Some thought it was teaching employees how to do something inherently cruel. Others thought it brought sensitivity to an emotionally difficult task. But Kyle Jackson, CEO of Talespin, said it was a very small part of what the company can help enterprises do with VR training.
I interviewed Jackson at Oculus Connect 6, Facebook’s VR event in San Jose, California, last week. We talked in a section of the exhibit area dedicated to Oculus for Business, which is one of the markets where VR is gaining traction.
Talespin is combining VR training simulations with artificial intelligence to create a virtual human platform that can be adapted for a wide variety of training scenarios. Over time, as VR hardware tracking gets better, such virtual humans will be able to gauge your engagement, emotions, and reactions — and then adapt to them. At OC 6, Talespin showed a simulation in which Farmers Insurance teaches claims adjuster trainees how to look for water damage.
Here’s an edited transcript of our interview.
VentureBeat: Where are you based?
Kyle Jackson: We’re in Los Angeles primarily. We have four offices. Los Angeles and the Netherlands are our two primary offices, and then we do our mixed reality stuff in Seattle, and our backend team is in Ottawa.
VentureBeat: When did you get started?
Jackson: We started the company in 2015. The first two years were focused on entering the enterprise space, anywhere they wanted to start. We were going and listening to everyone, whether it was a marketing use case, a training use case, a data visualization use case. We built more than 75 experiences in the first two years. That gave us the conviction to look at the verticals we’re looking at and build the platform we’ve been building ever since. Farmer’s was one of the first big customers we started working with.
Today we’ve bucketed the platform back into three big types of learning. We focus on object-based learning, which is the parts of things, if you think about it simply — the parts of a home, the parts of a car. Then process-based learning, which is what we’ve been doing a lot of with the insurance industry. And then interpersonal conversational skills. You probably saw the “firing Barry in VR” press that got out in front of us in the last month or so. That came out of a number of enterprise clients pushing us toward interpersonal skills and giving people a safe place to try that stuff.
VentureBeat: Did you guys create that?
Jackson: Yeah, we’re the developer. Everything’s in-house. We’re nearing 90 people now, between the four offices.
VentureBeat: What was that a part of? What was the context?
Jackson: It’s not something we actually sell. What we were trying to prove was that you could have an emotional experience with a virtual human. We wanted something that was really broadly applicable. Anybody who sat in it could empathize, whether they had been one one side of the table or the other. It’s uncomfortable and awful every time. If that experience translated into something that was impactful for you — you got sweaty hands, you felt uncomfortable — then we knew we had something that could be broadly applied for training use cases.
The actual platform, the modules being built and sold now, are more around manager feedback. Difficult conversations. You can think about businesses where customers have a bad experience and you have to try to rebuild the trust in that relationship. How do you practice that before you’re in the line of fire? There’s a long list of them — HR use cases, leadership use cases, and then use cases specific to the business unit.
We showed Barry to clients about a year ago, originally. For the last year, we’ve been building bespoke use cases for companies that have about 1,500 to 2,000 initial trainees that they wanted to focus on. Now we’re going toward a whole list of off-the-shelf modules based on that platform that will be broadly applicable and available.
VentureBeat: What’s your assessment of where the enterprise VR market is right now? Is there a place where there appears to be the most spending happening now? Is it on the scale of what people had originally been hoping for? If consumer VR has been slow to take off, enterprise VR seemed like a nice place to pivot. Is that materializing?
Jackson: Yeah, yeah. We started there pretty early on. In 2015, we were singing the praises of enterprise use cases. That’s where we wanted to go. The origin of why we’re doing what we do is [that] we looked at the landscape of how fast jobs are going to change with other technology innovation that’s happening — machine learning, computer vision, AI, robotics, automation, the gig economy. All the stuff that’s happening all that once. That’s impacting large enterprise clients.
We went and talked to them about that. We said, “Is this hitting you? How painful is this? What are you doing about it?” In insurance, it’s already there. It’s about three to five years ahead of a lot of other verticals, because of the workforce demographics.
VentureBeat: Is it a question of how much it costs to train people and tie up a veteran doing that?
Jackson: It’s even more macro than that. How are we even going to deliver this service in the future? In insurance they had an imbalance of supply and demand for talent. It was a very attractive field for a lot of years, because you could build up a book of business, and then it folded over on itself. You had a nice residual–it was a very entrepreneurial thing to do 20 or 30 years ago. You have a lot of baby boomers that have done extremely well.
Over time, though, it’s become viewed as stale, boring, whatever industry. You don’t have a lot of young people who are interested in that work. They have a real talent crisis and imbalance. We’re hearing that not just in insurance, but in telco, utilities. A bunch of industries are following the same thing. While that goes on, they’re also trying to rapidly insert AI and other algorithms to start eating away at some of the work. Instead of the basic data entry or data processing, they’re using drones and computer vision to do the first pass on a claims process. They have a people problem.
While all those things are going on, our thinking was, all this just means we need to accelerate learning. People are going to be moving in and out of new positions faster, or work is going to change and evolve so fast that they’ll always be reskilling. That was the 2015 hypothesis. We went out and tested that with a couple of key verticals to start.
The driving factors for us to select verticals were–companies that already historically had large distributed work forces, and that they operated in environments that they don’t control. If it’s not retail or hospitality, something like that, it becomes incredibly costly and difficult to train. When you have those two dynamics, the pain point is real. We focused there, which is why we ended up in insurance and telecoms.
Since then it’s just been expansion. About 50 percent of the business today is in the insurance sector. We launched an industry-wide platform at InsureTech in Vegas earlier this week, which was basically–after Farmer’s talked very publicly about our work in 2017, we got all of the major insurers reaching out to us. They all had “different” use cases. But as we started to distill it down, though, we realized that they were all very similar. We started building a platform and a product that addressed those common use cases.
VentureBeat: What’s being shown here?
Jackson: One of the most important things for us is obviously removing barriers. Controllers are a barrier, especially if you’re looking at reskilling a generation that’s not used to–they’re not the console generation, the gamer generation. We’ve been pushing on the idea of using only voice and gesture for your inputs. Our conversational, our virtual human platform, only uses your voice and gaze. The next step was to get rid of the controllers and go to hand tracking.
VentureBeat: Is this Quest only?
Jackson: Yes, it’s Quest only. We worked with the Oculus team for the last year.
VentureBeat: Is there a reason you wouldn’t want to use the hand tracking in the Rift? Or is the Quest just more intuitive?
Jackson: They brought it to the Quest first, to the SDK. That was where their focus was, because of scale. In terms of our R&D focus, we started there.
In this use case, you’re going into a residential home and you have to spot issues. There are some red herrings. It’s built for a demo space, so it’s not very deep. You don’t have to really know the job to get a good experience.
VentureBeat: You have a fire hazard here, that sort of thing?
Jackson: Yeah, yeah. You walk around and you’re like, “Ah, that’s something I need to tag.” You use your hands to open cabinets and close them and touch things. You tag different hazards. It’s very intuitive, very simple, which was obviously a huge part of what we were trying to prove. You can completely get rid of the friction at that point.
What we’ve seen is that with going to conversational, and now I’m sure with hand tracking, this becomes even less scary for that older generation that’s having to reskill. You’re not actually having to learn new interfaces. You’re using the things you know how to use — your voice, your gaze, your hands. We think that will become a huge acceleration for broader adoption of VR training.
VentureBeat: Is anybody telling you some good numbers about what they save?
Jackson: We do have a bunch of data there. Each client is very sketchy about letting that go out publicly, of course.
VentureBeat: I liked that virtual surgery one I saw today. Something like 82 percent no longer need to be hand-held in their surgery training?
Jackson: That was mind-blowing. I got goosebumps when she said that. In that use case you also have another factor, which is the actual training environment is extremely expensive for downtime. We have a different issue here, which is you don’t have any access to the training environment. The environment is damaged homes, or in a telecom use case it’s a corporate office park where you’re going to install fiber. You can’t replicate enough use cases to have meaningful training. What they used to do is try to build mock use cases, and then you’d travel there and train. This gives you the ability to go infinite in terms of the number of use cases.
When you A/B the learner populations from ones that used the old learning modality and the ones that had VR involved, once you put them out into the field, they saw more than 20 percent increase in decision accuracy. That’s meaningful when your average claim cost is between $7,000 and $10,000.
VentureBeat: Is there anything like firing Barry in the insurance setting?
Jackson: [laughs] Those use cases are obviously more intangible. When you get into the soft skills stuff, it’s really hard to measure. If you think about the old way of doing it, you’d have different facilitators come and give role-play at different offices. You’d have no commonality as far as how good or bad they were. You’d hope they were all good, but they facilitate differently. You’d have anecdotal measurement at best.
The idea of using simulation, soft skills simulation, is that one, you have a safe place to do it. You’re practicing by yourself. You’re not doing it in a group of peers. A lot of people aren’t good actors, especially in front of groups of peers. It’s uncomfortable. If I really want to practice a soft skill that I want to improve, that’s not the right environment. On top of that, there’s no consistent measurement.
Obviously here it’s a realtime engine. We can measure anything, from your sentiment to your gaze to what you said and how you said it. What conversational paths did you open up, or not open up? We’re getting new data. That starts to bubble up all sorts of additional performance conversations that they weren’t able to have before until it was too late.
You can also practice against a variety of virtual humans. We’ve seen some interesting data already where the same scenario with different virtual humans–if you go in and try to sell a male CEO services, and you go and do the exact same scenario with a younger woman CEO, we see people acting differently. You can confront them about that. It opens up a discussion around bias. There are these intangible, weird benefits that come out of this type of learning modality.
VentureBeat: What else is a good vertical for you guys, besides insurance?
Jackson: The virtual human stuff is going across. That’s HR-focused. It’s going across everything. We’ve done some stuff in health care. We just did a really big study in the professional consulting world, where they put 2,000 learners through a leadership bias module. That’s going to be published later this year. They did that for the purpose of seeing whether or not this was effective at scale. The data looks really good. We’re about a week away from finishing that, so we have a pretty good idea.
We launched a platform for the whole insurance industry. The goal there was to go from this bespoke development model, which has been what’s going on in the enterprise space for the last while, to broadly applicable solutions. We’re trying to demonstrate that across an entire vertical. We started launching off-the-shelf solutions last week. We’re working with three of the top 10 carriers already, and we’re growing. Once we’ve proved that works, the virtual human stuff is helping us expand horizontally at the same time.
VentureBeat: Did you raise any money at all?
Jackson: We raised a Series A, or a first round, of $5.2 million. That was last year. We operated the company profitably for the first couple of years. We acquired a mobile game studio in the Netherlands in 2017, and then we raised in 2018.