Taylor Bradley

▶0:00HostWelcome to the Built by People podcast, where we share the stories and insights of the world's top HR leaders. Join us as we dive deep into the minds of HR executives, uncovering their strategies, challenges, and triumphs in shaping today's workforce. I'm excited to welcome Taylor to the Built By People podcast. Taylor, thank you so much for joining us. And as a starting question, I always love to ask if you could share a little bit more about your career journey.

▶0:50GuestThank you for having me on the podcast. And absolutely, I started my career as a firefighter paramedic, as one does, before moving into HR where I've now spent over a decade primarily supporting tech companies in the late-stage IPO readiness journey. I'm a member of CNBC's Workforce Executive Council and currently the VP of Talent and Strategy Success at Turing. And at Turing, I lead a team of about 40 HR professionals. And one of the unique aspects of my role is the ability to experiment with and design AI use cases for HR in the force. For those not familiar with what Turing does, we effectively have two primary services that we offer. One is called AGI Advancement, where we help the large language models, which some of you may be familiar with, Claude or ChatGPT, become even smarter and do better jobs. The other aspect we have is Turing Intelligence, which is custom engineering where we help AI use cases and proof of concepts become full-blown AI applications and really scale the AI impact that those organizations are using in their workforces.

▶2:10HostTaylor, thanks for that great introduction. I want to start off by having you tell us about a time when you successfully implemented AI to solve a major HR challenge at Turing, particularly around your experience scaling employee onboarding by 100% last year.

▶2:28GuestYes, that was quite a journey. And the— it all started when I had our CEO say late H2 of last year that we would be scaling the organization by several thousand team members. And in order to do that, my initial response jovially, which then turned out to be our motto, was going to be, we're going to AI our way out of this. And one of the use cases that we identified early on was the number of tickets that our support desk was responding to from all of the new hires and all the questions they had amounted to about 52,000 tickets a year that we were responding to manually. And the team, after being trained on how to identify AI use cases, quickly came to me and said, we think an AI could do this or do part of it. And so the proof of concept we came up with was a custom GPT that we built in about 2 days using ChatGPT Enterprise license. And it would take an email from a team member that raised a ticket based on its knowledge base, would write a response, and then our human in the loop, our people operations specialist, would review it to make sure it was accurate, do some adjustments, and then they would send it. Just in 2 days of developing that custom GPT, we reduced our per-ticket processing time by 33%, which if you times that over by 52,000, had a huge productivity gain.

▶4:01HostGreat example. Taylor, how does your team approach balancing AI automation with human empathy and performance reviews? Can you walk us through your framework for keeping humans in the loop?

▶4:13GuestWorkforce engagement is one of the key barriers to AI adoption, not the technology itself. And when it comes to identifying— and there's an infographic that we'll share with your listeners— when it comes to identifying which task can be supported by AI, there's two questions that MIT research would tell us to ask. What can be augmented by AI, meaning you complement the human in the loop and their abilities and their impact? And what can be automated by AI, meaning that you replace the human with automation. Where we have found the most uptake, especially in getting the workforce engaged, is by proving the augmentation use cases like we did that I just described with the use case around the support tickets, is if you can show people how AI can support their work, make them drive more impact so they can focus on the more complex issues, That's where you get that engagement with the workforce. But more importantly, those immediate results.

▶5:16HostTaylor, when was the time you had to overcome skepticism or resistance while implementing AI in HR processes, and how did you navigate that challenge?

▶5:27GuestSo the AI use case that I recently talked about, we ended up calling the use case Alan, which is named after the namesake of the company, Alan Turing, a famous mathematician and computer scientist. However, the concept we developed is we didn't want the use cases to be isolated and independent. With our custom engineering team, we really wanted to turn Allen into an AI team member that would help us in a variety of different fashions. But early on in the implementation of Allen, the team members came to me and said, we don't like the way that Allen is writing responses. It's clearly written by an AI. And so where we educated the team is when those types of issues come up with AI, it's a simple matter of prompting. We just need to change the linguistics prompts that go into Allen's knowledge base so that when it writes responses, it starts to embrace more of a human-type tone. And the way that we did that was getting the team involved. I asked the team, go find the emails where you felt that you just wrote the perfect email, that you really did a good job at putting the human touch on it. We took those samples, we had AI analyze the writing, the tone, the style, and then we incorporated that into Allen's knowledge base. And that immediately resolved this concern that my team had, is that the outputs came across too robotic. So that linguistics prompt was several pages long, and that's where prompting is a key element to what teams need to understand, is they can resolve these types of bugs by simply telling in normal English the AI what to do and how to do it better. And that's how we resolve some of the consternation that we came across.

▶7:16HostCan you share with us the story behind developing your IT support bot from identifying the need to measuring its impact on team efficiency?

▶7:27GuestYes, I'm pleased to say we'll start out with the headline metric is by the end of 2025, We're on track to automate— remember, automate versus augmentation— automate 60% of the 52,000 tickets that we have annually. And we took Allen and we're now scaling it across the organization to make Allen available in Slack, available on the support portal, a variety of different fashions where our team members can interact with Allen. And it goes beyond— we're creating a chatbot to answer policy issues. We've expanded the knowledge base to make it feel more like a team member. Again, we're trying to break down those barriers for AI adoption internally and with our clients. And that knowledge base now includes coaching on how to interact with your manager, how to give feedback in structured ways. Of course, it answers policy issues and garden variety IT issues, but it also encourages team members on how to do better quality work, especially when it comes to using some of our proprietary tools around LLM labeling tasks, which are very code-centric type tasks to do. And we give feedback to our engineers on how to do their jobs better. So Allen is not just a chatbot, it's a team member. And that's been one of the great success stories that we've been experiencing so far this year.

▶8:57HostWhat parting advice would you like to share with our community, Taylor?

▶9:01GuestIt's a great question, and code alone doesn't drive change. It's culture. You need to have the workforce bought into AI adoption in order for it to be successful and relevant to your business. The biggest barrier to AI adoption today is not the technology, it's the inertia of the status quo within the business. And the best way to break that barrier down is by involving your employees.

▶9:32HostTaylor, thanks so much for joining us on the Built by People podcast. It was a pleasure to have you involved.

▶9:37GuestAnytime. Thank you.