgency is reshaping how software is sold and deployed.

"You're getting pushed by sea levels, by CEO, by the board, by the shareholders: use AI for the business, otherwise we're irrelevant. So you see people doing it both for the value that you get from AI and also for the fear that you get from AI." 00:29:57

Forward-Deployed Engineering Is Now Mainstream, Not a Services Penalty

A model once seen as embarrassing or niche (associated with Palantir) has become the dominant go-to-market motion for AI adoption in large enterprises.

"For the first time in many years you see how companies consume software in a brand new way. One example is what's happening with forward-deployed engineers — used to be something like a services penalty, were doing that, no one really understood what it means, and now everyone's doing that." 00:30:23

"The only way they can actually getting deployed and become AI much faster is by letting strong engineers who understand what they're doing and coming with a toolbox that they've created in top Silicon Valley companies, startups and sometimes larger companies, to come and transform those organizations." 00:30:52

AI Transformation via Acquisition Is an Emerging Enterprise Playbook

Rather than slowly adopting AI internally, some companies are buying enterprises outright and transforming them into AI-native operations — a radical new deployment strategy.

"There are really great companies, for example Longlake, that say: instead of you adopting AI, I know how to do it more efficiently. If I can buy the company and transport it up into an AI company, we can all win — we can create a higher margin, more efficiently. And this is a really radical new way for those companies to actually start using AI." 00:32:37

2. Contrarian Perspectives

Real-World Data for Agent Training Is Extremely Scarce — Synthetic Data Is Not Solving It

While the industry assumes synthetic data can substitute for real-world training data, Gonen Stein argues that existing real-world data sets are irreplaceable precisely because synthetic data doesn't capture actual organizational behavior.

"You don't find too many good data sets that don't look like real synthetic data that can actually be used to really look like the real world... It's very hard to find data that will help you to work like in the real world. Anytime you see someone building an agent or building a new application, you know, most of them don't really work. You have to go to the world." 00:07:15

He even cites the Enron data set — a public corporate dataset — as an example of how desperate the situation is: people are training on a decades-old bankrupt company's emails because there is almost nothing else.

"There's the — seriously, I'm speaking with companies asking what kind of data do you have, what are you training on — you hear finding stuff, for example, the Enron data is out there in public and people are actually using that as real data from a company. It doesn't know how a company works." 00:07:42

Dashboards Will Proliferate, Not Disappear, in the Agentic Era

Conventional wisdom holds that agents will replace dashboards and human-facing analytics. Gonen Stein argues the opposite: as agents multiply and become impossible to audit, dashboards become the only mechanism for humans to maintain any situational awareness at all.

"I actually think we'll see more dashboards because this will be the only way to kind of let us figure out what they have going on in the world. Because coding agents have started to write most of the code that's running in the world... Agents activating other agents would activate other agents and trying to keep track of the non-human identity — it becomes almost an impossible task." 00:18:59

Product-Led Growth Now Works for Enterprise Infrastructure, Reversing a Long-Standing Rule

Gonen Stein explicitly states he was skeptical of PLG for dev tools and infrastructure for years — and says that skepticism is now wrong in the AI era.

"In the past I was arguing that for the majority of things, PLG doesn't work, especially for dev tools, because the world is very fragmented, people don't want to move so fast, so forth. Now it became super hot. Looking at companies like Cognition, for example, which is you know an incredible company that were able to first go through a PLG motion, use that way in Eon, and then through the FD motion turning going to banks." 00:31:44

The Biggest Internal Security Threat Is Not Malicious Actors — It's Well-Intentioned Employees

The framing of AI security risk is typically about external attackers or rogue models. The more novel and underappreciated risk is non-technical employees accidentally connecting company data to external services through consumer AI tools.

"Think of the non-technical people building something with, you know, Wands as a coach or any other software that you have, for themselves — they are putting company data there. They're not even aware for things like security or compliance or who is going to use this data." 00:00:21

3. Companies Identified

Eon

Cloud backup, disaster recovery, and AI data foundation platform. The subject company — aggregates, classifies, and secures enterprise data across hyperscalers, then exposes it for AI/LLM workflows without compromising production systems or compliance.

"We've created a new data foundation that runs in the cloud and we provide multiple capabilities that allow customers to first map and classify their data across their environment, across multiple hyperscalers and identify what they have, where they have it, what's sensitive, not sensitive... and apply their AI models and LLMs on top of that data that's ingested from a variety of sources." 00:01:50

CloudEndure

Cloud migration and disaster recovery company, predecessor to Eon. Acquired by AWS; product became AWS Application Migration Service. Supported mass enterprise migrations across AWS, Azure, and GCP.

"Even before we sold our last company CloudEndure to AWS, we supported similar large-scale enterprise migrations with the other hyperscalers, with Azure and with GCP, where our product was integrated OEM into the console." 00:28:28

Databricks

Data and AI platform. Called out as one of the most important companies adapting to the agentic data wave, continuously reinventing itself to catalog and activate the growing volume of agent-generated data.

"You've seen companies like Databricks, one of the most incredible companies on the planet in my opinion... I have more and more data coming in, I don't necessarily know where it is, I'll help you catalog the data and make use of that... They are reinventing themselves all the time because they understand that more and more data has been generated by agents." 00:25:31

Cognition

AI software engineering company (makers of Devin). Cited as a standout example of successfully combining PLG with forward-deployed engineering to penetrate large enterprises.

"Looking at companies like Cognition, for example, which is you know an incredible company that were able to first go through a PLG motion, use that way in Eon, and then through the FD motion turning going to banks." 00:32:09

Longlake

AI transformation company that acquires enterprises and converts them into AI-native operations rather than waiting for organic adoption. Cited as a pioneer of a radical new AI deployment model.

"There are really great companies, for example Longlake, that say: instead of you adopting AI, I know how to do it more efficiently. If I can buy the company and transport it up into an AI company, we can all win." 00:32:37

Mercor

AI recruiting/labor marketplace. Named as a competing bidder on the Spirit Airlines data set in bankruptcy — notable signal that labor-focused AI companies are aggressively acquiring real-world enterprise behavioral data.

"The rumor, too, is that the other bidder on the data set was Mercor, right, in terms of the bankruptcy bid process. And so it's interesting — you had multiple different companies in the AI world bidding on a bankrupt airline's enterprise data set." 00:04:06

Google

Named for its $10M acquisition of Spirit Airlines' data out of bankruptcy to further train and monetize its travel-related models.

"Google obviously is in the travel space for a while, right? They want this type of data. They're already monetizing it. This allows them to understand, train it, understand that, monetize it even further." 00:09:11

Fivetran

Data pipeline/ETL company. Mentioned as an example of an excellent but increasingly narrow legacy data tool that was not designed for the breadth of the modern agentic data challenge.

"Fivetran was an incredible company, dbt and so on and so forth, but very niche, very specific tools for that purpose." 00:25:31

dbt (data build tool)

Data transformation tool. Co-mentioned with Fivetran as category-defining but purpose-limited infrastructure that predates the agentic data era.

"They were using some great companies, Fivetran, dbt, Monte Carlo, all the data tools that exist, in order to fulfill their tasks." 00:10:17

Monte Carlo

Data observability platform. Named alongside Fivetran and dbt as part of the prior generation of data tooling now being challenged by agent-era requirements.

"They were using some great companies, Fivetran, dbt, Monte Carlo, all the data tools that exist, in order to fulfill their tasks." 00:10:17

Applied Compute

Company providing fine-tuning services for open source models against specific enterprise data sets. Named by Elad Gil as an emerging player enabling domain-specific model optimization.

"There's companies like Applied Compute and others are starting to provide these sorts of services where you can fine tune models or open source models against specific data sets." 00:06:11

Hugging Face

AI model and dataset hub. Cited for releasing a new legal dataset — used as evidence of the scarcity of high-quality real-world data for agent training.

"You see that Hugging Face just released a legal data set a few days ago. But you don't find too many good data sets that doesn't look like real synthetic data that can actually be used to really look like the real world." 00:06:55

4. People Identified

Ophir Ehrlich

Co-founder of Eon and previously co-founder of CloudEndure (acquired by AWS). Deep background in enterprise disaster recovery, cloud migration at hyperscaler scale, and now AI-era data security. Particularly insightful on the security threat evolution from human to non-human actors.

"Up 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, from AI agents that essentially have legitimate access to the environment with legitimate permissions." 00:00:00

Co-founder of Eon and previously co-founder of CloudEndure. Drives the business and go-to-market perspective; highly attuned to how enterprise buying behavior, data economics, and AI adoption patterns are shifting simultaneously. Offers unusually candid market intelligence from conversations with enterprise leaders and AI labs.

"I'm hearing about labs going to Wall Street and trying to buy data from hedge funds and trying to understand how to map and analyze companies. So you see data that was accrued throughout the years by companies, which was usually like tapes, sitting on a shelf collecting dust. And all of a sudden, this becomes very important." 00:04:54

Elad Gil

Host of No Priors; prolific angel investor and entrepreneur. Surfaces the insight about Mercor bidding against Google for Spirit Airlines data — a detail not publicly prominent — and pushes on the applied use case for acquired data sets.

"The rumor, too, is that the other bidder on the data set was Mercor, right, in terms of the bankruptcy bid process. And so it's interesting — you had multiple different companies in the AI world bidding on a bankrupt airline's enterprise data set, which is fascinating." 00:04:06

5. Operating Insights

Assume Breach as the Default Security Posture — For Agents, Not Just Humans

The traditional breach-response mentality, developed for ransomware and external attackers, now applies equally to internal AI agents with legitimate credentials. Companies should operate with the assumption that something is already wrong inside the environment.

"We need to assume breach, whether it's malicious or not, and need to be able to handle it and act accordingly. It's a very weird situation today." 00:17:54

Operationally, this means investing in granular recovery capability — not just prevention — because agents can drop tables or corrupt data in milliseconds before any human notices.

Classify and Map Data Before Unlocking It for AI — The Sequencing Matters

Eon's own lesson from a major AWS customer was that they believed they were protected but were not, because resources were never properly tagged and classified. The same failure mode will destroy AI data projects: building AI pipelines on unclassified data exposes organizations to both security breaches and compliance violations.

"The customer thought that they were protected. They weren't protected because they didn't map and classify and tag their resources properly. So it wasn't protected. And so 60 percent of the environment was exposed by ransomware." 00:15:34

The operating implication: classify first, build pipelines second. Don't let urgency from CEOs or boards compress this sequencing.

Combine PLG with Forward-Deployed Engineering Rather Than Choosing Between Them

Cognition and Eon itself are used as proof points that the winning go-to-market in AI infrastructure is a hybrid: start with PLG to gain bottoms-up adoption and validate the product, then deploy forward-deployed engineers to convert enterprise accounts and compress sales cycles.

"Looking at companies like Cognition, for example, which is you know an incredible company that were able to first go through a PLG motion, use that way in Eon, and then through the FD motion turning going to banks." 00:32:09

6. Overlooked Insights

AI Labs Are Buying Financial Data from Hedge Funds to Model Company Behavior

This is mentioned in a single sentence and passes almost without comment, but it is a significant signal: frontier AI labs are so hungry for high-quality, structured real-world data that they are going directly to Wall Street to acquire proprietary financial intelligence. This has implications for hedge funds as unexpected data monetization businesses, for labs' competitive positioning, and for the kinds of reasoning capabilities being built into next-generation models.

"I'm hearing about labs going to Wall Street and trying to buy data from hedge funds and trying to understand how to map and analyze companies." 00:04:54

This suggests that proprietary financial data — analyst models, deal flow records, portfolio company operational data — may have a new, highly motivated buyer class in the AI labs, at prices that could far exceed what traditional data licensing has commanded.

Endpoint Security Is Being Completely Redefined by Local Agent Execution

Gonen Stein briefly flags that employees are now running AI agents locally on laptops — agents that simultaneously connect to consumer services (WhatsApp), internal corporate networks, and external AI APIs. This collapses the traditional network perimeter in a way that prior endpoint security solutions (even modern EDR/XDR platforms) were not designed to handle, and it is happening right now at scale without IT awareness.

"You see endpoint security, which looked like it solved them so many great companies around it... Now everything that's happening, you see people are running agents today on the laptops and the agents sometimes connected to other networks. And they are connected to think, maybe on OpenAI and connected to your WhatsApp, but also to your internal network and also to other applications." 00:19:50

This is a greenfield security category: endpoint security for heterogeneous, multi-network AI agents running on user devices — distinct from NHI (which focuses on cloud-side agent identities) and from traditional endpoint protection. No company name is attached to solving this, suggesting a product gap.