How Agentic AI Closes The Gap Between IRL And E-Commerce Shopping

By Brian Schwarzbach

When you think of personalization, what springs to mind? If you’re a boomer or member of Gen X, maybe it’s Netflix’s recommendation feed. Millennials may think of Spotify’s Daylist. For Gen Z or Gen Alpha, TikTok is likely the answer.

The personalization of our digital world impacts us. It’s a reflection of internal and external identities (MySpace) while revealing what we really want or like (FB/IG ads, TikTok feeds).

While the past decade has seen increasing levels of daily personal digital experiences, today’s advancements in AI have us on the precipice of entering a new era of personalization — far more targeted, tailored and timely.

It’s not personal, it’s business

The e-commerce experience is deeply stagnant and hasn’t meaningfully evolved in 20-plus years: It’s a feed-based scrolling process where you click on an item, look at an image, scroll, read, add to cart, input info, checkout. Consumers need better solutions.

How companies and brands communicate presents another issue. Think about app notifications. Many can relate to getting the same notification at the same time from the same company each day or simply irrelevant pings. This leads to a selective blindness to the notifications (and the company) because they’re not tailored and aren’t meeting you, the consumer, where you are.

The way these feeds or notifications were built has prevented personalization because it’s functionally impossible to build out all the edge cases for an individual. Instead, companies have catered to the largest common audience and default to displaying the most popular products.

Historically, most digital personalization was powered by recommender systems. These are machine-learning algorithms — using collaborative and content-based filtering — to recommend items, goods or content based on things like previous engagement, usage signals and user demographics.

But what if you could remove even more of the friction from the discovery or engagement process? What if the onus of engagement was shifted toward brands and companies to create more of a pull than a push? What if life’s digital interactions were noticeably tailored to consumers?

Wouldn’t that feel almost magical? Joyful? Frictionless?

This type of e-commerce personalization has long been a white whale. Today, thanks to AI, we will soon see massive adoption of these types of tools, so let’s talk about a few interesting use cases in e-commerce and digital consumer engagement.

Get online loser, we’re going shopping

E-commerce has always struggled to replicate the joyful discovery of in-person shopping. What many VCs get wrong about e-commerce is that it’s not always about efficiency. Shopping is also entertainment and the process — the joy of the hunt — is a feature, not a bug.

When you walk into a store and an associate helps you find an item that feels like it was meant for you — that’s an amazing, personal experience that quite simply makes you feel good.

Replicating that online is challenging with feed-based scrolling, relatively creating a lot of friction for the consumer and leading to lost sales. In an age of personalization, lean-back minimal- input-based discovery will become the norm (and much closer to that enjoyable, IRL experience).

For example — how would you typically describe bold heels you need for a wedding? Maybe it’s something along the lines of “spikey statement heels.” But historically, back-end product taxonomies have been rigid and narrowly defined by brand name, item name, SKU and maybe one or two basic attributes, making it difficult to find what you want without the actual product name.

There are several startups such as Daydream or Lily AI that are removing this friction in the discovery process by using AI to create “real-world language” product taxonomies. This enables products to be discovered by more attributes, vibe, style, occasion — words we would use to describe them to our friends.

AI enables these products to be scalably and effortlessly tagged with hundreds or even thousands of relevant attributes — facilitating better discovery. Products like this are much more efficacious and efficient than brand managers or merchandisers manually inputting one to three attributes across hundreds or thousands of SKUs, which is obviously not scalable (so no one does it).

Enter agentic AI — upleveling engagement

Let’s go back to that example of how poorly architected notifications and feeds can negatively impact users. How do we fix this?

At a high level, agentic AI acts independently on self-directed goals with decision-making capabilities that are not just premised on pre-programmed reactions to inputs or stimuli.

This is currently one of the most exciting areas of AI for both founders and VCs because it gets us closer to the optimal relationship between humans and technology: Technology should work for us, we should not work for technology.

What does this look like applied to personalization for consumer engagement? It’s proactive, tailored notifications that make us want to engage and meet individuals where they are. A brilliant example of this type of product is being built by a company called Aampe. Their agentic AI infrastructure delivers continuously personalized and optimized experiences to consumers.

With a tougher regulatory environment around consumer data privacy and third-party cookie usage, startups like this are poised to backfill a massive and fundamental part of consumer engagement and lead us to a new paradigm of tailored, “pull-you-in” experiences.

The exciting part of agentic AI is that it’s a horizontal technology that can be applied across many industries. Agentic AI startups that go deep into a vertical and leverage industry-specific data sets will be positioned to embed into many workflows. The beauty of agentic models is that the more they’re used, the more data is created which can further refine the agents — creating a powerful flywheel.

Parting thoughts

AI-driven personalization will become a foundational technology that will reshape how we interact with digital content on a daily basis. It won’t happen overnight. Regulatory bodies will try to figure out a balancing act for consumers and their privacy over time — just as they did for cookie-based ads and more broadly for AI.

However, similar to the early days of rapidly building strong D2C businesses by running Facebook ads (Bonobos, Casper, Allbirds, Warby Parker) — companies that build and/or adopt AI-driven personalization earlier, will reap asymmetric rewards. As the solutions become more widespread, the pricing-per-action arbitrage or engagement-per-ping effectiveness stabilizes and less customer value can be captured efficiently. In other words, for e-commerce companies hoping to capitalize on this AI trend, time is of the essence to make business more personal.

Brian Schwarzbach is vice president and investor at Cathay Innovation in San Francisco. Previously, he worked on ad monetization and marketplace dynamics at Pinterest, was a fintech investor at LLR Partners, and began his career in investment banking at Houlihan Lokey. Schwarzbach holds a bachelor’s degree in economics from The Wharton School at the University of Pennsylvania.

Illustration: Dom Guzman