Storytelling About TalentGPT By Beamery With Sultan Saidov
Beamery, a company that focuses on creating people-centric technology, has launched TalentGPT, a set of AI capabilities that allow them to serve every stakeholder with a new form of assistance. TalentGPT is Beamery’s own set of AI models that can understand what a job title means in a particular company and how it can help people figure out what skills they need to develop to apply for a certain role. The technology embeds in conversational experiences and assisted experiences to provide contextual guidance to job seekers.
The aim of TalentGPT is to provide personalization at scale by contextualizing the information within the company, team, and job. One of the main use cases for TalentGPT is for job seekers. The technology provides contextual guidance beyond just having a better careers website. It helps job seekers consider jobs they may not have considered before, provides feedback on resumes and cover letters, and even helps job seekers prepare for interviews by suggesting relevant questions and providing tips on how to answer them. TalentGPT also provides managers and employees with insights into their own career paths and skill development.
Give the show a listen and please let me know what you think.
Listening Time: 26 minutes
Enjoy the podcast?
Be sure to check out all our episodes! Thanks for tuning in to this episode of the Use Case Podcast!
Beamery’s Talent Lifecycle Management platform allows enterprises to create more human experiences for all talent, and unlock the skills and potential of their global workforce. We help organizations identify and acquire top talent, move the needle on DE&I, close important skill gaps and retain top performersFollow
Storytelling About TalentGPT By Beamery With Sultan Saidov
William Tincup: [00:00:00] This is William Tincup and you are listening to the Use Case podcast. Today we have Salton on from Beamery, but we’re gonna be talking about a product that they’ve just launched called Talent, G P T by Beamery. So, uh, why don’t we just jump right into introductions? Salt, would you do us a favor and let’s do two introductions.
Let’s do introduction of yourself, but let’s also talk about talent, G P T and be.
Sultan Saidov: Of course. So I, uh, I’m Sultan. I am one of the co-founders of Beamery. We’ve, uh, been building [00:01:00] Beamery for nine years now, and what we focused on from the beginning of building Beamery is creating people-centric technology.
You know, HR. Talent systems tend to be very process centric, and in being people centric, you have the opportunity to focus on the kinds of experiences and solutions that allow companies to hire people based on potential, develop people based on potential. And what we’ve announced with talent, G P T is something we’ve been working on for about four and a bit years now, which is.
A set of AI capabilities that allow us to serve every stakeholder, whether you’re an employee or a manager with a new form of assistance. Right. And what type G p T is, is our [00:02:00] own set of AI models. In 2018, we started working with the first generation of the new large language models. Right. Uh, at, at the time we actually started working with, um, the, uh, large language model by Google called Burt.
Right, right. And over the last four years, we’ve essentially created our own fine-tuned large language models that really focus on. The domain of companies, jobs, skills, organizations, industries, and allow us to use AI to understand what does a job title mean in a particular company, right? How do we help people figure out what skills they need to develop to.
Apply for a certain role to, um, uh, to develop themselves. Um, and that technology and that AI that we’ve built over the years to, uh, to do those insights. We, we often refer to those as inferences. How can we infer what’s relevant to you, um, is [00:03:00] now embedded in these, you know, new experiences, conversational experiences, uh, assisted experiences, and so,
William Tincup: You know what I love about this is one of the challenges of, of the large language models is that they can give you gen, they can give you information in general, which is great.
They can kind of do that. But what you solve for, which I think is super interesting, is how you can get personalization at scale. So within that country, within that, excuse me, company, within that team, within that job, it can be be contextualized, which I’m, I’m sure that wasn’t easy. Um, but, but how do, let’s give the, the audience a, uh, some kind of ways that they would use this.
So we’ll just kind of break down kind of Okay. Let’s say you’re doing this, this is how you would use talent gpt. Yeah. So
Sultan Saidov: we’ll, we’ll go through a couple of [00:04:00] scenarios. We’ll start with, I guess, the highest volume scenario if you, if you’re a candidate or if you’re a job seeker, traditionally. You know, your experience as a consumer is, uh, already night and day better than most applicable can experiences.
Um, you know, we’re used to a sort of very low friction recom, you know, recommend me a film through Netflix and give me a root map through Google sort of experience as a consumer. Now, talent, G P T, um, provides some equivalent in the context of being a candidate, but it’s more than just. You now have a better careers website.
If you’re applying for a job, it’s contextual guidance. So for example, if you’ve started applying to a company, One of the things that, um, talent G B T does, and this is an extension of something we’ve been doing for the last couple of years in our AI models, we help you consider jobs that you otherwise may not have applied for, right?
A lot of this is actually to drive inclusion. So, [00:05:00] you know, a lot of people might, especially people from underrepresented backgrounds and often, um, also this is gender-based, will, will not have the confidence to apply for roles based on traditional job descriptions and the experiences required. And if you help, um, analyze people’s backgrounds to suggest, Hey, here’s some things that you might have as skills and therefore that might make you able to do some of these roles in different departments.
We essentially analyze based on your background, your resume, what does this make you likely a fit for within this particular company? Because each company means very different things. Yeah, yeah, yeah. With different jobs, right? So, so it might be that, you know, you have a background in, let’s say, Data science and you’d never have considered applying to, um, a role in.
Accounting, but suddenly there’s some very interesting roles in accounting that are actually primarily about data science rather than what it used to be. Right. Um, you know, you have crazy new roles like prompt writers, um, fitting to this conversation, but there’s, there’s [00:06:00] certain backgrounds that, um, make that.
Uh, a role that you might be more likely to be good at, you know, people with creative backgrounds and so on. Um, there’s lots of niche roles that have appeared over the last year, like blockchain engineers, and we can see that people from certain backgrounds might find it easier to train into those roles.
So, as a candidate, you, we give these recommendations around, Hey, you know, here’s paths that you might wanna consider in any company you look at, but we also show you, um, uh, potential content that may be relevant, um, about that company in the context of what you’re looking for, you know, content about.
Companies, policies, programs, approach to diversity can be spread across a lot of different places
William Tincup: and usually are Yeah.
Sultan Saidov: Yeah, exactly. And you, and you can, um, contextually bring that to something, not generate any new content. You’re not sort of, uh, imagining something new about the company that’s, you know, something a company could choose to use these tools for.
But the stuff that a company’s already got out there about themselves, you can make it easier for candidates to find this stuff. And, um, and to bring it up even in conversational interfaces. Now, [00:07:00] if you. That sort of mindset of, you know, I’m a candidate. Further, let’s say you’ve become an employee at a company.
We can go even deeper in terms of this sort of journey map or Right, right. Ticket as a sort of career. Yes. Well, it’s one of the
William Tincup: thing, cause it’s like you look, you’re re, you’re re-imagining internal mobility where it’s, it’s push-pull. Right now we have to put out a job post and then people then self-identify and then apply for it, et cetera.
Whereas what you’re talking about, it seems like it could tap people on the shoulder and. Your experience says that you’d probably be pretty good at this if you, if you’re interested.
Sultan Saidov: Yeah. That’s, that’s spot on. And I think it’s, oh, um, it’s also not just about full-time jobs, but about Right, right, right.
Continued development. You know, the, going back to kind of what, uh, what I said about when, why we started the company to try and create inclusion and people centricity. A lot of experiences, both when you’re applying for a job, when [00:08:00] when you’re an employer, very transactional. You know, you have like a performance review, whether it’s once a year or otherwise, and then you might have a job that you apply for.
But we are living in a world where it is possible to encourage continuous self-development, whether it’s by helping you find colleagues who you can learn from, whether it’s mentors or buddies that you can develop skills with. And help you have continuous insights into, hey, if you want to try something out, whether it’s a company that might have projects or gigs or opportunities that you can explore, how do you connect somebody to that window of opportunities in a way that is exploratory?
So for example, if you imagine, you know, you are using Google Maps, you can say, I wanna get to this destination and it can give you choices. Do you wanna ride a bike? Do you wanna go on a bus? Like what happens if there’s rain? There’s equivalent exploration that you can give people just. Providing transparency.
You don’t need to create a new process. You can say, Hey, you wanna become a product manager? Here’s people in your team you’ve done who’ve done that. That’s just data that exists and sort [00:09:00] of buried away. Um, but we can bring that to the surface and help somebody browse, Hey, here’s some colleagues you can speak to who’ve been on this journey.
And it’s amazing because what we found from, you know, we’ve been doing this, this type of ai, career pathing technology for years. This talent G p T piece is an extension, um, that is, uh, more conversational of things that we’ve, we’ve been doing for, for some time. But what we found is, you know, transparency is one of the biggest drivers.
Retention, particularly for people you know from underrepresented backgrounds. Because often the reason people leave is not because there aren’t any colleagues they could talk to is because it’s not visible to them. They don’t know somebody like them who’s been through that journey or they don’t know what roles they might want to consider unless their manager tells ’em, Hey, this is a role that you might have the skills for.
Right. Right. So empowering people to do that can, can have a big impact.
William Tincup: So, dumb question. Um, can folks just buy talent g p t by often by themselves? Or is it a part of, uh, Beamery and like their, you [00:10:00] know, thousands of clients? Is it a part of the Beamery experience? And this is an added bonus of just being a Beamery client.
Or do they have to, or can they buy it by the off to themselves?
Sultan Saidov: Yeah, it’s, it’s a great question. Um, today it’s a benefit of being a Beamery customer and a Beamery experience, and you are a Beamery customer. Um, you know, these experiences can be embedded in what you offer your candidates, your employees, your recruiters, your managers.
Um, but it’s an interesting question because, you know, we, the, the technology behind talent G P T are fine tuned large language models. You know, we now have, um, APIs that. Some of our partners. Identify skills for jobs and u use this in their own products. Um, so I think over time it may be that we start offering this as, you know, direct experiences, um, both for candidates and for partners and other people we work with today, that this is essentially an embedded benefit.
Uh, and the talent G GPT offering is embedded [00:11:00] for every stakeholder. You know, we’ve started with candidates and employees so far. You know, there’s equivalent things we do for managers to help ’em figure out, right, do I actually open this job? Do I look internally? It’s this, uh, assisted experience. That users are, you know, carefully trained and bias audited models in, in, in different circumstances.
But, but yeah, you’d have to be a Beamer customer today to use them.
William Tincup: Right, right. So how do you, you know, some of this is okay, I don’t know what, I don’t know when, when they’re a Beamer client, you’re rolling this out. How do you onboard them or how do you see the onboarding experience to this in particular?
Cuz some things within Beamer are just, they’re just super simple. Like there’s, it’s just self-evident. You can do this and just click, click, click, and it’s done. I could see this one needing a little bit, a bit more explanation on how they could use it to their betterment. Totally. So how do you, how do you see that plan out over the next couple of of months?
Sultan Saidov: You know, this is something that we’ve spent a lot of time testing over the years, so, [00:12:00] A couple of years ago, you know, unsurprisingly, we use our, our own recruiting candidate processes.
William Tincup: I’d be angry if you didn’t, but Yeah, fair, fair.
Sultan Saidov: Yeah. Um, so, so a couple of years ago we, we started some experiments, you know, even in candidate experience, like what actually, um, creates better outcomes, more inclusion, more engagement, and one of the things that is risky is, The extent to which something conversational might be taken at face value as opposed to something that is non conversational, right?
For example, if you tell somebody, Hey, you know, here’s some jobs at the company and here’s how, uh, how many of these jobs have like relevant skills to you, and you sort of have insights about jobs, you’re just helping someone browse better. It’s a little bit like, you know, if you are, uh, on Netflix and we flagged like recent films versus recommended films, and I tell you why they recommended.
It’s different when you suddenly have a, you know, a piece of text tell you, Hey, look at this job. You know? And I [00:13:00] think as human beings, inherently we, we’ve evolved to be conversational, you know, storytelling. Um, uh, Thinkers and the way that that kind of information’s consumed can make people jump to conclusions, even if that’s not what you know is a text intended.
Right? So things that we’ve um, been careful about is where do these things become safely embedded in other visual products where there’s not actually any sort of conversational chat for ins, for example, you know, if you are a customer of ours and. You connect a learning management system, which is, you know, where people do their training content.
We use some of, um, these new technologies to help extract and understand learning courses and say, well, if we’re gonna recommend a learning course to someone when they’re trying to figure out what job to do, it helps to. Understand better the learning course and be able to sort of map it to different career parts.
So that kind of stuff can be, you know, in the background invisible. Everyone just gets it because from your perspective, all that’s happened is [00:14:00] you’ve just been recommended some learning courses and clearly some technologies doing that and bringing them out of a different system and showing them to you.
Um, and so that kind of stuff is very natural. You know, you are already browsing this stuff. Suddenly the browsing just got a little bit. Um, when it comes to new experiences, like suddenly there’s an assistant, um, in an experience that you’ve been using before, what we’ve been careful about is how does the assistant get shown up and how do we avoid introducing some kind of, you know, bias decision making in how you are, you are using that assistant.
So, for example, we only use very careful. Audited and long tested models for anything that involves a recommendation. We don’t suddenly sort of use text to tell you a recommendation or user model like, you know, chat GT or something like that. Right. Um, so, so in the process of, you know, figuring out how do we roll this stuff out, one of the things we’ve done is made a lot of the new conversational pieces op, both optional and something we spend a bunch of time testing, not not testing, just to figure out, [00:15:00] you know, what is the impact, but how does it compare in terms of.
What does it allow a person to do that they haven’t done before? Because it’s not just a question of you ask the same question, you get a slightly different experience. You know, if you are a manager trying to figure out, do I open a job and you’re using a, you know, traditional product or interface, you’d probably start by just clicking on open job and just going through that motion.
Whereas if in that process, rather than having a screen with a bunch of menus and something you click on, right, you ask you. Let someone ask, Hey, can you help me open this job? And in response, you don’t actually help you say, well, before you do, here’s some people in your team that you might wanna talk to or continue developing.
That becomes a whole new thing, right? Um, and, and we have the option of helping people th. Think about, uh, you know, scenarios that wouldn’t have necessarily come to mind, right? Like, here’s some people you can train or do this, um, or consider this. Um, which I think, you know, we, we, we want to be thoughtful about to make sure people really get the most out of this.
So you’re spot on. You know, we, we we’re careful in those [00:16:00] areas, um, whereas other areas is just automatically plugged in, in a better experie. Right. What
William Tincup: I, what I also, uh, love is, uh, obviously with what you’ve done with Beamer, you’re sitting on a huge database of, uh, folks that do, that we’ve recruited, that we’re marketing to, et cetera.
So candidates, some of which got jobs, some of which did not get jobs, and then you’ve got employees that you. You’re changing the nature of kind of push pull. So when you use that scenario of, help me, help me, you know, sign up or help me create this job, which I think is fantastic on the employee side or even candidate side, it’s, it can then go and see who would be, who should apply to that.
So what I love about that is it’s, it’s actually you as a candidate or employee, you don’t have to. As much like, again, thinking literally, um, or, or, or hierarchically of, of moving, you can think about different things because [00:17:00] actually, you know, talent, G P T is actually looking at your experience and then serve, offering up different things that you might not have thought about.
Yeah. So it’s changing the nature of job search in a, in a way, or even mobility in a way.
Sultan Saidov: I love that. I think part of this is trying to make it, um, about continuous self development. Right? Right. How do you make that, you know, we live in a world where, The time it takes for most people’s skills to be redundant in the role they’re doing has shrunk from, you know, 10 years, not long ago to every couple of years.
And so everyone, no matter what you do, you know, has to think about continuous self development. But it’s hard, right? Because you don’t know. Where the world is heading when you dunno where your company is heading. And so being able to have, you know, a continuous, interactive experience around, Hey, you know, what do I try, how do I develop myself?
Is there an opportunity that’s interesting? Um, and making those things embedded in, you know, the tools
William Tincup: you use, that sounds like fun, right? Like that’s not work. That actually sounds like fun.
Sultan Saidov: I think [00:18:00] that’s one of the things that, you know, outside just talent G B T, you know, with, with the AI evolutions that are happening now, I think the element of making something fun and interesting is gonna be perhaps one of the most impactful things in different spaces.
I was, uh, talking to someone recently about, you know, take something like politics and elections. Not many people. Engage in reading candidates policy documents. Right. But if you, if you created a Yeah, exactly. But if you made a chat saying, Hey, like, talk to this nominee, you might be like, Hmm, maybe I’ll ask them a question.
Suddenly something that was incredibly tedious. Become somewhat interesting.
William Tincup: Um, and that might legislation, I mean, you know, taking it to anything that’s boring, uh, technical documentation, things like that, or anything in the healthcare space. It’s like, you know, whenever you get a prescription here in the States, it, it comes with this piece of paper that’s stapled to the prescription that no one reads.
So it’s like if you could interface with that and say, Hey, [00:19:00] uh, got a quick question. Uh uh, this. And it answers back. It says, yeah, okay, you shouldn’t do that. Like, uh, I, I love it. I, two things is, I think
Sultan Saidov: touches on, sorry, go ahead. No, no, no. Finish your thought. I think this also touches on, you know, what, where the risky areas come in, because if there’s certain areas where, you know, something being inaccurate.
Poses less risk, but being inaccurate at a prescription or telling you what a presidential candidate thinks is very risky. That’s right. Um, that’s why, you know, we’ve taken the approach of creating these, these fine trade models and these, um, AI things that we can audit in the areas where we actually have to be careful because I think Right, this stuff is very impressive, but it carries a lot of risk if you, if you aren’t thoughtful about it.
William Tincup: Oh, a hundred percent. That’s where the ethical AI and audited AI conversations really get interest. You know, when you, when you start thinking about, okay, we need to go about this and check our algorithms both internally, but also [00:20:00] probably occasionally have a third party come in and take a look at what we’re doing and make sure that there’s no unintended, uh, consequences of, of what we’re doing.
Yep. Good. Couldn’t it be more how does. So I’m super sarcastic, right? So how dumb question, dumb question alert. Um, how, how does, how does talent G P T or LA large language models, how is it consuming sarcasm?
Sultan Saidov: You know, I, um, I think this is one of the areas where. It’s really impressive how well it seems to, especially G P T four, the latest model, right? How well it seems to understand, uh, jokes and recreate jokes. You know, this is the kind domain where talent, G P T, what we do doesn’t really touch that, you know, we don’t touch, um, because we’re very fine tuned to helping recruiting candidates.
William Tincup: There’s there obvious sarcastic there. Yeah.
Sultan Saidov: Yeah. But the large language model that do [00:21:00] this, I. I don’t know. Um, you know, obviously they’ve looked at a lot of data points and analyzed, um, enough language to be able to create these interpretations, but I, I don’t know how well, you know, the, these models actually work in practice versus how much of it is in the eye of the placebo.
You know how, well, I think it’s what,
William Tincup: I think it’s what y’all build is the, the more the large language model knows. Like, the more I interact with it, the more it, it knows me. It’s, it’s kinda like voice to text in a way I’m thinking of it is that at one point it knows, cuz I’ve corrected it so many times, it knows what I mean when I say something.
So I think again, Maybe not, maybe not the large language models, but more, more what of what y’all have built that’s specific and you can see how it’s personalized to the company, to the job, to the manager, to the candidate. You know, like all that stuff. It’s, it’s personalized. So I think it gets [00:22:00] to know you better and can start to kind of suss out whether or not something was said in a sarcastic tone or whatever.
Um, 2, 2, 2 questions. Uh, left. One is, You know, you’ve been building this thing behind the scenes for a long time and uh, and that’s something that Beamer does a wonderful job of innovating. So when you get to show people talent, g p t for the first time, what do you love to show them?
Sultan Saidov: There’s quite a few things. I think for me, first thing, the thing I’m, I’m most excited about is the ways in which we. Give people something completely new as opposed to just making an existing thing slightly better, right? So for example, example. We have a lot of insights about how, you know, how long a job’s likely to take to fill and like where is the supply of talent if you’re gonna start recruiting for this job that historically, you know, a manager [00:23:00] is really gonna look at because you know systems, most managers don’t use HR systems.
They leave that to HR and talent teams, right? Where suddenly we can, we make that insight so accessible and relevant to someone who is actually hiring people and managing teams. And it allows people to really. You know, change the boundaries of where are you spending your time. It means that you spend more of your, you know, actual human time on people’s development and conversations.
And it means that if you are partnering with your recruiting teams and your HR teams, they can focus on the people time too. And, and that’s not just a case of taking something existing, like, write me an email and. Saving time, right? Fun. But it’s, it’s creating something completely new, which is we no longer default to just opening a job.
We’re now really using these insights cuz they’re at our fingertips. And to me that, that’s pretty exciting because it’s creating very new behaviors out of something that, you know, in theory was possible before, but was just very inaccessible and hard to do and therefore, right. It didn’t just pay people time, people just didn’t
William Tincup: do it right.
And people are busy and, uh, they don’t have enough time in the day. And this. [00:24:00] Gonna be able to give them some of their time back because it’s just an easier interface.
Sultan Saidov: Yeah, exactly. And the same thing on the employee side. You know, you, we were talking about, you know, what, what difference does making a small fund make?
But it can be the difference between you pursuing learning and not, it’s not just a case of, um, it, it being slightly better. It’s a case of. You know, I guess if, if you live right next to a place you probably visited, if you live really far away, you might never go even though, um, it’s, uh, it’s possible. And I think there’s a similar, you know, level of friction where after a certain point people just start adopting things they’ve never done before.
William Tincup: right. Last question, um, at the, you know, we’re, we’re one quarter in at the end of the year, what’s success for you with talent, G P T? What does it look.
Sultan Saidov: Honestly, our success is our customer success, right? We, for us, that looks like really two main things. One is a lot of organizations [00:25:00] are going through, you know, pretty chaotic time when it comes to talent over the last couple of years, right?
There’s every function’s been an industry’s been impacted differently, but the common theme is people need. Become more agile and be able to react to things faster. But people also need to take care of their people. You know, the idea of, uh, CEOs saying people are our greatest asset. People have talked about it for 30, 40 years, but in the last couple of years, You know, companies are really leaning in and suddenly, if you have the technologies to do this, you really can make people center stage and really invest in development.
Um, and for us it’s a really exciting opportunity cuz it’s a marriage of, we have this technology that can make a huge difference in people’s experiences as candidates, as employees, as recruiters, as managers. And now you have, you know, a lot of organizations where, The theme of wanting to lean into this is, um, is more pertinent than ever.
So for us, it’s really the helping our, um, customers meet their people outcomes and, and drive, you know, adoption of [00:26:00] these better experiences, um, more seamlessly
William Tincup: jobs. Mike walks off stage. Sultan. Thank you so much for carving out. Tom. I know you’re super busy, but I really appreciate you coming on the podcast.
Thanks so much.
Sultan Saidov: Really enjoyed the conversation.
William Tincup: Absolutely, and thanks for everyone listening to the Use Case podcast. Until next
William is the President & Editor-at-Large of RecruitingDaily. At the intersection of HR and technology, he’s a writer, speaker, advisor, consultant, investor, storyteller & teacher. He's been writing about HR and Recruiting related issues for longer than he cares to disclose. William serves on the Board of Advisors / Board of Directors for 20+ HR technology startups. William is a graduate of the University of Alabama at Birmingham with a BA in Art History. He also earned an MA in American Indian Studies from the University of Arizona and an MBA from Case Western Reserve University.