Today, we’re talking to Grin Lord, the founder and CEO of mpathic AI. Their technology uses AI to analyze text and voice communication to strengthen a company’s culture and improve the performance of public-facing teams.
Transcript
Mark:
Welcome to PeopleTech, the podcast of the HCM Technology Report. I’m Mark Feffer.
Mark:
Today, we’re talking to Grin Lord, the founder and CEO of mpathic AI. Their technology uses AI to analyze text and voice communication to strengthen a company’s culture and improve the performance of public facing teams. They call this “empathy as a service.” We’ll talk about what that means and how it works on this edition of People Tech.
Mark:
Grin, thanks for being here, and what is mpathic?
Grin:
Great question. So right now, mpathic is being used in organizations to really transform and create a culture of empathy. Pragmatically, we’re a B2B SaaS company, and we have an API that it integrates into existing HR SaaS platforms and analyzes conversations for empathy, and makes suggestions or corrections to what people are saying, if that’s applicable. So for folks that are doing interviews, that’s more of the case. In other settings, we’re really just returning metrics around how folks are performing with empathy collaboration, partnership, synchrony, and associated things like that.
Mark:
How does it work?
Grin:
Yeah. First of all, it’s an AI. We’re an AI company, and we have machine learning models that we’ve trained on conversational data. So that’s, in an HR context, could be between a manager and employee, or it could be with someone who is interviewing or a recruiter and a candidate.
Grin:
But we have a lot of different conversational data that our models are trained on, and we have folks that review those conversations and essentially identify, what are the key ingredients in those conversations that are empathic? And we label those. So, what are the things that they did really well in the conversation that objectively can be ranked as associated to empathy?
Grin:
And once we’ve labeled those conversations with tons of different aspects, so it’s not just like, “Was this good or bad or a felt sense?” We actually have years of psychological research around specific phrases and things that you can say to increase empathy that cross-culturally are perceived as being more empathic. For example, something like an open-ended question versus a series of closed ended questions driving the discussion into a certain place. The open-ended question is perceived as more empathic.
Grin:
So, we label this data and then we train a machine learning model to recognize those same skills and novel conversations that it hasn’t seen before. So, I think probably most folks in tech spaces are familiar with AI, but for anyone that isn’t, I think of our models like a human baby in some ways, and they only recognize the patterns that you show them.
Grin:
So, we train them on all of these different patterns and we say, “These are the good things. Look for these,” or, “These are the really bad things. Definitely don’t do this,” and show the model many, many examples of that, and then we it’s faced with a new data set, it can pick those out.
Grin:
So, that’s how the detection works, where we are detecting, “What are the good and bad parts in these conversations?” And then we have another step, which is, what do you do about it? What is that actionable insight or the thing that, “Okay, now I know I really need to improve my quest.” Let’s just use that example. Like, “What am I supposed to do?”
Grin:
And then we have another set of models, deep transformer models, that basically make a correction to the thing that people said and say, “Here’s a better way to be more empathic. Here’s some choices for you to rephrase.” And those are using generative AI, which is a little bit of a different process, but has similar things where we have to train it on all sorts of examples of how to improve speech and interaction.
Grin:
So, that’s one set, and then we have some other things, other models and detections that aren’t so concrete or based on exact skills of what people do and say, and more things like, how are people synchronizing over time in their voice, vocal patterns, pausing, words they use how they phrase things… Which there are several research studies on that have shown that when people synchronize over time, they’re actually feeling more alliance, rapport, and empathy.
Grin:
So, those are defined in a different way than saying, “These are the good things,” and, “Do more of that,” and, “Here’s a correction.” Those are more unconscious indicators that people are getting along. So, our models look at both of those and we’ve trained them using experts in psychology.
Grin:
And for myself, I’ve been doing this for 15 years in a university based setting. So, we’re taking some of the academic concepts that psychologists and folks in IO and in other organizational psychology have worked on for a long time, and trying to really operationalize that in conversations and train machines to do the same thing.
Mark:
Now, why does this matter to a business?
Grin:
I mean, at the core problems with empathy, especially now in the great resignation, are related to churn and satisfaction with one’s job. Even in the interviewing context I think, we’re seeing there’s scarcity for certain roles and people are no longer satisfied at working in places that don’t treat them well, and culture is important.
Grin:
And especially in larger organizations where it’s harder to have a unified cultural approach, using an AI to back up the company culture is really helpful. So for example, if you have teams coming from different places all over the world, and we actually have this with some certain customers where they’re onboarding a team, let’s just say in Romania and Hong Kong to work with a team in Seattle, and everyone’s on Slack talking together or sending emails together, there are going to be certain misinterpretations or things that happen cross-culturally that can lead to reductions in productivity or problems with this agreement. We’re trying to nip all those problems in the bud. And it’s not just a cross-cultural problem. It happens between any person, right? It only takes one message that’s misinterpreted in text based formats, which we’re all doing in remote work, to lead to a series of problems.
Grin:
So, having an AI robot that’s consistently evaluating you and prompting you will help to communicate that company culture in a consistent way, and then will help the individuals to realize, “Oh, I have different choices I can make here.” And we’re not telling you, “This is the only way,” there’s still agency. That person can decide, “Yeah, I want to do this.” Or, “I really want to grow in this area, give me more corrections like this,” or, “I want to learn more about that.” It’s all about doing this in a way where there’s feedback in an objective manner that helps companies retain their employees longer. So yeah, I think at the core that’s what we’re trying to do is reduce churn and help with retention, especially in a time where people aren’t interested in working for companies that are non-empathic.
Mark:
It just occurs to me to ask, how is it capturing these conversations? I mean, I’m assuming it’s recording it somehow, but could you step me through that?
Grin:
Yeah, that’s a really good point because our technology can be applied in a bunch of different areas. So texts, email, Slack, text-based things, but can also be used in phone conversations as long as there’s a text layer. So for example, if you have customer support or agents, or if you have Manager 360 reviews or one-on-ones, things like that where there’s actually recording, we can give close to real time feedback on empathy in those settings.
Grin:
And different companies implement this in different ways. So again, we’re not a set product. We integrate with what other people are using. So, a lot of different companies have different platforms for how their managers give feedback or how employees talk to each other, and we integrate into that seamlessly to give those assessments and feedback. So, it would really be up to the company where we integrate and how we work.
Mark:
It’s really fascinating, and it’s really different. I never thought that I would see someone saying, “They’re applying technology to empathy.” Are there competitors, or are you the only people in this space right now? What’s the landscape look like?
Grin:
Yeah. It’s different because we’re really… We’re taking a lot of the research and the evidence about, what do we know about human behavior and speech, that we can take some of these academic lessons and lessons from psychotherapy and other contexts and apply them in these commercial settings?
Grin:
I haven’t seen anyone doing that exact thing where there’s this very… We’re not just a writing assistant. We’re not just trying to say, “Here’s how to phrase your stuff better.” It’s like, “Maybe you need to have a meeting with this person.” “Hey, don’t send this.” “It’s time for a one-on-one.” We’re giving specific behavioral prompts on, what do you do with all of this data analysis of your speech?
Grin:
So, we’re definitely in this era where there’s a lot of performance tools and review, but I think from my background as a psychologist where it breaks down is that people have endless amounts of analysis and data at their fingertips, but they don’t know how to integrate that into what they’re doing.
Grin:
So, we’d like to come into those conversations and say, “This is exactly how you do it. Here’s what you need to try. This is the human way to apply all these insights.” And I really don’t see anyone else in the market doing that.
Grin:
But folks doing similar to things would be Gong.IO or Cogito. They’re using it for more call centers and saying, “Okay, there’s a mismatch. I’m trying to do collections for a bill and there’s a mismatch and empathy and the tone. Change your tone so you can do collections,” which is totally different use case, but has some of the same principles.
Grin:
Again, we differentiate because our labels and our models are proprietary and focused on this specific data set of humans talking, not in commercial sales or other things. So, nothing’s off the shelf. We created it ourselves specifically for this use case.
Grin:
But yeah, I guess there are similar companies out there taking the idea of, “Wow. Now for the first time in ever, we have technology where these large scale AI models can be run in close to real time.” We can take enormous pieces of data,, put them through huge models, and output something immediately.
Grin:
When I first started working on this problem in the university context, we would take a one hour recording of a call and analyze it for different skills. The latency would be sometimes I want to say hours to a half a day, depending on how long the call was, because we just didn’t have the AI computing power. So, these ideas have been around a long of time, and now we’re catching up to this moment where the technology’s there. And then we have this huge problem of employee churn, satisfaction, remote teams that have never met each other that don’t have time to build trust and rapport. All of this is convening together to make a really interesting application of some of the principles and technology that we’ve been doing, in a sense, manually for a while
Mark:
Now, as you look out toward the future, how do you see the business growing? And I mean that in two ways, the business meaning you and your competitors that might spring up and all of that, but also just how this application and this approach to empathy is going to evolve. What do you think?
Grin:
That’s a really good question. I think there are some things that need to be worked out in terms of… With this really novel technology, how do we make it acceptable and how do people, how can they integrate it so they become truly empathic and not just corrected in their speech or at the time, but really how do they change their behavior? And that to me is the most important, new learning that we can make is… For me, the end goal is not increasing retention in sales for a company. It’s like, “Can we actually take these learnings and apply people to be better listeners, to be more compassionate at scale with AI?” I’m very interested in that.
Grin:
I think broadly we’d like to expand beyond empathy and into essentially becoming an auto-tune for any kind of tone or content. Some folks are highly empathic, to the point where they’re non-assertive or can’t even get a point across because they’re listening so much. Our models have the ability to do that as well. So I think in the future, it will be more than just empathy and more like, “Let’s assess and tune to the type of feedback that’s required in this moment.” Could be empathy, could be assertiveness, could be being more dynamic or whatever it is. And we have those capabilities now, but I think it’s more figuring out the use case.
Grin:
And then, the technology will own only get more impressive. Some of the generative AI models that are being used today, things like GPT-3 and coming on to GPT-4, huge advances in speech signal processing and production of language. I mean, there’s also far applications into social robotics and, “How can we use these very human things to help in those other contexts?”
Grin:
So, I think this is just a really exciting and ever evolving field, and even from the time I founded my company till now, there are huge advances. In some ways the future is now. If I listen to this a few months from now, I’ll be like, “Oh yeah, we’re there.” The field is moving so fast.
Mark:
Well, Grin, thanks for taking the time today.
Grin:
Yeah, no problem at all. Thank you.
Mark:
My guest today has been Grin Lord, the founder and CEO of mpathic AI. And this has been PeopleTech, the podcast of the HCM Technology Report. We’re a publication of Recruiting Daily. We’re also a part of Evergreen Podcasts. To see all of their programs, visit www.EvergreenPodcasts.com.
Mark:
And to keep up with HR technology, visit the HCM Technology Report every day. We’re the most trusted source of news in the HR tech industry. Find us at www.HCMTechnologyReport.com. I’m Mark Feffer.
Image: iStock
