AIMinds #035 | Derek Wang, Co-founder at Taalk
About this episode
Dr. Derek Wang is an accomplished entrepreneur who founded Stratifyd, Inc. in 2014, a global leader in big-data analytics. His journey into entrepreneurship began when he left his tenure track position in Computer Science at UNC Charlotte to pursue this venture. Under his leadership, Stratifyd expanded into a global force, offering AI-driven customer experience analytics to a wide array of Fortune 2000 enterprises. Dr. Wang earned his Ph.D. in Computer Science from UNC Charlotte. Notably, he served as the Associate Director of the Ribarsky Visualization Center, providing essential services to various U.S. Government agencies, including the Department of Homeland Security (DHS) and the Special Operations Command (SOCOM).
Listen to the episode on Spotify, Apple Podcast, Podcast addicts, Castbox. You can also watch this episode on YouTube.
In this episode of the AI Minds podcast, Derek Wang, co-founder of Taalk, delves into his transition from academia to entrepreneurship, the origins and services of Taalk, and the significant role of compliance and ethical AI practices in business communications.
Derek shares his experiences starting Taalk, which utilizes a sophisticated AI system to enhance outbound and inbound customer interactions through improved natural language understanding and data visualization. He discusses the importance of consent and regulatory compliance in ensuring ethical AI deployment, emphasizing the company's commitment to upholding standards that prevent misuse and protect privacy.
Derek also touches on the practical applications of Taalk's technology in various industries, including insurance and financial services, showcasing how AI-driven tools can streamline operations, boost customer engagement, and ensure compliance. The conversation underlines how his entrepreneurial journey and vision for Taalk are shaping new frontiers in AI applications and business communications.
Fun Fact: Derek met his current co-founder, Paul, through a mutual investor in a somewhat humorous situation where the investor was tired of hearing Derek's wife complain about him rearranging furniture at their home.
Show Notes:
00:00 Launched NLU platform improving global company operations.
03:46 Assisted pharma with diabetic drug deployment feedback.
06:47 Developed GPC bot for ethical product marketing.
10:21 AI bots provide factual product information unemotionally.
15:52 Taalk: Self-serve SaaS platform prioritizing market compliance.
17:38 Security system, partnerships ensure protection, zero tolerance.
21:54 Ensuring accurate AI-driven CRM updates with guardrails.
24:41 Autonomy's potential is limited by current unreliability.
26:47 Prefer getting direct pricing info, without sales.
More Quotes from Derek Wang:
Transcript:
Demetrios:
Welcome back to the AI Minds podcast. This is a podcast where we explore the companies of tomorrow being built. AI first. I am your host, Demetrios, and this episode is brought to you by Deepgram, the number one voice API on the Internet today, trusted by the world's top conversational AI leaders, startups, and enterprises like Spotify, Twilio, NASA, and Citibank. In this episode, we're talking with Derek, the co founder of Taalk. How you doing today, Derek?
Derek Wang:
Doing great. Thanks for having me.
Demetrios:
So let's start out with something that is a little bit peculiar. You were a professor and you left the wonderful world of tenure, and you decided that you wanted to become an entrepreneur, not necessarily the path that most would take, but what were you a professor in and why did you leave?
Derek Wang:
So I was teaching machine learning, data visualization at University of North Carolina. And really, it's the entrepreneurial spirit. When you build a good product or you build a different research stuff, you want the world like what deep Graham is doing. You want the world to take advantage of it. You want to apply that to the real economic and businesses. That's why I decided leaving the professorship was one of my, two of my previous CO2 co founders. One was my PhD student, the other was my postdoc student. So three of us just started a journey back then.
Demetrios:
Incredible. And did the. Did the University of North Carolina invest in your startup?
Derek Wang:
No. So that was another story for another episode. It was a lot of back and forth. North Carolina wasn't known for investment or undo investment or whatnot. But we got all our resources from Canada, overseas and all different areas. But we do contribute back to the university and start sitting on their chancellor's boards and really helping future entrepreneurs to get more resource and funding in this neighborhood.
Demetrios:
Yeah. So it's not really, it doesn't have that muscle like Stanford does then, I guess.
Derek Wang:
Oh, no, no. Stanford is a machine. It's so beautifully done. Work out.
Demetrios:
Yeah, exactly. A little more time with you on that board and you'll be able to make it happen. So you went out and you raised some money, you had quite some success. Can you tell us about this first trip around the sun as an entrepreneur? What'd you learn and what happened?
Derek Wang:
It was. It was very fortunate and lucky for us at that time. We started around 2014 2015 focusing on customer experience. Right. So we build natural language understanding machine and platform to help global Fortune 2000 companies looking at the multiple channels of their customer experience feedback and by analyzing the taxable data there. It was really good that we helped company in Finserv pharma and large all around the world to improve their satisfaction, improve their operations in the back office. And I've been running for nine years of the organization and then transition off in 2022, just looking at the world after Covid and say, hey, what else can we do? How can we help organizations?
Demetrios:
Interesting.
Derek Wang:
Mitigate the networking, the engagement factor.
Demetrios:
Okay, so you have deep understanding of machine learning and data visualization. You went out there and you started to tackle the hard problem, especially pre chat GPT of how can we understand what is out there on the Internet about our companies? Or you were helping companies understand what was being said about them.
Derek Wang:
Correct. Right. For example, we helped one of the large pharma companies when they deploy a new diabetic drugs out in Australia. They did a first trial run from the kind of clinical trial people, and we got all the data back and a lot of the fun feedback coming out of that is people don't know how to use it. People don't know how inject where to inject, how to inject all those information. Now you are said, exactly. Beautiful is pregpt, right? We got natural language processing. We extract all those information, visualize that for the decision maker, and they move on really fast.
Derek Wang:
They're one of the largest company on the planet Earth for pharma. They build a website interactive. So for people come in and say, oh, I don't know where, oh, here it is. What's your type? How many days? Blah, blah, blah, blah. You can insert that, hey, here's where you want to inject it. Just based on the feedback we gave to people, we feel like we save people's lives, right. By providing those information, that makes us feel really nice about the work we have done before.
Demetrios:
Yeah, the pharma use cases are high impact use cases. And I imagine the fin serve also. You're dealing with people's lives and people's money. So it's very crucial for us to.
Derek Wang:
Get all the good services over those industries. Yep.
Demetrios:
Incredible. And so then you stepped away from that business and you didn't have enough. You still had the entrepreneurial itch. How'd you meet your now co founder? What's that story about?
Derek Wang:
Yeah, it's really fun. Like, we met through a mutual investor. So our investor in New York, hedge fund manager, was like, hey, Derek, I heard enough. Your wife complaining you're moving all the furnitures in the house like three times a week. So enough. There's enough fly up here. I want to meet your guy. That would money Paul Roberts by themselves.
Derek Wang:
He bring a company to Nasdaq. Right. And then he transitioned out of that. He said, I don't know what you and Paul, me can do, right? But just go me some.
Demetrios:
You're both great founders. You need to hang out for a little bit and sparks will fly. Dating. And Paul had previously ipo'd his last company.
Derek Wang:
Yeah, QBN, it's a programmatic fraud detection company. They do a lot of real time fraud detection for performance industries and et cetera.
Demetrios:
Also in the machine learning space.
Derek Wang:
In the machine learning space, right. So they have sub milliseconds to decide whether this transaction is a fraud or not, which is very much dear to my heart, is we want to solve problem at scale and then really solving more societal impacts. And that's quite honestly led to talks. DNA is you need to be enterprise grade. You need to solve a problem, and then they need to be at scale. That's why we started this together.
Demetrios:
Incredible. So you've been coming at it from this idea of enterprise first, and you then, well, sorry, I interrupted you. Can you finish the story on how you came about creating Taalk and then enterprise?
Derek Wang:
It's really fun like that, right? So from the professor, when GPT came out and when I had more time on my hand, it's like, hey, I need to talk to something. So I build this GPC bot, like a therapist that I always talk to. So when I met Paul, they said, hey, other than the blockchain they were talking about, you can order chinese food using blockchain by now, which we didn't. So all that would just joke around about it, but then say, hey, Paul, pick up your phone, call this number, get a feeling about what it is. He said, dude, you have something, right? His background from performance marketing space, really, to say people don't like other humans to pitch products to them, right? It's very tiresome. We have seen a rock chasing the sun from offshore call centers to neutral call centers to Vegas, and come back out to the offshore. All this start coming to the point of, why can't we have something that people actually like and something that would be consistent and ethical, giving them the product and introduce them the things that may be benefit for them. So that's what we're talking about.
Derek Wang:
And Paul and I are entrepreneurs, were like, we can't just invent a product in vacuum, and we won't. The first thing we went out and said, hey, let's find somebody that may benefit from this. So we got. Went back to our investor, who said, oh, you're in finance, right? You want to reach customer, right? We have this idea. Do you want to use the product? He was like, yeah, I'll give it a take. And then once he saw that, he was at, hey, where can I sign up? Where can I sign up? This service, this is so good. I want to use it to reach, organize people, and that's where important. I say, there's something we can build out.
Derek Wang:
There's something we can pull down together and just make it happen.
Demetrios:
And so the inspiration was really most people, like myself. When I want to go and buy something, I just want to go know everything about it, watch reviews on YouTube, hear different people talking about it in different communities, maybe search a few subreddits, make my opinion, and then go figure it out and buy it. What I don't want to do is I don't want to go onto the webpage and hit that contact sales button and then have to book a meeting. And from that meeting, I get the whole runaround on a sales call where they say, yeah, let's have another sales call to decide what we talked about in this sales call. And it's like, the last thing I need right now is more sales calls on my agenda. Can we just get the information that I'm looking for? I want to know if you do this, this and this. And I've heard some bad things about your performance on this. How have you addressed that?
Derek Wang:
Yep.
Demetrios:
Those kind of questions are what I'm really interested in when I get on a sales call. But of course, the, me being someone who used to do sales, I totally understand the incentives of the salesperson. They want to slow play it. They want to make sure that you're the right fit. They want to get, you really, like, understand how much can they take you for that type of thing. And so it's, it's almost like this painful dance that you have to do. And you said, the hell with all that. Yeah, let's just give people what they want.
Derek Wang:
Exactly. And then to your exact point. Right. So what we also, you from the sales side we're looking at this is, it's the alignment. Salespeople has the alignment. They want to make more KPI. They want to be able to say, hey, I want to extract every single freaking dollar out of you and just make it happen. AI doesn't we build AI to be the one that really to ease your pain of, hey, here's a product.
Derek Wang:
If you want to know the product, you can ask ten. Now we're 40. Whatever you want to do with the bot, it will tell you what the goods and bads factual, right? And then you can make the decision. And then sometimes people will pick up the phone and say, curse the bot out. And bot doesn't care. You can tell them to jump off the east river 50 times, 51 times, one over. Hey, Demetrios, how about this product you want to buy, right? So all those things where we have seen this is a new way of, for us to get information, right? And the last comment on that is to your coin sales. I got one of my friends to say I purposely bought Amazon prime with advertisement.
Derek Wang:
I was like, why did you do this? He was like, because I won't get some information. I want to know what's cool out there. I want to know what people are building product wise. I want to get educated. And that's where Taalk AI would be helpful is like, let's spread the word out, right? Let's let the people know what's going on and be factual about it too.
Demetrios:
Wow. Okay. And so is it primarily outbound services or is it inbound?
Derek Wang:
So it's both, right? So we look at this for different industries and different use cases. So for industry, we're really strong in insurance warranties, financial services, and we're marching to risk and healthcare right now. For industry wise, for use, case wise is more outbound engagement, inbound customer support, cost necessity. We built Taalk. One thing there other than our security compliance first. We don't want this just a Taalk. I don't want it just to be able to, as a bubble head talking. We built a lot of downstream automation.
Derek Wang:
So we actually took the time, built a subsystem, which we call the plugin that can help you to book appointments, write everything to a CRM system, do form filling, updating, polling information, updating, you think about it. So that's why we work with warranty companies where when I call in, hey, I want to update my VIN number for the vehicle instead of talking to a real human. The bot will get it done for you. The bot will confirm that it is updated the CRM and you are done. Think about the power of integration there.
Demetrios:
Wow. Okay. Now let's touch on this idea of compliance first and really thinking about the enterprise because both co founders came from what it seems like backgrounds where you had to have that in mind. I mean, when it comes to fraud detection, you have to be compliance if you're dealing with that. There's so many regulations that are around money and how things work. So it makes sense that it's in your blood. What made you want to do that now? Because as you said, maybe you're going after small to medium sized businesses now, maybe that isn't top of mind for them. Or am I mistaken? Are you only thinking about enterprise use cases? And of course that's going to be top.
Demetrios:
But yeah.
Derek Wang:
So we primarily. So there's two pro, right. The landscape of AI and what people may call ruble calling AI. Dialing consent is a rapid changing landscape. Right. And this landscape, there's two ways we fundamentally believe AI need to be ethical.
Demetrios:
Right.
Derek Wang:
It need to be compliance. But also from a regulatory point of view, ftcs always have different ways that making sure it's a healthy environment for people being contacted, like the TCPA regulation, the DNC regulation and all this. Right? Almost, Dimitri. Almost every business who want to leverage in AI for outbound or inbound support need to be aware of this. But many of them just don't.
Demetrios:
Right.
Derek Wang:
They. They would be like, they either don't know or there's no services. Remind them this is key. That's where we have built in essentially a federal DNC list. We don't allow you to call anybody that's a federal did not call list. Right. So what that stub do, that will actually save you quite a bit of a hassle if you accidentally call somebody on DSC list, because that is findable behavior that is out there. Right.
Derek Wang:
It's very key there. We do work a lot with larger entities they have. So for example, we would dial about 6 million minutes per month and we can go out to 100,000 dials per hour. Yeah. So it's massive there. But just like the programmatic fraud detection my co founders build is, we want to check out every single one of them, right. So that everybody is out of any issues, want to do this. And they're always in the compliance eri.
Demetrios:
Incredible what scale that is. Amazing.
Derek Wang:
And that's powered by you. So that's a shameless plot. Back to divgram. We're not working with you guys. That's awesome.
Demetrios:
Well, now talk me through this a little bit more because I think I understand how you're saying folks are not necessarily going to have compliance top of mind or what they can and can't do top of mind, especially if these regulations are changing. So you're almost taking that upon yourselves to educate your end user. Is there a little bit of danger that they might still do something incorrectly and then blame you because you let them?
Derek Wang:
Very interesting. So Taalk is a platform, right? So we're a full self serve SaaS platform, you can change different things. But for us, we believe as a second time or fourth time as audit entrepreneurs we have the ability to influence the market. We can go to say, hey Demetrios, are you trying to be compliant with this or not? If you come to us and say no, no, no, we're not, then we have the rights to not serve you. It's very important for us to make sure that we can go to sleep tight at the even. I don't want to do anything shady knowing that the whole market as we bring into, that's the fundamental operating mindset we have. But concretely speaking, so where people get into trouble is without the one magic keyword, consent. So anywhere you can do a lot of things with a customer, as long as you have their opt in and consent, we often remind our customer consent is your best friend in outbound or inbound process.
Derek Wang:
And with our system also building is all the checkpoints. Like say I mentioned earlier, we got DNC check out every single call that dials out. And then if people when you dial out to that individual say they're hey, stop calling me, I don't want to be contacted again. Our AI will automatically add them to a DNC list. So this is not going to be offended, right? So that they don't get kissed off. And then we built in, we really took the time, built the diner system and we built in the entire TCPA holding regulation. So Demetrios won't got a phone call at 04:00 a.m. and that's outside the TCPA kind of permitted time zone.
Derek Wang:
We have all those checkpoints built into our system and we are partnering up with other security minded companies with APIs just to make sure we can maximize the ability to protect you. And then at the end of day if you come back and say, you know what, work, I'm going to do something shady. And we got our partner like Twilio help us capture this. We got our self compliance team help us to capture that. If we saw your offender or repeated offender knowing back to the first principle, you're done. We're not, we're not going to serve you. Right? Is that I think piece by piece we can have a very healthy and communication environment that's so key for my kids and probably your kids to grow up into.
Demetrios:
So what are some results that you've been seeing? What are some things that make you really excited about how Taalk has been implemented?
Derek Wang:
Yeah. Nope. It's really opened up the top of the funnel for organization and in a fast amount of time. Right. So typically what we got say in the auto warranty space we can get the people start dialing within 24 hours. They can start looking at this and start reaching out their customers or providing supports within 24 hours. And the result is that we have two KPI's. I will always tell people, number one is our AI conversion rate.
Derek Wang:
We call it Aicr AICR. So the AI conversion rate is when I call an individual gigabit how, what's the percentage? I can convert you to any actions. Right. Talk to a closer or something. The other one is the answer rate. Like it's really how we can help you to reach out. All those people pick up the phone.
Demetrios:
Yeah.
Derek Wang:
So we, we just had a customer who recently started outbound the auto warranty and they came back, we transferred about 200, 300 calls for that customer a day. And they came back say, dude, I need to, I need to hire more people. I just hired more salespeople on the backend because we're start dropping calls that you guys are transferred for us and which to which we have about 27% conversion rate on the AI. Conversion rate for a customer, which is fascinating. Right. Again, I don't want to take all the credit to Taalk as I then the quantity of the leads people opt in and then the rest is Taalk and then your closers.
Demetrios:
Right.
Derek Wang:
So those are pretty exciting to me. It's very exciting to see the real world impact.
Demetrios:
Wow. And how nice me thinking back to my days when I was smiling and dialing in sales and having to listen to so many dial tones and basically 80, 90% of my dials were phone calls that went unanswered.
Derek Wang:
Yes.
Demetrios:
So that is huge. And then I, if I understood correctly, it's the first part of the call will be with AI and it is an AI voice. But then when somebody says the right things, they get transferred to a live closer or salesperson. So that's a human who's ready to chat with you and they know that, hey, this call was happening and here's what they just said. Boom, I'm in. I drop in on the conversation and I say, hey, cool, let's, let's book you for some time.
Derek Wang:
Exactly. Right. And in between this, you have talk AI to whisper to Demetrios as a closer say, hey, I just talked to Derek. He has this needs and go ahead. Right. And boom, connected.
Demetrios:
So you're giving context to there.
Derek Wang:
Absolutely. Right. It's making the sale. That's where the closer raid on their end is skyrocketing as well. It's just no longer just a blind to blind calls is calls with contacts, calls with transfers. And then that also echoed by our ability to appointment making. Right? Hey, you say I really don't want to get this call transferred right now. Can I book appointment? Yeah, we can offer time being real time, converse with you.
Derek Wang:
Hey, you can book something next Wednesday at 09:00 a.m. our sales team is going to call you back. All those things are combined.
Demetrios:
That is so cool. And now I know, working with AI and machine learning, you're talking about some of these features being a little bit agentic. Like you. If someone says, I want to update my VIN number, and then you automatically update that on the CRM, that is incredible. But I also know from playing with agentic tooling, it can be a little unstable sometimes. How do you make sure that what the customer says is actually what gets put into the CRM and almost like put these guardrails on it so that somebody wants to update their VIN number and then they get charged $1,000 because they needed a new Cadillac converter or something.
Derek Wang:
So I think there's multipoints of the check and balances. Number one, the natural defect of a large language model, and the conversation is hallucination. So if you're in a support city scenario, as you talk about this, we actually build our own knowledge base around the large language model, which we tailor to this purpose. And the knowledge base needs be, no, absolutely zero hallucination. That's actually the facts that we can provide to people. Number one, all different layers of optimization has been put in there. And then outside of this is where we wrap around the automation layer. Our automation layer is not derived by any large language model.
Derek Wang:
Is really our ways of saying this need to be 100% accurate. For example, in auto insurance space, one of the key factor is for the new insurance carrier to ask, what's your bodily injury? That could be 10,000, 20,000, you have it right? Different list. And the customer may not know. They may have said, oh, I have 1020. If you just have a normal conversation, they don't know what that is. But because of our plugin system, we can very precisely know what your bonnet injury is and pass that information back to people with 100% confidence of what that is. So it's really combined factor between machine learning, our automation, and then the leverage model that we built all together. It's a full package.
Demetrios:
So I was hoping you were going to say that because it is. You're using AI when you need to use AI, but you're not using AI as much as possible and you're replacing it with heuristics so that it can be more reliable.
Derek Wang:
100%. And the automation need to be purely API to API, right. That's the most robots, and I like the term use for our agenda, right? That's what we call our AI agents. It needs to be performing those actions.
Demetrios:
That autonomy, that's what is the beautiful part. And it is one of those things that has so much interest and there's so much going on around it. But also I think we've been promised the stars and really we're not even at the moon yet. So it's like I a lot of times want to know how people are doing the agendic stuff because I've seen everything from an agent that is able to navigate a web page by the pixels that it has and click on different buttons and that stuff. But that isn't reliable. It can do it, but it doesn't do it ten out of ten times. So when you're in these situations, as you were talking about where you're dealing with enterprises or you're dealing with companies that need this to be reliable, you can't take your chances on that kind of thing.
Derek Wang:
You can do anything funny essentially, right? So those are mission critical ways. And that's where our prior experience was. Enterprise really helped us to think how we can. Well, there's two things, as we always tell people is we cannot ask you to use our CRM system or anything. Right? You are enterprise. We have to bring ourselves to you. And how do I bring ourselves to you? This is what I did back in stratify. We developed 130 different data connectors to all different parts of the organization.
Derek Wang:
We are looking the same way of saying, hey, how can we API into most other systems? Mission critical was we have a strong integration like a Salesforce HubSpot Zoho, where they bidirectional integrate into talk. But then what if you have the warranty updates, which is a third party system, let's API into it. Let's build out a lookup table, right? We got guys in kind of maze service will reach out and say, hey, can you help me do a pricing quote? Great. We can actually have a pricing lookup inside our subsystem to say, oh, defender, you got two kids, two rooms. Here is how much you have to in real time for the prospects.
Demetrios:
I love that. That is so cool to be able to do that. And like I said before, a lot of times I'm not necessarily interested in talking with somebody from sales right now. I just want to get the information, and it feels like you can give me that information. Like a pricing lookup. That's probably the number one thing people want to know. If there's the. Oh, when you go to the pricing page and it says, just talk to sales as the call to action, and it doesn't give you any pricing, usually it's like this sounds expensive.
Derek Wang:
Exactly.
Demetrios:
So if you get that from the call, then it's. It's way better. Man, this has been a blast talking with you, Derek. I'm so happy for all your success, and I look forward to hearing what else you all can create.
Derek Wang:
Sounds great. No. Thanks for being a strong partner, and thanks for having me on the podcast.
