AnthologyAI wants your data. It offers to pay.

Imagine the small biography a phone could write: a ride home, a pair of shoes, three episodes watched instead of going to bed. Each service holds a fragment. The person who performed all these ordinary acts rarely gets to see the fragments assembled. AnthologyAI’s business begins with that peculiar arrangement. Its consumer app, Caden, asks people to connect accounts and choose what to share. Money enters the conversation. Suddenly, the person supplying the raw material has a seat at the table.

The proposition has two audiences. Consumers get a way to inspect and monetize their digital activity. Businesses get a connected view of behavior that extends beyond their own checkout or subscription records. The same transaction must feel worthwhile to both. A beautifully informed retailer is little use if the people supplying its intelligence decide the bargain is dreadful.

A better bargain needed plumbing

John Roa founded the venture in 2021. His previous company, the digital consultancy ÄKTA, had been sold to Salesforce in 2015. By January 2022, Caden was introducing itself as a personal-data advocacy startup, backed by a $3.4 million pre-seed round. Yahoo co-founder Jerry Yang and Starwood Capital’s Barry Sternlicht were among its backers. The pitch gave consumers ownership, control and a financial interest in data they were already producing.

There was an engineering problem tucked inside that agreeable proposition. A streaming account and a shopping account do not arrive speaking the same language. In May 2022, Caden selected Stardog to organize disconnected datasets into knowledge graphs. HotLink would help retrieve information held by consumer services; Vault would give the person a place to manage it. Before anyone could sell a fuller picture, someone had to connect its pieces.

That work defines the company’s expertise more precisely than the letters AI. Account connectivity, permissions, data relationships and analytics sit in the same chain. If a connection fails, the picture loses a fragment. If a permission is misunderstood, the relationship loses something harder to replace.

The customer brings the ingredients

Caden announced its public mobile-app launch on March 21, 2023, with a 15,000-person waiting list. Its instructions were concrete: establish an encrypted vault, connect services such as Amazon, Uber and Netflix, examine personal insights and earn rewards for eligible participation. The launch materials described points that could be redeemed into a qualifying card or bank account. This was a product you could operate, rather than a philosophical position on the internet.

A September 2023 setup guide added a revealing detail: cash redemption required at least 40,000 points. The number sounds substantial, but without a conversion rate it tells you about the redemption gate, not the size of the reward. Readers considering this model should make the same distinction. A points balance and an income are different things.

Privacy also needs a careful reading. Early launch materials allowed aggregated analytics, advertising identifiers and even identifiable uses with explicit consent. Later AnthologyAI messaging emphasized an anonymous consumer panel. Consent is a set of permissions attached to particular uses. It cannot sensibly be reduced to a reassuring adjective on every product the company has ever described.

The app was also a supply chain

In August 2023, Caden announced a $15 million Series A led by Nava Ventures. Its three announced rounds amounted to $24.4 million. The proceeds were intended for product development, market expansion and hiring. The company also reported that it had distributed hundreds of thousands of dollars to consumers. These are financing and payout milestones; neither establishes profitability or the operating cost of collecting a useful record.

The enterprise suite, CadenOS, gave the consumer app a commercial destination. Its announced uses included aggregated intelligence for hedge funds, retail-media event streams and data for AI models. Caden AI, a separate personal-assistant offering announced that year, aimed to turn a user’s supplied activity into recommendations. One side of the operation offered personal utility. The other sold intelligence to organizations.

By May 2024, the enterprise brand AnthologyAI was being introduced publicly, with Caden retained as the consumer app. Comcast NBCUniversal LIFT Labs lists the company in its portfolio. That is evidence of participation in an enterprise innovation setting, rather than a license to put Comcast on a list of paying customers.

Reading behavior, rather than paragraphs

AnthologyAI’s February 2025 announcement extended its platform to large language models. The company reported more than 45 billion accumulated signals and over 100 million data points collected daily. It also said it served multiple Fortune 100 enterprises. Those are company-reported figures, and a data point is not a person. The scale describes an information stream, not a census.

The proposed applications make the idea easier to grasp: tracking wellness spending, examining local viewing habits or understanding shifts in travel demand. A retail analyst could use connected behavior to investigate where spending is going. A media business could examine audience patterns beyond its own service. These are applications the platform proposes, rather than published demonstrations of forecasting accuracy.

The attraction is a broader field of view. A business can know its own customers intimately and still misunderstand what they do elsewhere. AnthologyAI sits between consumer research, alternative data and enterprise analytics, selling evidence that crosses those boundaries. Its distinction is the combination of consumer recruitment, compensation and connected behavioral records.

Permission is useful. Evidence still needs work.

The approach belongs to a wider market of data-reward apps, consumer panels and transaction intelligence. Digiday reported that even Roa was skeptical of the fashionable name zero-party data. The phrase matters less than the actual exchange. What was shared? For which purpose? With what benefit to the participant?

A practical lesson follows from the design: make participation understandable, provide visible utility and build permissions into the collection process. For a buyer, the next task is methodological. Check whether the panel covers the population you care about, whether account connections remain dependable and whether the records answer your question. Paying volunteers does not automatically remove selection bias. Connected observations can reveal relationships without proving what caused them.

Roa’s own story has continued. In June 2026, Raptive appointed him Chief AI Officer and acquired his separate venture, AlchemyAI. The distinction between those names matters. AnthologyAI’s story remains an instructive experiment in a very ordinary human incentive: people may cooperate more readily when they understand the bargain and receive something for their trouble. The intelligence business starts there.

“I do think that the definition and not the term is righteous.”John Roa / Digiday / 2023