Aaron Lee, Co-Founder & CEO at Smith.ai – Interview Series

Aaron Lee, is the Co-Founder and CEO at Smith.ai, a platform that combines AI and human intelligence to offer 24/7 customer engagement with live North America-based agents to capture and convert more leads.

Aaron is also the Ex-CTO of The Home Depot and co-founder of Redbeacon, which was acquired by Home Depot in 2012. He is a founding engineer on Google Video & led YouTube monetization.

What initially attracted you to computer science?

I was a huge fan of video games and computer graphics when I was younger, and that fascination pushed me to study computer science. My first computer was the Apple IIe – I think you can still buy old models off of eBay, but it’s nowhere near the capabilities of today’s computers. It had a 6502 microprocessor and 64KB RAM (for context, my current laptop has 64GB of RAM, 1024 times more). But I loved that computer because I was coding in the assembly language, and I was able to program games on that computer.

That passion from a very young age prompted me to study Computer Graphics for my Ph.D. In the end, it served me well. A fun fact, after I graduated from Princeton, I almost took jobs with Nvidia and Activision, because I was excited about their upcoming game World of Warcraft. But on the insistence of a friend, I joined Google instead, where I had the really valuable opportunity to be the founding engineer for Google Video, and later worked on YouTube’s monetization features.

Your first startup RedBeacon was an online home services platform that connected consumers with contractors for their home maintenance, repair, and remodeling needs, and was later acquired by Home Depot. What did you learn from this experience?

That experience taught me how to build and scale a company, the myriad of challenges SMBs are facing, and ways to cultivate a stellar company culture.

Up until RedBeacon, I had only worked for massive companies – throughout my time at Google, the company grew from 2,000 to 20,000 employees. With RedBeacon, I was building and growing the company myself and we were a small team. I had to learn about every single component of building a business, like creating and marketing a product, finding customers, and hiring. I think because I had that experience and learned so much about what it requires, I was more successful with Smith.ai.

I initially started RedBeacon because I saw a gap in the market. I was personally frustrated by how hard it was to find a contractor, and I realized that contractors felt the same way. They wanted an easier way to find customers. Later on, when I was the CTO of Home Depot, it became my job to speak with these home services pros to understand the challenges they were facing, and how technology within the Home Depot umbrella could help. Those personal conversations helped me understand that much in the same way growing and scaling a startup is tough, being an SMB is equally challenging – there’s a lot you have to navigate, and rarely are people given explicit guidance on what it takes to run a business.

Another area where Home Depot excels is providing lots of training and promotion from within. During my time there, I worked closely with the former CEO, and my mentor, Frank Blake, and he instilled this desire in me to think about how we support and encourage our team. When I later founded Smith.ai, I tried to build that culture from the ground up. We offer lots of training to our 600+ employees and try to promote from within wherever we can. Something I am very proud of is the fact that nearly every specialized team and management role is filled by employees who moved on from frontline roles to customer success, recruiting, training, product, and sales. Promoting people who know our business and do good work for us has helped us elevate the company and be more successful.

You’re an angel investor in multiple start-ups including NerdWallet, what differentiates successful founders from average founders?

I see a few common characteristics among successful founders:

First, they have to be great salespeople who can sell a mission or vision even when they have nothing. You have to be able to convince smart people to follow and join you. Most of the “overnight successes” take years to make and you have to be passionate about the underlying mission.

Second, great founders are always resilient. In startups, you encounter more downs than ups, and you have to be persistent and adaptable to changes. You also have to be willing to keep grinding and figure out a way to accomplish a lot with very few resources.

The last trait may seem somewhat contradictory and counterintuitive, but you need to be stubborn and yet willing to listen and course-correct. I think oftentimes founders get misunderstood because they think about the present and future at the same time. But being able to find that balance and knowing which instinct to follow in a given situation is key.

Can you share the genesis story behind Smith.ai?

As I mentioned, my experience with RedBeacon and Home Depot exposed me to the challenges that SMBs face. One problem I encountered constantly from thousands of pros was that it’s hard to be good at your job if you don’t have time to do it. There are only 24 hours in the day, and if you focus on your specialty, it can be hard to make time for the other components of running a business, like responding to customers as quickly as they’d like.

In looking at the existing solutions to this problem, there wasn’t anyone that felt like it addressed this challenge. Most virtual receptionists, answering services, and traditional call centers relied on scripts. We’ve all experienced how they feel forced, distant, and inflexible. However, SMBs need a cost-effective way to manage this problem because they lose out on valuable revenue if they ignore it. I realized that if we empowered agents with AI tools, we could offer a better service that didn’t feel like the typical outsourced agents. So in 2015, I left Home Depot to tackle this issue and build this technology with my co-founder, Justin Maxwell.

Smith.ai offers more than webchat, what are some of the different features and use cases that are offered?

We offer multiple services to ensure that businesses can engage with their customers on different mediums. The first product we brought to market was virtual receptionists. When a customer calls one of our clients, that call gets routed to a human Smith.ai agent who handles the customer interaction. Our clients can personalize their greeting and tone, so it feels like an extension of the business. We later built on this offering with Live Web Chat, which is powered by AI but also fully staffed by humans to make sure that no message goes unanswered. Our clients integrate it into their websites, and it allows website visitors and potential customers to directly message a business 24/7.

We saw significant traction with these tools and found that more of our customers were looking for support with outbound sales. Fast follow-up is the key to securing new business, and data shows that to convert 90% of leads, sales teams or outsourced providers need to make five call attempts! So in 2021, we launched another tentpole product, Outreach Campaigns. Our agents have the same support as virtual receptionists but are trained as sales development representatives. With these solutions, we’re supporting businesses with every type of customer engagement: inbound, outbound, calls, and messages.

What are some of the machine learning algorithms that are used?

We have built proprietary, in-house AI capabilities so that we can offer the best possible service to our clients. One of the key ways we use AI tools is to support our virtual receptionists and outreach campaign agents. For each call, we provide them with prompts and relevant information that is tailored to the specific business and conversation. Our AI is also trained on over eight years of proprietary customer engagement calls, so our models have become very sophisticated at navigating all of the different nuances of a business. One personal example I love is that our model can understand the different meanings of the word “gross” – it has very different implications in a legal context than in a home services context! Before, we used to present all of the relevant information to our agents at once, but we found that people were overwhelmed by that, so over time we adapted our AI tools so it would give agents select tailored prompts based on the flow of the conversation. That way, agents could provide callers with the right details, while also collecting the relevant information, and knowing exactly what to say at the right time.

We also transcribe and record calls using Assembly AI’s technology, so if a client opts in, they can have a searchable record of all caller engagements. As part of these transcriptions, we’ve incorporated technology that automatically identifies and blocks out personally identifiable information (PII) like Social Security numbers and credit card numbers. One other very popular feature with our agents is that we use algorithms to block over 20 million spam calls – a service we offer free to our customers.

How does Smith.ai integrate with existing platforms such as Salesforce or Slack?

Part of our operating philosophy is that no tech exists in a vacuum. To provide value, you have to integrate with the other tools and platforms that people use. Ultimately, what we’re offering is ease and expediency. With integrations, when one of our agents talks with a customer, we can update all of the relevant platforms with key information – we can put call transcripts in a CRM, book a meeting through Calendly, or set up other automations with tools like Make or Zapier. It’s making the process as smooth as possible to work with us. That’s why we have a team of solution architects who make sure that our services can be integrated into other tools, and help our customers with the initial setup.

Can you share your views on why agents working in unison with AI offer a better experience than agents working alone?

This is a topic I am very passionate about, so I’m glad you asked! AI is the secret sauce that allows our agents to feel like a natural extension of a business, rather than an outsourced operator. AI is a powerful tool – it allows us to give very tailored and personalized responses to customers. Often, our clients have nuanced customer interactions: they are dealing with a high-value situation, or something very emotional, like a flooded house or divorce. Human agents, guided by AI, best navigate these interactions because compassion and emotional intelligence are innate and ever-present. AI helps agents feel supported – they know that they will have guidance on what to say or do next in a conversation.

What is your vision for the future of AI for customer engagement?

I think we are still in the early days of this technology, and businesses will need to figure out how to maintain the relationship between humans and AI to optimize efficiency without sacrificing personalization. Technology is indisputably a valuable tool for customer engagement, but I want to caution against fully automating customer service with AI. Many interactions are very straightforward, but there’s an equal number that requires a higher level of emotional intelligence, and AI isn’t equipped to traverse these on its own.

There are ways that AI will undeniably be helpful. For example, AI answers the phone and directs callers to the appropriate extension, or one AI copilot converses with another AI copilot to schedule a follow-up meeting. We are still very much in the early stages of technical advances for customer service, but I’m excited to see how AI can support both efficiency and personalization. However, I still firmly believe that to reach the full potential of customer engagement, we still need to foster human-AI collaboration.

Thank you for the great interview, readers who wish to learn more should visit Smith.ai.