Shadow AI and Other Things to Consider, with isolved’s Pragya Gupta

[00:00:00] Welcome to PeopleTech, the podcast of WorkforceAI.news. I'm Mark Pfeffer.

[00:00:16] Today I'm joined by Pragya Gupta, the Chief Product and Technology Officer at isolveds.

[00:00:22] They've been leveraging AI for some time, so we're going to talk about the state of AI today,

[00:00:28] shadow AI, AI's role in developing HR, and the expectations around implementation and adoption.

[00:00:36] All that and more on this edition of PeopleTech.

[00:00:41] Pragya, welcome. It's good to see you again, and thanks for being here.

[00:00:45] Always good to see you too, Mark. Thank you for having me.

[00:00:48] Sure. So I wanted to talk about the state of AI.

[00:00:54] It's been a couple of years now since ChatGPT showed up and sort of got everybody to pay attention.

[00:01:03] And solutions providers have been working pretty hard to include AI features.

[00:01:08] I'm wondering, you know, what progress have you seen since, say, 2023 with the capabilities of AI and also the adoption by customers?

[00:01:21] Yeah, that's a great question, Mark.

[00:01:24] Mark, as you stated, it was a couple of years back when with the advent of ChatGPT and particularly LLM models,

[00:01:33] AI became far more pervasive than it used to be.

[00:01:37] AI has been around for a while, but it wasn't so pervasive, especially in the SMB and mid-market spaces.

[00:01:45] And now it has become widely available.

[00:01:48] In fact, you know, one of our, we did a study and according to our HR leaders report,

[00:01:54] 47% of HR leaders believe that AI, it can benefit the employee onboarding process, as an example.

[00:02:03] So it is, you know, we are seeing a lot wider adoption.

[00:02:06] We are seeing a lot wider curiosity.

[00:02:10] You know, HR leaders, when we speak with customers, when we speak with prospects,

[00:02:15] that's one of the biggest questions on their mind.

[00:02:17] How can AI change the way I do business?

[00:02:21] How can it make me more efficient?

[00:02:24] In my view, business leaders should consider solutions like, you know, HR chatbots,

[00:02:29] which can give, for instance, HR teams time back in their day-to-day so they can be more strategic.

[00:02:37] They can use things like predictive analytics, which can drive business decisions.

[00:02:41] We have launched, and I know many, many vendors in this space are thinking about job description

[00:02:49] generation using AI, candidate job matching to streamline the recruiting efforts.

[00:02:57] Another example that comes to mind is payroll anomaly detection to ensure accurate payroll every time, right?

[00:03:04] So there are so many solutions that are out there, and HR users, practitioners should be thinking about them, if not already.

[00:03:14] And what are you finding adoption is like?

[00:03:18] Like, are most of your customers using it or expressing a desire to use your AI solutions,

[00:03:25] or is it sort of an active minority?

[00:03:29] What we are seeing is that they're very curious, right?

[00:03:35] As I said, they're asking us questions, because what is happening is internally,

[00:03:40] most organizations are having discussions about how can we become better with the use of AI.

[00:03:47] Some organizations are asking themselves questions, should we maybe have less people because we have AI?

[00:03:54] So it's not about reducing your workforces, but making your workforce do more with less,

[00:04:00] or become more strategic, focus on things that are driving the business forward,

[00:04:06] rather than just the very minutiae tactical work.

[00:04:10] And which is why HR leaders are asking us the question,

[00:04:14] what can we do to become more efficient, more process driven?

[00:04:21] And in AI platforms, they can drive employee education, training, benefits.

[00:04:27] And with the advent of technology, all these tools and technologies,

[00:04:34] they are becoming more and more, they're becoming better.

[00:04:39] And also the models are learning from the plethora of data available to them.

[00:04:44] So we are seeing a lot of inquiry.

[00:04:48] The other thing I would say is that, you know, for small and mid-sized businesses,

[00:04:53] we are seeing that AI, it's really driven by internal experts, right?

[00:04:59] There's like a pocket of individuals who are more curious than others.

[00:05:04] So they are asking the question and they're propelling the organization forward.

[00:05:07] Whereas in large enterprises, it's also about scale.

[00:05:11] How can we build the scale to be able to grow more?

[00:05:14] So it is a little bit nuanced, if you will,

[00:05:17] but there is a lot of anticipation, interest and discovery that is going on.

[00:05:24] Are you finding that the employers and their employees are aligned on the use of AR?

[00:05:31] I've been reading a lot lately about shadow AI,

[00:05:34] where people are basically bringing their own AI solution to their jobs.

[00:05:39] And it seems like there is a set of employees who are just very, very enthused

[00:05:45] about the tool and may have sort of passed by their company or moved ahead of their company.

[00:05:51] What's your experience been?

[00:05:54] That's a very, very interesting insight.

[00:05:57] And actually, I 100% agree with you.

[00:06:00] You know, there are, as humans,

[00:06:03] some of us are more curious or, as you said, enthused than others.

[00:06:10] So there are a lot of individuals who are coming up with smart ways to explore and use AI.

[00:06:17] Even with ChatGPT as an example, as you start using it, you understand that your prompts drive so much of the output.

[00:06:27] Right. So the better prompts you use, the better output you get out of it or outcomes you get out of it.

[00:06:32] So definitely there are, you know, cheer.

[00:06:36] There are people in the organization who are ahead of others in their use of AI and exploration and discovery of AI.

[00:06:47] And in my view, you know, organizations need to embrace those individuals and ask them to be the torchbearers for their departments.

[00:06:58] In the beginning, there was like about a year ago, I think there was a lot of, you know, some people were scared.

[00:07:06] What does this mean for me?

[00:07:07] What does this mean for my job?

[00:07:09] But with time, especially with those, some of the leaders who are, you know, who want to explore more,

[00:07:17] those are essentially paving the way for others in the organization.

[00:07:20] And we are seeing a lot of that.

[00:07:23] Have you ever been to a webinar where the topic was great, but there wasn't enough time to ask questions or have a dialogue to learn more?

[00:07:29] Well, welcome to HR and Payroll 2.0, the podcast where those post-webinar questions become episodes.

[00:07:34] We feature HR practitioners, leaders, and founders of HR, payroll, and workplace innovation and transformation,

[00:07:41] sharing their insights and lessons learned from the trenches.

[00:07:43] We dig in to share the knowledge and tips that can help modern HR and payroll leaders navigate the challenges and opportunities ahead.

[00:07:50] So join us for highly authentic, unscripted conversations, and let's learn together.

[00:07:55] Now, you've been, both you personally, but also iSol has been immersed in this pretty much from the beginning.

[00:08:04] So what have you learned in the last year or so about AI and its use by HR?

[00:08:11] Yeah, definitely.

[00:08:13] So according to recent research and particularly from the HR professional survey,

[00:08:20] you know, 37% of employees, they still feel threatened in some ways that AI can replace their job.

[00:08:29] And what we are seeing, what we are seeing from the survey is that on an average, AI is handling 34% of HR-related tasks in the organizations we surveyed.

[00:08:42] So 34% is a big number if you think about it, because this means that that time is now being freed to go do more strategic things.

[00:08:52] And, you know, just HR, I'm so happy to see that HR is driving a lot of, you know, embracing of AI.

[00:09:04] And what we've seen in our survey is that 77% of HR professionals believe that AI training is critical to enhancing their roles.

[00:09:14] So, you know, we've gone from people being threatened, but ultimately like 77% are saying that we want to learn, give us more.

[00:09:24] How can AI make us better?

[00:09:26] And also that's a really good stat that happy to see HR professionals jumping on the bandwagon, so to speak.

[00:09:35] And HR professionals that are using AI, they are saying that automating those routine tasks, you know, I talked about HR chat bot as an example.

[00:09:46] Like imagine how many times HR is just being pinged day in, day out.

[00:09:51] What is my PTO balance?

[00:09:53] What is the tuition reimbursement policy?

[00:09:56] What is the travel policy?

[00:09:57] These are questions that can be answered by bots as an example.

[00:10:00] So now what we are seeing from these professionals is that 81% of these professionals that are using AI,

[00:10:08] they are saying that automating these routine tasks has allowed them to focus on building more meaningful connections,

[00:10:15] more strategic discussions and discovery with the employees in the organization,

[00:10:21] rather than just, you know, that answer, go back and forth answering sessions.

[00:10:28] It strikes me that implementing AI is kind of a nuanced thing from the human point of view, from the user point of view.

[00:10:36] There's been some research done that shows employees look at their employer as the source of training, basically.

[00:10:45] They're relying on their employer to train them in its use.

[00:10:48] Do you find that employers actually get that and they're supporting their employees when they implement an AI platform?

[00:10:59] I do.

[00:11:00] I think, you know, it comes down to if you, you know, if we as organizations, we don't train our employees, there is a skill gap, right?

[00:11:10] And it is always going to be harder to, you know, to get, we could just say, okay, let's go outside and get more, get employees who know AI.

[00:11:23] But then you lose all the organization context, all the growth that has happened in the employees.

[00:11:28] So it is better for organizations to train their existing population, whether that is sending them to relevant courses or, you know, helping them enroll in learning programs, which is what we are seeing quite a bit of.

[00:11:44] It is better for organizations to do that.

[00:11:46] Invest in their employees, let them get to the point where they are not scared of AI anymore.

[00:11:52] AI is becoming an asset.

[00:11:53] And what you could do, what five individuals could do in 10 days, now they can probably do that in two days, which doesn't mean that, you know, you don't need as many employees.

[00:12:03] What that means is that they can go do more.

[00:12:05] So organizations have this ability to propel themselves forward.

[00:12:10] And that is where the training is extremely important.

[00:12:14] And building those relevant curriculum, sending them to those relevant programs is critical for organizations to embrace.

[00:12:24] So where do you think this is all heading?

[00:12:27] Where do you think AI is going to be especially useful to HR?

[00:12:33] We all talk about automation and such and efficiency right now, but are there other things that it can bring to the table?

[00:12:44] So I think most definitely automation efficiency, but also the personalization of experience is something AI can do.

[00:12:57] Another thing that is very relevant is, for instance, think of in the benefit space.

[00:13:02] For as long as we've been in this space, even when we have to enroll in our own benefits, you know, should I use, should I choose HMO?

[00:13:12] Should I choose PPO?

[00:13:13] What is a hospital indemnity plan?

[00:13:15] Is this the right plan?

[00:13:17] What should I use HSA?

[00:13:18] How much should I contribute to HSA?

[00:13:20] Did I select the right plan or not?

[00:13:21] These are questions that we as individuals ask ourselves every day, every time we have to do open enrollment.

[00:13:29] Now, as much as, you know, we've talked about efficiency, we've talked about automation.

[00:13:33] If we can give individuals the right tools, you know, AI recommended assessment of benefits or, you know, AI enabled recommendations when we are choosing our benefits plan, as an example.

[00:13:47] So that now is taking it to a whole different level.

[00:13:51] It's not only about productivity.

[00:13:52] It's not only about automation.

[00:13:54] It's about personalization.

[00:13:55] And it's about in helping AI make us make better choices.

[00:14:00] Or, for instance, you know, can we creating models that can be trained on data, on plethora, on a lot of data.

[00:14:10] For instance, someone who is in my demographic, you know, two children, you know, maybe more doctor visits.

[00:14:21] This type of plan is better for you.

[00:14:23] Now, that is powerful information that helps me make a good decision, good health decision, good financial decision for my family.

[00:14:31] So that ties into financial wellness.

[00:14:33] Another example we touched on earlier was payroll efficiency or, you know, making ensuring that payroll is correct.

[00:14:46] You know, whenever we do a survey of employees, what is their number one reason for leaving?

[00:14:51] Payroll is actually still in this day and age a big reason for why employees leave the organization.

[00:14:58] It could be incorrect payroll or, you know, not feeling adequately compensated.

[00:15:03] There's a lot of factors that come into play.

[00:15:05] But, you know, whenever I talk to payroll admins, all of us have at some point dealt with the stress of a payroll Thursday, right?

[00:15:17] You know, processing or, you know, there were changes.

[00:15:22] You know, some employees was doing part-time hours.

[00:15:25] They've been there now doing full-time hours.

[00:15:27] But we have to make sure that the HR classification has been done so we don't have compliance issues or we don't have financial implications.

[00:15:35] So think of payroll assist using AI as a smart assistant in your car, right?

[00:15:43] Driver's assistant in your smart car.

[00:15:45] You clearly know how to drive.

[00:15:47] But when you are barreling down that interstate and your car goes left or right or there is a discrepancy in your driving pattern, the car tells you, hey, you're not driving well.

[00:15:58] Go take a break.

[00:15:59] Similarly, you know, we're working on this model.

[00:16:04] It's called perfect payroll.

[00:16:06] Essentially, it's enabling the payroll admin to make the right decisions ahead of processing payroll, not after.

[00:16:14] Not after payroll has been processed and then the employee says, hey, my hours were incorrect or I was unclassified from a part-time to a full-time or whatever it is.

[00:16:23] You can proactively take care of these anomalies.

[00:16:26] You can proactively address them.

[00:16:27] So, I mean, it is the payroll example, the one I just gave.

[00:16:32] It's still a little bit of productivity and efficiency.

[00:16:34] But now it's not only productivity and efficiency.

[00:16:37] It also is experience because now my receiver doesn't have to call me and tell me there was a discrepancy.

[00:16:44] You took care of it ahead of time.

[00:16:48] Now, AI is obviously a big part of the vendor's messaging right now.

[00:16:54] And I'm curious about the customer's expectations when they start the discussion with you.

[00:17:01] Are they believing all the hype that's out there and setting goals that aren't realistic?

[00:17:08] Or are they taking a more pragmatic approach?

[00:17:12] What's it like out there?

[00:17:14] That's a great question because you are absolutely right.

[00:17:19] When we talk to prospects, when we talk to customers, they ask, you know, as much as they are happy to hear about productivity, efficiency and experiences,

[00:17:30] they are also very much focused on data privacy, ethical practices.

[00:17:37] They are asking that how are you making sure that your models are not intrusive?

[00:17:42] So they are thinking about these things.

[00:17:46] And, you know, like, are we going from just prediction and guidance to oversight and control?

[00:17:55] What does the ease of use of these AI tools look like?

[00:17:58] So customers are focusing on not only what the efficiency is, but they are asking us the right questions, especially around biases, misalignment with company values, intrusion, human oversight.

[00:18:15] They are asking us questions about them and rightfully so.

[00:18:18] And are they satisfied, do you think, with where we are right now?

[00:18:22] Yeah.

[00:18:23] So how I think about it, let's take one example.

[00:18:26] So let's maybe take example of ethical practices in AI, right?

[00:18:33] When we build an AI model that is, let's maybe take example of candidate matching, right?

[00:18:43] So when we match the skills of the candidate with the skills of the job.

[00:18:47] So think of when you post a job opening, right, job posting, you get a thousand resumes.

[00:18:55] Now, instead of manually going through each of these resumes, we have built an AI-based candidate matching model, which will say that these top 10 candidates are the best fit for the job based on their skills, based on their experiences, as an example.

[00:19:18] Now, what a customer or a prospect wants to make sure is that there is no bias in this model.

[00:19:28] There's no bias based on gender.

[00:19:30] There's no bias based on age.

[00:19:32] There's no bias based on race.

[00:19:33] That's what they want to make sure, right?

[00:19:35] So as the developer of these models, what we do is two things.

[00:19:42] First, whenever we are building this model, we strip the data exchange of any of these personally identifiable information.

[00:19:52] So then when we are training our models, our models are not trained with anything personally identifiable.

[00:19:58] So then the models are skills and experience based only, as an example.

[00:20:03] That's one thing.

[00:20:04] The second thing that we have to do is there are AI and ethical consortiums that, you know, do audits.

[00:20:14] They, we participate in, in, in forums that make sure that our, our models are continuously being audited.

[00:20:25] Our models are meeting the best practices that, that have been set from these, these forums.

[00:20:33] So, so that's something that every vendor should be thinking about.

[00:20:38] Every vendor should be proactively self-screening themselves.

[00:20:42] Well, Pragya, thanks very much for, for talking with me.

[00:20:46] It's always great to talk with you and I appreciate your time.

[00:20:50] Wonderful.

[00:20:51] It's always good to speak with you, Mark, as well.

[00:20:53] Have a great day.

[00:20:54] You too.

[00:21:05] My guest today has been Pragya Gupta, the chief product and technology officer at ISO.

[00:21:11] And this has been PeopleTech, the podcast of workforceai.news.

[00:21:16] We're a part of the Work Defined Podcast Network.

[00:21:18] Find them at www.wrkdefined.com.

[00:21:26] And to keep up with AI technology and HR, subscribe to Workforce AI today.

[00:21:32] We're the most trusted source of news in the HR tech industry.

[00:21:36] Find us at www.workforceai.news.

[00:21:41] I'm Mark Theffer.

[00:21:42] Thank you.