He Left Google to Flip Who Bids for Whom in Online Shopping Raphael Cohen Founder & CEO of haggl.ai

What we talked about.

In this episode of the Sunny Ray Show, host Sunny Ray talks with Raphael Cohen, founder and CEO of Haggle.ai, calling in from Tel Aviv, Israel. Cohen explains why advertising has always been a workaround for the fact that a store cannot talk with every customer one on one, and what happens to that workaround when AI shopping assistants replace human browsers. Drawing on his years as a product director working on Waze and Google ads, he describes how platform incentives pull drivers, advertisers, and shoppers in different directions. He outlines how Haggle lets merchants set a goal and budget while shopper agents prove they are a good fit without revealing personal data, using cryptography such as zero knowledge proofs. Cohen also covers how a neutral venue keeps agents honest, how personalized offers differ from manipulation, how merchants discover the product through agencies and a Shopify app, and why timing is the hardest part of building for a market that barely exists yet. He closes with a view of which shopping categories will move to agents first.

Raphael Cohen says AI shopping agents will make ads obsolete, replacing bought attention with direct, verifiable negotiation between merchants and shoppers.

The questions, and the answers.

What did your years in product at Google teach you about the incentives between merchants and shoppers?

At Waze I saw the driver wants one thing, the advertiser wants another, and the platform gets paid to bend the first toward the second. In e-commerce, shoppers pay for the ads that convince them, and merchants overpay for customers. Nobody is evil, but incentives point three ways. Remove the middle platform and they align.

Why did consumer data need to move onchain before this could work?

I'd push back a little. It's not really about the blockchain itself. What mattered was decentralization, with no platform deciding who sees whom, and proof of value. My agent can prove it's a valuable customer with minimal or no data leakage. We use cryptography like zero knowledge proofs, which can be onchain but doesn't have to be.

How do you keep a shopper's AI agent from gaming the negotiation?

Honesty comes from the venue, not from the agents. My background is algorithmic trading, where markets work because the exchange is neutral, not because algorithms are nice. Our matching runs inside verified code anyone can check. Neither Haggle nor the merchant can peek at orders or tilt results, so nobody has to trust anybody.

What's the line between a personalized offer and a manipulative one?

Manipulation is when the seller knows things about you that you don't know they know, like fake countdown timers or prices rising because you visited three times. Haggle is the opposite. Your agent decides what to prove, the offer is explicit and in writing, and your agent can compare every store and walk away at zero cost.

Can you describe Haggle in one minute?

Haggle is performance marketing built for agents instead of humans. You tell it in plain words, for example give me 20 new customers this month with this budget. The budget goes to improving the offer for the shopper, not to a platform. The budget is a ceiling, and you only pay when it works.

How do merchants usually find Haggle?

Increasingly the agent traffic finds them first. E-commerce agencies are the best channel right now, since one agency runs dozens of brands and notices visits that don't behave like humans. There's also a Shopify app, and bigger partners in payments, energy, cellular, and insurance who want it for their whole customer base.

What is the hardest thing about building infrastructure for shopping that barely exists yet?

You have to guess future pain points instead of listening to current customers, and you need luck in timing, because being right too early is the same as being wrong. It's also a chicken and egg problem between merchants and agents, and the ground shifts daily with new launches and protocols.

In five years, will shoppers still visit online stores?

Some will, because many people enjoy shopping and won't want agents to take that fun away. But for utilities like internet, cellular, electricity, and insurance, people just want the best price handled for them. Those verticals go first. I don't see humans still using those markets directly in two years.