Trellis builds industry-leading solutions that make insurance easy for everyone, from consumers to partners to insurers.
In This Video
Daniel Demetri, CEO & Founder, shares the details on Trellis and what it’s like to work there.
We discuss:
- Details about Trellis & how the service works
- Team details & growth plans
- What motivates the team
- What’s next for Trellis
Video Transcript:
We call about 100,000 people a day. Maybe 10% of them pick up, but if they do, they’re talking to a virtual agent initially. That virtual agent is qualifying their interest in talking to a human agent. Our agents are equipped to sell auto, home, motorcycle, whatever the person needs, and they deliver best-in-class service.
Technology empowers the agents through what I call the “local agent at scale” model. A local agent is going to know that in this part of North Carolina, National General is a better fit than Progressive for a certain profile, or that Chubb won’t look at them. That’s the local agent knowledge. However, we want the at-scale benefits of being open seven days a week. If you call a local agent, they might be on vacation for a week or unresponsive because they’re busy that day. So there are benefits to being at scale. We want the best of both, and technology has enabled our national contact center of agents to operate with that local wisdom and expertise.
Machine learning is saying, “Jack is coming in with this zip code and these types of vehicles.” It gets quotes from insurers and ranks those quotes for the agent. The agent can then say, “Given your profile, this is the best fit for you. Let me ask you a few questions.” If you don’t have the patience for those few questions, they can offer, “We’ll go to this other option. It’s going to be faster but maybe a less good price.” That’s how the agents are navigating the system, delivering local expertise with the benefits of being at scale.
Within technical functions, we’ve got a few essential teams. Our Conversation Engineering team is the team that’s actually talking to customers over text and voice and interoperating with everything else we do.
Another team is the Real-Time Bidding team, which is in charge of the machine learning that powers and allocates our marketing spend. We spend about $100 million a year on marketing right now. One team of four engineers is in charge of all of that spend. Per engineer, they’re spending $25 million—that’s a high-impact job. We’re hoping to spend $200 million in the next 12 to 18 months, meaning the impact of even 1% improvements to the algorithms has massive returns that pay for the salaries. We’re really excited about growing that team in particular.
Then we have the Core Engineering team, which is in charge of the intelligence around the orchestration and the ranking of someone’s actual insurance needs. They answer questions like: What is the best policy for someone? What’s the right insurance company for them? Which agent should they speak to? Is now a good time to reach this person, or should we try again after work?
Finally, we have our Data team. We’re dealing with a huge volume of data about each shopper. Every interaction is getting logged, and we’re trying to make the best use of every millisecond of the customer’s time. For every text message they get, every text message they send us, every click on our website, we’re thinking, “How can we paint the pixels to be slightly better? How can we use language that’s slightly better? How can we pick up the phone with slightly better language, intonation, or voice quality?” Our data team has been responsible for ensuring we have ultra-high-quality data and then using that data to drive impact.
We’re solving hundreds of thousands of people’s financial problems every day. It’s motivating to see the five-star reviews on Trustpilot and hear when customers are crying on the phone because they’re so thrilled and blessed by what we did for them—the impact on their lives is meaningful. That kind of individual impact really motivates my team and me.
We’re investing in the conversations and the insurance intelligence so we can have personalized, conversational, and “agentic” recommendations that are limited to that channel and moment in time. If someone says, “I need full coverage,” the system needs to be able to limit the search to full coverage. We’re building that conversational, “phone-you-just-talk-to” experience, but just for insurance.
Our customers are saving $800 a year. A lot of our customers are making about median national income, maybe even less, so $800 a year is a lot. We’ve had people say, “This is the difference between being able to put Christmas presents under the tree for her grandkids and not being able to.” Our average customer is about 50 years old, so we have a lot of senior citizens on fixed incomes, and this is a particularly challenging time for them. Our solutions are really amazing for those individuals.
Insurance companies have a million questions they need to ask you, and it’s faster and easier to get all those questions done on the phone. The phone is a better channel because you can hear intonation, and someone can ask for clarification. Some of these questions are super confusing. It allows for a dynamic workflow: a customer can ask for clarification, an agent can provide examples, and we can make inferences from information you’ve already provided to skip questions that don’t matter (e.g., if you’re uninsured, the insurance companies that ask about that don’t care about your education level, so we can skip that question).
This dynamic workflow is high-bandwidth, exchanging information about how impatient a person is getting. Should the agent go for the faster, slightly more expensive option, or is this person willing to hang out and answer more questions to get a better price? Those things can be accomplished on the phone. We deliver people a better return on their time on the phone than we can on the web.
