Transcript – Veritone’s Analyst Update and Tech Demo Day

EPISODE 12

“BH: Today I’m going to be walking you through Veritone Tracker. This product was built for the person of interest tracking use case, it leverages technology that looks at the entire person as an object and allows you to then track and follow that person’s whereabouts across different camera angles and media sources. Let’s get right into it. In this example here, we see a person of interest spotted in the subway station. All we need is a single appearance of that person across any footage and we can use that to actually run a search, based off of their appearance. So it’s not looking at facial recognition or any biometric features but it’s actually doing this as looking at the visual appearance of that person.”

[INTRODUCTION]

[0:00:39.4] RS: Welcome to Veritone’s Adventures in AI, a worldwide podcast that dives into the many ways technology and artificial intelligence is shaping our future for the better. I am your host, Megan Mintchev and today, we have a little bit different episode for you. We are covering our analyst update and tech demo day where you are going to hear from Ryan Steelberg, co-founder and president of Veritone, talking about a brief overview of the business as well as some of the new initiatives that we have underway, followed by three of our business unit leads, each discussing new products and initiatives and showing you some of the demos on the products in each of their respective divisions. It is my pleasure to hand things over to Ryan Steelberg.

[INTERVIEW]

[0:01:25.2] RS: Good morning, good afternoon, and good evening. Welcome to Veritone’s 2022 Analyst and Tech Demo Day. We’re thrilled that everybody could join us today. We’re really excited to talk a little bit about some highlights on our numbers from our recent quarterly earnings in Q3 as well as we get to jump in and walk, through and share with you some fascinating and amazing new product innovations that we’ve rolled out here over the past few weeks and a few months. 

First, I want to acknowledge our talented team and staffers at Veritone, who helped put together another fantastic quarter in Q3, with total revenue growth over 64% year over year. Customer growth over 43%, to now, over 618 SaaS and software-related customers, while maintaining over a 90% retention rate of customers. It’s a fantastic job and everybody should be very proud of that.

Over the past several quarters, we have been making intentional disciplined investments in both the AI work platform and associated market-specific applications and services that leverage it. I’m proud to state that while maintaining and extending our historical applications and services that continue to enjoy a very high customer retention rate, we have also continued to innovate.

[0:02:37.9] Today, I’m excited to introduce you to several of these AI driven innovations and applications, all of which have recently been moved into general availability for our customers. For a while now, we have argued that artificial intelligence in good times and bad will continue to thrive. 

Whether it is our advertising services, our HR solutions, or our SaaS offerings for meeting entertainment, our customers are realizing the improved ROI inefficiencies that is enabled by Veritone’s AI Ware Solutions.

We have an exciting and agenda for everybody today, starting off with the deep dive in our voice and voice technologies and then transition over to our exciting new HR technology, specifically pandoSELECT application. And then finally, we’re going to bring it home with the introduction of our exciting new, people tracking or person of interest tracking solution called Tracker and now, I would like to turn over to Sean King, our senior vice president and general manager of commercial enterprise. Sean?

[0:03:34.5] SK: Thank you Ryan. I’m pleased to be here to share some of our ongoing advancements with Veritone Voice. We continue to see broad adoption of our voice product across media, entertainment and advertising groups with specific emphasis on audio groups both broadcasting and digitally consumed audio like podcast and streaming as well as sports.

As we look ahead, we continue to see great opportunities to grow existing use cases, as well as expanding to new ones. Veritone Voice has a unique ability to help content creators across many different industries, localized and personalized their continent at scale.

Veritone Voice is uniquely differentiated from our point solution competitors because our solution is built on aiWARE. Not only can Veritone Voice create best-in-class synthetic voice models in two modalities, text-to-speech and speech-to-speech, we have the ability to create end-to-end enterprise workflows and audio automation that allow an organization of any size or even an individual to rapidly scale their content. 

A recent example of this is our announced partnership with Stats Perform. We’re very excited to showcase our capabilities here shortly. In this partnership, we’re ingesting data feeds, directly from Stats Perform and using our natural language generation or NLG models to convert this data in near real-time to speech to provide play-by-play for sports.

This is a game changer for the sports industry, as it allows the ability to provide full accessibility to thousands of global sporting events and can be made available in a variety of different languages and now, I’m going to turn it over to my colleague, Corey Hill, who will take you through a demonstration. 

[0:05:13.1] CH: Hi everyone, this is Corey Hill, bringing you another sneak peak of our live sports offering through our recent partnership with Stats Perform. In this partnership, we will be ingesting match statistics and meta data from Stats Perform, in order to generate engaging and customizable digital experiences for sports fans around the world.

With the help of a little NLG and digital sound effects processing, we’re able to not only generate the appropriate commentary but also apply timely sound effects in the background as well as targeted content such as historical facts and advertising for a truly personalized, enhanced experience at the speed of AI, with Multiple personalities and full match coverage:

“Man City is running a four-three-three formation with Edison as goalie and Jay Stone, Joao Cancelo and Mac MG. And Nathan Ake on defense, Roderic.”

“Well, had that been a successful attack and that play would be the youngest competitor to score a world cup final goal since Pele.”

“Rodri weaves right past defense.”

[0:06:09.0] CH: Across various languages.

[0:06:11.0] “[Language]”

[0:06:18.5] CH: And with the opportunity to combine with engaging visual content:

“[inaudible 0:06:22.4] he sends a pass to the center in the attacking AT. Batubinsika clears the ball, goal, it’s a goal, ESG closes the gap within Haifa over a low left shot by Messi just barely finds the net. 36 minutes in and the score is one to one. Lionel Messi has scored his first goal in the UEFA champions league this season. PSG have scored in their last 10 UEFA champion’s league games.”

[0:06:52.2] CH: This puts Veritone at the center of some compelling opportunities in the digital sports space. Thank you very much.

[0:06:58.1] SX: Hello, my name is Steve Xeller. I’m the chief revenue officer here at Stats Perform. Stats Perform is a global leader in sports tech, provided the most trusted sports data in latest innovation AI to lead next generation solutions for teams, sports books, broadcasters and media.

Our clients are companies like Google, Apple, Amazon, ESPN, NBC, BBC Sport, CANAL+, Bet365, Professional Teams and many more. We’re excited to recently announce our global partnership with Veritone to give sports data a voice. Opta Voice powered by Veritone combines Stats Perform industry-leading data with Veritone’s award-winning, synthetic voice AI technology.

We will create new ways to engage sports fans with life-like AI content at scale and in multiple languages. Together, we will enable voice-enhanced commentary for automated match previous, player recaps, and play-by-play information.

We’ll empower fans to choose audio content that fits their preferences, it will reach a global audience in their native language, providing access to underserved markets and sports reporting to the visually impaired. Veritone and Stats Perform share a passion for innovation, working with them as a partner has been a seamless and rapid evolution of bringing this type of solution to market. 

We’re engaging our respective client bases and prospects and are seeing huge demand for AI voice-powered sports data. We look forward to our continued partnership and commercial success. Thank you.

[0:08:30.3] RS: Thank you Steve for being here and being able to share and now, I’d like to take a moment to introduce you to our head of PandoLogic, Terry Baker.

[0:08:38.3] TB: Hello everyone, PandoLogic’s operational theme this year has been all systems go. For us, that means the development of multiple business units to penetrate the varied and diverse talent acquisition market with multiple products designed for specific market fit across what is a 13-billion-dollar in opportunity just for job advertising.

I’m happy to report that in Q3, we have ample examples and great results on our all-systems-go approach, so let’s get started. Let’s start with our growth. In the HVH business unit, our client expansion efforts have resulted in revenue growth up 374% in Q3 compared to the prior quarter. 

One of the biggest objectives this year for high-volume hiring has been to expand our client base within the supply chain and logistics market including trucking and delivery services. We’re already working with providers such as FedEx and JB Hunt and one of the largest expansions in Q3 has been with Amazon DSP or Delivery Service Providers. 

Which is a program created by Amazon corporate to help the entrepreneurs launch and operate their own package delivery business, which we moved into our new franchising business opportunity. Where the enterprise business unit, we saw an expansion of 25% quarter-to-quarter growth and 118% growth year to date versus our prior year.

[0:10:07.0] In the last quarter, we have seen some big client names being added to our expanding portfolio. Major new clients like Regis Court, Lincoln Tech, Behavior Frontiers, and even a roofing company, Bartlett Roofing.

One of the strategic things we’ve done is differentiated our vertical market and industry penetration. There is ample opportunity for continued growth across new vertical markets, which for us, becomes really strategic as we see changes happening in the labor market. 

Another area of massive growth has been our partner relations effort as we launch 10 new partnerships in Q3. These partnerships represent not only revenue opportunities for cross-selling to partner clients but also faster implementation times for new PandoLogic clients as well as a seamless workflow for recruiters that use multiple talent acquisition platforms.

And finally, growth with our new franchise business. We’ve seen steady growth with our intent to become the leading top of funnel recruitment marketing and talent engagement toolset for franchise operators by the year 2025. We’re investing in new product offerings to help us get there. 

Franchises currently have a total of over 8.5 million open jobs, that’s a lot of opportunity and we’re providing complete talent management toolsets for multi-unit franchise operators and we’re investing across the board to meet this demand head on.

[0:11:34.3] Part of that investment is to create a product offering that specifically caters to the unique needs of SMBs and franchise business owners. As a result of our acquisition of Wade & Wendy, we’ve launched pandoSELECT and you're going to see a demo and a client testimonial of pandoSELECT in just a moment.

And as you could see on the right side of the slide, the market has taken notice of the work we’ve done. There’ a list of awards and recognition by some of the industry’s most prestigious organizations that PandoLogic has received since our last invest to date and now we’d like to show you our newest product, pandoSELECT in action. 

Let me introduce you to Whit Walker, our VP of product. Take it away, Whit.

[0:12:18.2] WW: Great, Terry. Thanks so much for the introduction. My name is Whit Walker and I’m going to walk you through PandoSELECT. This is the PandoSELECT dashboard. It’s a dashboard that houses all the jobs and applicant information that we acquire for our clients using pandoIQ.

So all the information that you see on this dashboard here relates to the client jobs and applicants that are actually being distributed by pandoIQ. On the left side of the screen, you can see we have the various jobs here and as I click around, you’ll see that the right side of the screen updates. 

That’s because the right side of the screen houses the different applicants that we have received via pandoIQ algorithmic AI power job distribution and each of the applicants on the right side of the screen has various information, such as when they applied, their phone number, email address, resume and in certain cases, they have answers to questions that we can get via our chat product. 

The information here is all organized to make it super easy for the recruiters to do their job of evaluating talent. So they can look at the resume, they can understand what sort of credentials and qualifications somebody has and they can either advance them or they can reject them for a selected reason. I’ll quickly now show you what the chat looks like, so you can get an understanding of the information and how it’s collected that powers this. 

So to demonstrate the chat, I’ll take this applicant here, Nida, and this is what a chat would look like for Nida. This chat would be branded for the company’s, you know, have the company’s logo, appropriate avatar, we can make sure that it is in line with corporate brand standards. It is customized to Nida as well and it follows a very familiar dynamic sort of interaction that we all know because we’ve all chatted before and done SMS and messaging and what we do is the chat basically chops up a lot of the relevant information to sell the company and also get information back from the candidates so that we can qualify them for the job. And we’ve present things such as videos, we saw it just earlier. We can have lots more in a corporate information you know, about perks, benefits, FAQs to really help the applicant understand what it’s like to work at the company in question and within the experience, we have a couple of different types of questions. We have directed questions such as this one, “Are you authorized to work in the United States.”

We also have open-ended questions like we’ll see here in just a minute, such as, “What are your pay expectations?” Put in the information, we’re going to see it populate in the dashboard in just a moment and throughout the whole interaction, you know, candidates really love this because it allows them to answer it at their own pace. They can do it at their own convenience and they are not pressured to like you might be in an interview. 

We also have directed types of questions where we’re going to prequalify somebody based upon experience with different types of tools and you know, kind of knock-out type of answers there. So now that we’re done with the chat, let’s split back over and we’ll see what this looks like in the dashboard. 

Previously, Nida didn’t have any of the questions answered but upon a quick refresh, we see that we can now have an evaluation of all of Nida’s qualifications for this job and we can take action on Nida. All right, thank you. 

[0:15:34.6] TB: One of the pandoSELECT clients is Blake Quinlan, franchisee for expressive point professionals. Now let’s hear about his pandoSELECT experience. 

[0:15:44.1] BQ: Yeah, before we got connected with the PandoLogic team, we were basically using ZipRecruiter, Indeed, having to go and then post onto those websites, managing to those website postings and then kind of you know, just very manually go into all of that and use that. So obviously by being able to work with PandoLogic and the pandoSELECT team, the AI base and just the time savings has been a huge thing for us. 

We have seen about a 30% savings in our per-candidate hire. While our cost for hiring has increased, the number of people that we’re able to place has increased by far more than what we’re paying. So this year as an office, we’ll grow by about 110% and our cost for everything is only increasing by about 60% and that doesn’t include the time savings of not having a full-time employee that I would normally have to have in order to just manage the job postings, reach out to all of the candidates, organize it, get it into our system and all of that. 

So you know, I mean, you throw on the full-time employee on top of that, I mean, you are talking about like 400% saving. Having a partner like pandoSELECT that can adapt with us, that can create features for us that we need for our business is invaluable. PandoSELECT as an organization ability to cut through the weeds, recognize our problem and come up with a solution that works for us, it’s awesome. So it keeps me coming back.

[0:17:20.7] TB: Thank you Blake and everyone for joining us here today and now it’s my pleasure to introduce Jon Gacek, who will discuss how tracker fits into our rapidly expanding product suite. 

[0:17:31.6] JG: Thanks Terry, I appreciate the introduction. My name is Jon Gacek. I am the GM of aiWARE Enterprise and I am excited today to introduce to you our new application, Veritone Tracker. Veritone Tracker is based on aiWARE but it’s an additional capability, where you can create a unique scenario around looking for human objects in very, very crowded scenarios. 

So without further ado, I’d like to introduce Ben Ha. I think seeing is believing, Ben is going to walk you through the demonstration. Take it away Ben. 

[0:18:03.0] BH: Hello everyone. My name is Ben Ha from the Sales Engineering Team here at Veritone. Today, I am going to be walking you through Veritone Tracker. This product was built for the person of interest tracking used case. It leverages technology and it looks at the entire person as an object and it allows you to then track and follow that person’s whereabouts across different camera angles and media sources. 

Let’s get right into it. In this example here, we see a person of interest spotted in the subway station. All we need is a single appearance of that person across any footage and we can use that to actually run a search based off of their appearance. So it’s not looking at facial recognition or any biometric features. What it’s actually doing is it’s looking at the visual appearance of that person. 

What type of hair do they have? Do they have a cap on? Are they wearing a backpack? What color is their upper versus lower clothing? All of those different visual distinguishing attributes are now taken into account and then what it allows you to do is run a search and say, “How similar is that person compared to the other possible detections of that person?” So now we’re using this [confidence 0:19:04.0] slider, I have the ability to very quickly go through these different detections that are potential matches. 

I have the ability to review it. Now, we can see that it picked up that same person of interest in a completely different camera angle, different shot within the subway station and then I can continue to build on this. As I am going through this process, I have the ability to now select these different matches of the person and now I could add them into a slide of matches there. So think of this as your user-verified list of all of the moments of where that person of interest has appeared. 

Now, what’s nice about this process is it doesn’t have to be linear. For example, right now I am looking at the front facing shots of this individual but as they start to go lower on the confidence, I might run into shots like this where I actually get the back shot of that same person of interest as you can see over here. So now using this, I could potentially get additional matches that I wouldn’t normally get using the front facing approach as my reference detection. 

So now using this approach, this gives me the ability to now go through and run a search and now I can get back facing shots of that same individual spot at the different shots throughout the subway station like we are seeing over here and now we’re going to get the complete picture of where that person’s whereabouts are and so now you can continue to build off of multiple searches into your selecting matches area.

You can send that off to illuminate when you’re ready and now it can give this personal label, submit it and now it’s going to give you the ability to access Veritone illuminate and all of its analytical capabilities. So for example, I can search on that individual, that person of interest I just submitted through. This will show me all the different results what that person of interest has appeared along with where in the video they are located, as you can see over here. 

You pick up that same individual like you saw earlier in Tracker and then what I can do is run further types of analysis. So I have narrowed down my huge library of videos from that event to a smaller subset of videos that are relevant where I know my person of interest appears. So now I can run additional types of processing on that data, whether it’s transcribing the audio of its present or maybe detecting other objects. 

Maybe recognizing textile for the signs or license plates off of vehicles in the parking lot, logos off of people’s clothing, whatever it might be, it opens up the world of possibilities of what the area of platform offers starting off with Tracker’s really starting point to call it information done. Let’s look at some additional examples. So here we have another person of interest. We’ll call him Brian. 

You can see he’s in the yellow shirt and cap and now what I can do is run a search and it’s now going to go through and reveal additional matches of that same individual and I can see the same individual has been picked up in a different small part of the subway. As I start to go lower on that confidence slider, it also actually gets shots of that individual from the front facing as well and so much like we saw in the previous example, by using a combination of both the front facing and the back facing shots, it’s really going to allow me to build a complete list of the matches of where the individual has it here. 

Let’s look at one final example here. In this case, we have a more difficult situation. We had this person in the black suit as a person of interest. If I run a search off of this, you’re going to see that’s going to return a lot of possible matches because there is a lot of people in the subway station as you can imagine who are going to be wearing the same black suit and so as I am looking through my potential matches, I am going to have a lot of false positives to get through. 

But one strategy that you can take is to now rely on people who are next to that person of interest. So you can see this person is actually walking with this other individual and so now if you focus in on her who is a lot more visually distinguishing the black suit individual, this gives me the ability to now run a search off of her instead and now if we find her, we find the person she’s walking next to. 

So using that search, I can now look at my potential matches and you can see it now spots the same individual in a lot of different other areas. So here is that same couple now from a different shot in the subway station. As you can see right over there, there is another example where she’s spotted actually front-facing and now we get a front-facing shot of her as well as that person of interest. 

So between these detections, we have more than enough information to find that person of interest, the two of them actually across all the other locations they might be. That concludes our demonstration of Veritone Tracker. I hope you found this informative. Thanks for your time everyone. 

[0:23:07.1] JG: Thanks Ben, great job on that demo. I want to give a little more context on Tracker. As I mentioned, we just announced it. We are currently positioning it as a post-event review platform where humans will use it to call down data, to find an individual. What’s really interesting is those individuals could be a crime scene. It could be a missing person, human trafficking is a use case we’ve been informed about. 

We are also demoing it to major sporting event locations around the country, you know, football stadiums, baseball stadiums, racetracks, again, lots of resources are spent trying to find people in those situations and this tool speed that up dramatically. Overtime, you will see us at more and more capability to it but we’re really excited about getting started and the feedback that we’ve gotten this far. 

As a matter of fact, we have one of our partners on, Mitch Thompson from WSI. He is going to share his perspective on Tracker. He and I have done a number of demos together. So Mitch, take it away. 

[0:24:07.0] MT: Hi, I’m Mitch Thompson. I am with SWI Technologies based in Indianapolis, Indiana. We’re an audio, video and digital evidence management specialist that has been around since 1977 and we’re proud of our partners that we work with and Veritone is a very valuable partner at WSI. At WSI is being an audio-video specialist, we like the bold on tools that we provide with our product line and that Veritone provides a capability with aiWARe. 

As everyone knows probably around Illuminate and redaction and it helps our clients be more successful. When Jon showed us the power of Tracker, it really just took it to another level and we were super excited about showing that to our clients because we know the value that’s there and the ability to be able to identify and see what folks are doing but we also know that there’s challenges associated with that with facial recognition. 

Whether it’s the inability to get a clear face recognition or the fact that agencies are just shy from using facial recognition, Tracker gives them that option to begin and have that ability to process people across many video streams. What is great about Tracker is it provides technology that helps law enforcement be able to solve crimes and it also has that wow factor if you will in terms of technology. 

When speaking to our clients and letting them know that there are options out there, when I described Tracker to them, they immediately say, “Yes, we want to see it” and we’re very excited to bring in Veritone to have the ability to show them the capabilities of Tracker, which then opens the door to many other discussions as well. 

[0:25:49.1] JG: Thanks Mitch, we really appreciate your support. 

[0:25:51.7] MM: This concludes the Veritone analyst update and tech demo day. There is a question and answer round available, which is linked in the shownotes of this episode. So feel free to tune into that section as well. 

[END OF INTERVIEW]

[0:26:04.0] MM: This has been another episode of Veritone’s Adventures in AI, a worldwide podcast that dives into the many ways technology and artificial intelligence is shaping our future for the better. Talk with you next time.

[END] 

Guests

Veritone Leadership

Ryan Steelberg, Sean King, Corey Hill, Terry Baker, Whit Walker, Jon Gacek, Ben Ha

Partners

Steve Xeller from Stats Perform; Mitch Thompson from WSI Technologies

Customers

Blake Quinlan from Express Employment Professionals

Subscribe

Subscribe to the podcast updates and never miss an episode again

bg contact