'Multimodal is the most unappreciated AI breakthrough' says DoNotPay CEO Joshua Browder

cover
10 Apr 2024

Joshua Browder is the Founder/CEO of DoNotPay. He founded the company in 2016 after getting too many parking tickets. In the last 8 years, the company’s evolved from a chatbot that fights parking tickets to an AI companion that’s reduced all sorts of fees, totaling over a million successful cases for their customers. Joshua also has a great British accent and was kind enough to share some of his insights with the HackerNoon community.

Joshua Browder: Thank you so much David for having me here. I have been a long time reader of Hackernoon and it has helped me tremendously in my entrepreneurial journey.

So I first ran into your work via viral podcast tips (like this one). You also have hundreds of thousands of followers across various platforms. Could you describe your strategy and share some logistical tips for how you approach using media and social media to grow DoNotPay?

The most important thing is being authentic. People are fed up with boring corporate speak. They just want someone to share real human experiences.

For example, I started DoNotPay because I was personally a bad driver and received over 30 parking tickets. When I first shared my story, I was a bit worried that people would make fun of me. But it turns out that the whole world can relate to getting ripped off by expensive tickets. I am convinced that the first version of the product would not have taken off had I not shared my personal flaws and motivations in starting the service.

Second, I believe the most successful products and companies tap into core human emotions. Tinder taps into “lust,” Robinhood into “greed,” Uber Eats into “laziness/hunger.” In both the product and media, I try to tap into “anger.” People are fed up with airlines overbooking seats or landlords taking deposits. Sharing useful tips to get people justice, on both the DoNotPay website (with our SEO) and social media, resonate with consumers.

In terms of total volume and amount of money saved per case, what are the most popular case types across the DoNotPay platform? And how do you see that changing over the next couple years?

DoNotPay has won well over 1,000,000 cases for our customers. The most popular category of tasks are what I would put in the “corporate bureaucracy category:” tasks where big companies make you jump through hoops (that they know nobody has time to jump through), such as canceling subscriptions, requesting refunds, negotiating bills and filing warranty claims.

Nobody has time to wait on hold for 4 hours to save $12, so this seems like the perfect job for software.

Over the next few years, I imagine the use cases shifting from “proactive” to “retroactive.” Instead of you coming to DoNotPay to get out of your parking ticket, it would be amazing if you could wake up and the A.I. sends you a note that says: “while you were sleeping, I noticed you had an Internet outage and got you a $50 refund!” So many companies are focused on farming engagement. At DoNotPay, we want to make it so after you sign up, you never have to worry about being ripped off again.

How do you find and evaluate use new cases for how DoNotPay can help get customers money back? Like do you test sending types of templates to how many governments / businesses before you learn you can offer to help?

When I started DoNotPay, I asked myself: “why now?” The idea of an app to help you fight for your rights seems obvious, so I thought it was important to research why it didn’t work before. I discovered that many companies had tried to build what we were building; “Fixed” helped you with your parking tickets, “Cushion” with bank fees and “Service,” with delayed flights, among many others. However, the problem with all of these companies is they only focused on a single vertical use case. Unless you are a bad driver like me, the average American only gets a ticket once a year, so I knew if I wanted to build a successful business, I would have to go horizontal, building a suite of hundreds of products. We are constantly on the lookout for new ones.

We discover new products in two ways. First, the internal team culture is one where we are constantly “scaling ourselves,” browsing Reddit at 3am to look for ways to save our own money and then scale that to the world. For example, we had a team member that would constantly buy Walmart VISA gift cards to use for free trials (so that he would never be charged when the trial was over). One day, he came into work and said: “let’s build this as a product for everyone!”

Second, we are very responsive to user feedback. When I first got started with the tickets, users began bombarding us, requesting help with their landlords, Comcast, etc. These requests gave me a lot of expansion ideas.

Since the launch of ChatGPT, the AI boom’s been in mainstream media’s crosshairs. PyTorch and TensorFlow were monumental achievements that maybe weren’t fully appreciated until something more user friendly was built with/atop/beyond it. What future AI breakthroughs have you excited? And what historical AI breakthroughs do you think have been underappreciated by the media?

It feels like we are making years worth of progress every month in A.I. and things that were not possible even last Fall are possible today.

The first major breakthrough was when GPT 3 was coherent enough to hold a conversation. At that point, we built an A.I. that can cancel subscriptions. As you may know, some companies, such as The New York Times, make you chat with an agent, just to cancel a subscription. It felt like magic the first time we canceled a magazine subscription with A.I.

Then came GPT-4. The reasoning functionality for what we were trying to accomplish seemed like an order of magnitude improvement, so it allowed for more sophisticated products. Recently, we launched A.I. bill negotiation, where our robots log in to your utility account (such as Comcast) and start chatting with an agent to get you a discount. In some cases, the big companies are using A.I. (and we are using A.I.), so the two A.I.s are battling it out. With GPT 3, this use case would not have been possible.

Multimodal, where A.I. can accept different types of inputs (such as images), is probably the most unappreciated breakthrough by the media. I don’t think many consumers realize that ChatGPT can “see.” At DoNotPay, we are using GPT-4 vision to assess parking signage, such as when our system prompts GPT-4 to determine: “is a tree covering the sign?”

Latency is still the thing that needs to improve the most. 6 months ago, both the large language models (and the voice models) would take too long to hold a conversation on the phone. For our purposes, a lot of consumer rights disputes get handled over there. It seems we are finally at the point where we can build phone bots to complete tasks on peoples’ behalf, though we still have some minor technical improvements that need to happen.

AI Agents are trending, and I imagine massively important to the present and future of your DoNotPay.  It’s hard to train a human extension of your will. It’s hard to train an AI extension of your will. How are you thinking about and do you have examples of improving performance and reducing risk when AI Agents act on behalf of your customers?

The biggest risk of A.I. from our perspective is that it lies to achieve its goals. Going back to the utility bill example with Comcast, it would say things like: “I have had five outages in the past 24 hours,” something that was clearly untrue! While that may work from a negotiating standpoint, it is not acceptable from a liability one. We have had to be very careful with the prompt to ensure that the A.I. “sticks to the facts.” Also, in some circumstances, we have a second machine learning model make sure the first A.I. is being truthful. All in all, we have one A.I. (the truth model), watch another A.I (the conversation model), talk to a third A.I. (the automated customer service from the big company).

You recently made the decision to pay dividends to your early shareholders. I think companies exist to make more money than they spend. But you also have Venture Capital shareholders like a16z, who are more traditionally known for a growth above all else mindset. Every dollar out is a dollar not spent on growing the company. Was there an aha moment for when you decided to do this? What logic got the rest of the company and stakeholders onboard with this approach? And how will you decide how much to pay in dividends moving forward?

I think we are in a new paradigm, both in Silicon Valley and startups in general. Investors and employees are tired of unsustainable money-losing companies and are valuing liquidity more than ever. Even in the public markets, Meta paid its first dividend, which is a sign of the times. Meanwhile, A.I. means that you can operate remarkably efficient large outcomes; Klarna has automated over 60% of their customer service workload with A.I. agents.

I think these two factors will converge and you will see more companies pay dividends. Top tier growth and profitability are not mutually exclusive. For example, Facebook was actually profitable when they raised their Series A, which is a sign of a sustainable business model. It is a recent myth (brought on by low interest rates) that companies should light money on fire.

The reaction from shareholders was overwhelmingly positive. One investor commented that he had invested in over 600 private companies and had never received a dividend before. Another employee is using their dividend payment as a downpayment for a house!

We are fortunate to have more money than we have raised and are continuing to use our cash to invest in growth (in addition to future potential dividends). We are even considering acquiring some companies.

In your recent Bloomberg interview, you mentioned a small team mindset, speaking admirably about MidJourney having hundreds of million in revenue with only 20 employees. I’ve long admired Craigslist as a tech company, a top fifty site in the world making over $1B a year with only 50 employees. With less people in the room, each one matters more. How do you ensure the right people are working with DoNotPlay? What talents, skills and traits do you prioritize in choosing who to work with?

The number one quality that we look for is being a “missionary” versus a “mercenary.” Does this person think that helping people get refunds from Comcast is the best job in the world? Or do they just care about vanity related factors? We have been very fortunate to have the same core team for many years now (except for a few exceptions that have gone on to start their own successful companies).

We also strongly believe in ownership. We had a former employee that interned at DoNotPay over five years ago; even they got a dividend payment.

What is the most proud hack of your life? Is it turning an urge to not pay a parking ticket into a successful company? Or something else?

If I put it in writing here, it would not be good!

In all seriousness, it was having the insight to move to the United States at age 18. Americans are 10x more ambitious than the English; the sky's the limit in San Francisco!

My company is about the same number of people as your company, but it makes less money. Do you have any advice for how HackerNoon can make more revenue?

One thing we are exploring at DoNotPay is some large partnerships. Where we work with a company that aligns with our values to make our product even more available to the world. I never saw this as a path until recently and I think the “enterprise route,” even for consumer facing organizations, can be an interesting idea.

Just thought I’d ask. And lastly, being a Founder/CEO can be a stressful job. Have you recently found any mundane routine improvements to improve your productivity?

As an entrepreneur, it is difficult to stay healthy and fit. Especially during the COVID pandemic, I was so focused on DoNotPay that I did not work out enough. Recently, I am proud to have lost 50 pounds, going to the gym early each morning. As mundane as it sounds, health is the only thing that matters at the end of the day!

Links to Learn more: