Agentic Commerce & CRO: Winning When Your Next Customer Is an AI — Itai Horn (Resident)
Host: Eyal David · Guest: Itai Horn · About the host
In this episode
- A mattress is a rare 'event' purchase with a high AOV and a lot of upfront research — so the whole decision rests on trust (365-day returns), a strong brand (Nectar, DreamCloud) and explaining the offering as clearly as possible.
- Resident measures everything: Revenue, Conversion Rate, Gross Profit and Product Mix (driving the user toward the more expensive product) — and every test that goes live is data-driven, with a clear sense of when to stop a variant to decide faster.
- Building a CRO Agent fed with every technique, internal data and industry research (Baymard) grew velocity from 2–4 to 8 concurrent tests and the success rate from 10–20% to 35%; even those without much traffic can run fewer but more precise tests.
- Agentic Commerce splits into three: Nanagentic (the user as operator — the classic model), Semi-Agentic (the LLM builds a comparison table and the user moves to the site) and Full Agentic (the agent researches and buys on its own) — and full agentic is the goal retailers aspire to.
- Just like the race to be first on Google, now the race is to be discoverable and first inside the LLMs — a new 'LLM Search' expert role is born, transitioning from the SEO world and working with tools that measure your ranking in ChatGPT and Gemini.
- The way into product today is hands-on: try as many AI tools as possible (GPT, Perplexity, V0, Claude, Google AI Studio), build agents, and realize that even the C-level is getting its hands dirty again — the direction is 'builders', not just product/design/engineering.
Listen to this episode
Episode Description
Itai Horn, a Senior Product Manager at Resident — one of the leaders in US e-commerce for mattresses — returns for a second episode to talk about conversion optimization in a world where your next customer might be an agent, not a human. He breaks down Resident's data-driven A/B testing work, reveals proven best practices (Affirm, dual CTA, Good-Better-Best), and explains how they built a CRO Agent that doubled both the number of tests and the success rate. From there he opens up the hot field of Agentic Commerce — the three agentic models, what Stripe sees from the front line, and the new challenge of being 'first' inside the LLMs instead of on Google.
Full transcript with speakers and timestamps
Selected quotes
“If you're simply not there, you'll lose the traffic, you'll lose your positions, you're no longer a leader — you have to be there in agentic commerce.”
“This whole thing helped us grow test velocity from 2 to 4 to 8 tests simultaneously, and the success rate from 10–20% to 35%.”
“You see $100 a month versus $2,000 total — it really shows the user it's not so scary, and it's actually not that expensive.”
Questions & answers
What is Resident and why does it matter in the market?
Resident is a US leader in e-commerce for mattresses under a DTC (Direct to Consumers) model, with the brands Nectar and DreamCloud. It was acquired by Ashley Home for over a billion dollars to add online and digital-marketing power alongside Ashley's strong physical retail. The US mattress industry is estimated at about $17 billion.
Which A/B tests proved themselves in practice?
The Affirm test — a component below 'add to cart' that shows a monthly payment (e.g. ~$100/month) versus a one-time payment (~$2,000), which worked great on pages with a high AOV and raised Conversion, Revenue and Product Mix. And the dual-CTA test on the homepage — two buttons leading the user to both a product page and a bundle page, to serve several types of intent.
What are the three Agentic Commerce models?
Nanagentic (the classic model, 'User as an Operator' — the user arrives from a traffic source, enters the site and completes a purchase); Semi-Agentic (the user asks an LLM like ChatGPT/Gemini/Perplexity/Claude for a recommendation, gets a comparison table and moves to the site to complete it); and Full Agentic (the agent itself researches, chooses and makes the purchase end to end).
What does Stripe identify as the key points in agentic commerce?
Four points mentioned by Stripe's Chief AI: control (the brand's control even inside the LLMs), discoverability (being found and being first in the LLMs, not just Google and Facebook), fraud (labeling good bots vs. bad ones instead of blocking everything), and checkout (an experience that supports completing a purchase even for an agentic user).
What is the Good, Better, Best tactic?
Presenting the same product at three price levels — at Resident they're called Classic, Premium and Lux — so the user sees that the more they spend, the more they get. It supports the Product Mix KPI, and in the LLM era it connects to smarter personalization (e.g. tailoring the package by the customer's location and delivery time).
Related Episodes
No More Moats: How Harmony Reached 50 Customers Without a Single Salesperson | Ran Ribenzaft (Harmony)
01:00:56 · September 24, 2026
Ran Ribenzaft, Co-Founder and CTO of Harmony (and formerly founder and CTO of Epsagon, sold to Cisco for $500M), is an unusual CTO: 90% of his time go...
Tony Arad Felik | Not Just Surviving AI — How to Thrive in It
01:06:04 · July 28, 2026
Tony Arad Felik, a mentor to managers in high-tech and a former R&D Director, on management in an era when roles, hierarchies and expectations are shi...
Aviad Harel | Team8 | How to Build a Thesis About the Future Worth Millions
00:54:32 · June 22, 2026
Aviad Harel, Managing Partner at Team8 and previously a co-founder of Sisense, on how to build startup theses at the earliest stages — before product,...
"Stop Thinking Like an Employee. You're a Startup." | Lior Frenkel (Adaptation)
01:08:36 · May 17, 2026
Lior Frenkel, founder of Adaptation and author of "The Little Book for the New Manager," argues you should look at your career as a startup — with fea...