In the wise words of Spider-Man, with great power comes great responsibility.
Nowhere is this more true, perhaps, than in the tech industry. In just a couple of years, consumers have become “AI native,” said Anda Gansca, CEO and co-founder of customer journey intelligence platform Knotch. This means that consumers have higher standards for the technology they’re using – from B2B software to ecommerce sites – and it’s up to the tech companies to live up to these rising expectations.
Knotch’s most recent solution is called ACE. Announced on Tuesday, ACE is an AI layer that sits on a brand’s existing tech stack to provide a more personalized experience for its customers by prompting natural language conversations and analyzing historical data, as well as brand guidelines and regulations.
One of ACE’s early adopters was Zillow’s B2B arm, whose clients include rental managers ranging from enterprise property managers down to “your aunt who has one or two investment properties that she’s trying to rent out,” said Ben Levine, head of marketing for Zillow Rentals.
It’s still early days (ACE has been in beta for roughly three weeks), but Levine is optimistic about the results so far. Zillow has been able to “get smarter about understanding who our customer is,” he said, by allowing users to provide more granular information about their background and what services they need. The goal is to create a web experience that keeps people on the website longer. The immediate bounce rate (or people clicking off the page after just a few seconds) has been “going up and up and up,” Levine said.
Building blocks
ACE is built on two Google-owned AI models, Vertex and Gemini, and is also supported by Knotch’s intelligence platform, Knotch One.
Knotch One determines the best way to target a user based on data sources like the brand’s CDP, past data Knotch has from working with the brand, the brand’s preferred SEO and GEO partners or Knotch’s own data partner, Conductor.
In addition to all of the preexisting data ACE has on a customer, it also prompts customers with questions (like how many properties they own) when they arrive at a brand’s site. Users can ask additional questions, too, like “what do I do when a renter stops paying rent?” From there, Levine said, the agent can “basically manifest a curated answer” by pulling information from Zillow’s legacy content.
A mix of agentic and generative AI tools ingest the brand’s content, and then break the content down into smaller pieces, or building blocks, like Legos, each tagged semantically, said Gansca.
Then, when a user asks a question, Zillow responds in a way that feels dynamic and personalized, “but really just reassembles atoms of information that have already been approved,” said Gansca.
With ACE, Knotch has created a “closed feedback loop” that is still governed by a brand’s guidelines and compliance standards, but automates the consumer journey so the user “can command the experience,” she said.
It’s “incredibly important” that all of the content ACE is accessing has already been brand-approved, Levine added, since it ensures that the AI won’t “go haywire” and spit out inaccurate information.
Dynamic change
It’s not as though having access to customer data – like their past engagement with Zillow’s website or behavior across the web – is a new concept. What’s new is the ability to act on it, according to Levine.
Zillow’s user experience “wasn’t particularly dynamic” in the past, he said. It used customer data to optimize on a mass scale for property managers, but reaching people on an individual level wasn’t in the cards.
Now, when one of Zillow’s B2B customers opens the website, they’re immediately asked several (optional) questions, ranging from the size of their business to their particular role in property management. Zillow can then serve them follow-up content and products based on their needs (for instance, products that help them bump their listings up in search or syndicate their listings into social).
LLMs have “conditioned us,” as consumers, to expect a dynamic and personalized experience on every site, said Gansca.
In the past, she added, humans had to learn “the language of the internet,” like keywords and topics, to get the information they wanted. But with the advent of LLMs, the conversation has changed. Now, said Gansca, “the internet should come to us.”
