It’s been nearly 18 months since ChatGPT launched and sparked a lively debate about the future of Talent Acquisition. While it’s important to consider the long-term implications, a crucial part of building a strategy is embracing the potential of current technology through experimentation and skill development.
My guest this week is Andy Headworth, Deputy Director of Talent Acquisition at HMRC. Over the last few months, Andy’s team has gone from lunchtime AI experimentation sessions to rolling out the platform they created across their organization. This has driven radical improvements to their hiring process, improving its quality while simultaneously saving many hours of hiring manager and recruiter time. Andy shares its impact, the lessons they have learned, and his advice for TA leaders on building strategies for an AI-driven future.
In the interview, we discuss:
• The recruiting challenges at HMRC
• Experimentation, learning, and being brave with your thinking
• Understanding the problem you are trying to solve
• Revolutionizing the hiring manager experience
• Building platforms using publically available tools and data
• Implications for TA and HR software vendors
• The use of AI by candidates
• Advice to TA leaders
• What does the future look like?
Follow this podcast on Apple Podcasts.
Matt: Support for this podcast comes from CV Wallet and it’s a verification platform for recruiters. In a world of jobseeker adoption AI and too many unsuitable applications, CV Wallet puts trust at the heart of your hiring by verifying your applicants at the start of the recruitment process, saving recruiters time and money by getting straight to the right candidates for your jobs. Its bluetick verifications are very easy to set up. Work globally are completely bespoke and can be set at an organizational or job level depending on recruiter requirements. Typical verification includes identity checks, right to work, qualifications, licenses, and skills. Enabling jobseekers to fast track their applications while securely saving their verifications for future use in their free CV Wallet app. It’s free to sign up and with no integrations required, fits seamlessly within your existing application process. Get started today at cvwallet.com.
[Recruiting Future theme]
Matt: Hi there, welcome to Episode 607 of Recruiting Future with me, Matt Alder. It’s now nearly 18 months since ChatGPT launched and sparked a lively debate about the future of talent acquisition. While it’s important to consider the long-term implications, a crucial part of building a strategy is to embrace the potential of the current technology through experimentation and skills development.
My guest this week is Andy Headworth, Deputy Director of Talent Acquisition at HMRC. Over the last few months, Andy’s team has gone from lunchtime AI experimentation sessions to rolling out the platform they created across their organization. This has driven radical improvements to their hiring process, improving its quality while at the same time saving many hours of hiring manager and recruiter time. Andy shares its impact, the lessons they have learnt, and his advice for TA leaders on building strategies for an AI-driven future.
Hi Andy, and welcome to the podcast.
Andy: Hi Matt, it’s good to be back. I think it was number two or three many, many years ago. I was on it last time.
Matt: It was and I think you win hands down win gap between first podcast appearance [Andy laughs] and second podcast appearance because it must be eight or nine years. But it is brilliant to have you back on the show. For people who may not have come across your work before, can you introduce yourself and tell us what you do?
Andy: Yeah, absolutely. Andy Headworth, those of you who have not come across me for I’ve been in the talent world for many, many years. For the last eight years I’ve been working at His Majesty’s, we said the wrong one there– His Majesty’s Revenue and Customs, known as HMRC the dreaded taxman. I head up talent acquisition and recruitment. I’ve been doing that for the last eight years.
Matt: Fantastic stuff. So, give us a bit of context here. What challenges does your organisation have when it comes to talent acquisition?
Andy: We have quite a cross section, as you might imagine. So as an organization, we’re a 65,000 strong, primarily UK based, although we do have operations outside that size organisation. We have a sizable contact centre, number of people that are frontline. So that’s one pool where we recruit heavily every year. And then across the organization, we have a huge range of skills that we recruit for which people don’t, they just don’t know. I mean, everyone thinks we just deal with tax, but to deal with tax we need everything, from IT to policy to legal, to finance and everything else in between. And so, our challenge is to make sure as like every other large employee, we’ve got the right skills in the right place at the right time.
It sounds so corny, but it’s just so simply true. And then within my team, we look after– So, we then break that down further. From a recruiting point of view, we have hiring manager led recruitment. If the hiring managers can’t manage it themselves, i.e., they can’t find the people through sticking AdWords, then that’s where my teams step in and we are all the proactive teams. So, we look after specialist proactive recruitment, we do the exec recruitment, we do contingent labour and we do the graduate recruitment within my team. My area is wide and we have all the challenges that come with that.
Matt: I saw you do your presentation a few weeks ago and it was fascinating by the amount of cyber security talent that you need to do the work.
Andy: We are one of the most attempted hacked organizations out there, understandably, because of who we are and because of the financial aspects. So, we have a very large department of security people from all aspects, which is just an ongoing challenge. And anyone who’s in the marketplace for recruiting cyber people knows how hard that is to bring people in because they’re in massive demand.
Matt: So, I know that you’ve had a big focus on using generative AI to improve your processes and ultimately solve some of the recruiting challenges that you have. Give us a bit of an overview about how you’re using it, but also how that came about and how you mobilized people to work together to build the system that you currently have.
Andy: Yeah, absolutely. It started with the good old ChatGPT launching in November 22, was it 23? whenever it was now and I mean, you know, me being as an early innovator of things straight away, the following year when it launched, a couple of us could see the benefit of how AI could be used in talent acquisition. And so, we got super excited about it. And for three or four months, we’re looking at how we could use it in different ways. And it very quickly evolved in those three or four months if people remember. And then by about the summer or June of that year, it’s like, okay, we figured it out now. I figured out how we could use some of this in HMRC and how we could use it in talent acquisition.
So, a few of us started to have lunch and learn sessions. And this is how it all started, to try and get the buy in from people that really, at that stage, still didn’t even know what ChatGPT was. So, we started just doing an hour, every hour, every Thursday, come along, let’s look at the new tools, let’s look at what’s going on, let’s play with it, let’s test it. At that stage, the tool count was growing every day by about 20 or 30 tools a day. So, we had plenty to go through. And then we had about 15-16 of my team that would come along every Thursday. And we had a team’s channel, we were very active in learning.
And then the penny dropped with me was when, this is just my team, was when DALL-E 3 came as part of ChatGPT and Microsoft DALL-E 3 went live. We changed having competitions in the team. So, we had picture creation competitions and did strange things like that and other things and marketing campaigns. And suddenly the interest from my team doubled literally overnight. They wanted to get involved. It was competitive, as you can imagine, with recruiters, but also it was something different, the ability to play with it. And they had the license to play with it in a certain way. And what was fantastic was they were bringing people in and they were challenging each other with the prompts because they couldn’t, we obviously had difficult tasks to do including, I was doing them as well because they were so impossible, they were learning and relearning prompts and doing different things. It was fantastic.
And then we sort of cemented that. And then that carried on. In the meantime, in early June, I started then doing presentations across CPO and other business groups about what we were doing. And then I carried on doing that across the organization and then outside the organization. And then in September last year, we broke down the recruitment process to a number of different sections. We all know what they are in Talent Acquisition. And we looked at half a dozen sections in recruitment and went, right, okay, what can we do with AI and build to make the life of that section of that, whatever it may be, easier, more manageable, time saving, more efficient. We chose two to start with, one, which was the hiring manager. Come back to that in a second.
The second one was looking at insights around locations. So, talent acquisition insights, we’ve got 18 locations that we are focused on. So, I wanted to know exactly what’s going on within an hour’s travel of those locations. But coming back to the hiring manager one for a minute, so we identified these two that we could build ourselves. Now, this was at the time when the GPTs came out within the paid version of ChatGPT. So, we were able to quickly build a GPT. So, the hiring manager one is very simple. Hiring managers have no time. Therefore, they don’t give any time to recruitment in most organizations. They also have drawerfuls of old job descriptions, which, shall we say, some of them are not legal, some of them are way old, and they don’t have the right current technologies on.
And so, it’s like, how can we make this process easy? So, we basically built a tool where you’re hiring manager Matt. You put your job title in. It then asks you for your grade, and then it starts working. And it comes back immediately with six other job titles that you might want to consider and you haven’t thought of. Then it gives you a full job description of the job you’re going for. Then it gives you a job AdWord for a job board, and then an AdWord for a LinkedIn post and a social media post. Then it starts to look at interview questions, further interview questions, if you need those. We use something called success profiles in the civil service, which is a behavioral tool, and that’s already pre-programmed in there.
And so, it then examines the job description that you’ve got, and then it’s telling you, it recommends the behaviors that you should then plan for in the interview, and then provide you with the questions for those. It provides you with outreach for social media. So, if you want to engage candidates you found on LinkedIn, for example, it gives you the text to do that, gives you some hashtags, and then you can take all of that into a Google Doc. So, it produces all that itself, then opens up a Google Doc. You can use AI, then to customize any aspect of that. You might want to improve some parts, you might want to add some stuff, change it.
And then when that’s all done, you just download and you’ve got your full PDF with your full end to end with a golden thread from start to finish. And all a hiring manager has to do is go to his diary. Well, sorry, go to their diary book, some time to do the sifting and then some time to do the interviewing. And everything else is taken care of. And we just say we are saving so much time and helping hiring managers with that. And that’s what I think AI does. It doesn’t take it away from the fact that the hiring manager still has to do the recruitment, but we take away all the bits around it. As you saw it, Matt, it’s just so lovely. It’s just a piece of art.
Matt: Yeah, no, it’s pretty impressive. And to clarify, your team built this using the publicly available tools, didn’t you?
Andy: Yes, there was all publicly available data, so weren’t infringing on any data rules within HMRC. We started it on GPTs, but then we quickly realized there were limitations because people could only look at it if they were paying their $20 a month. So, that was a no, no. So, we use MindStudio as a product, which is a drag and drop tool where you can load your large language model above. And we use ChatGPT Claude, and I think it was barred at the time, I can’t remember which were the generative AI above it. And the people that built this are just members of my team in my recruitment function. They are not developers, they’re not technical, they’re just people with a passion.
And we work the problem through and then the tools are simple to use, you just work through them. And we’ve created them. We’ve done, I said another tool now around the insights that’s equally as powerful. And we’ve got two more in the offing. One for selection and probably one for onboarding is the last one we’re looking at the moment same principle. We try to understand what the problem is we’re looking to solve and then use the AI to solve it.
Matt Alder: A quick message from our sponsor, Wonolo.
Hi everyone. I want to tell you about Wonolo, that’s w-o-n-o-l-o. Wonolo stands for work now locally. Wonolo enables businesses to find quality workers for on-demand, seasonal, short-term, and long-term work. Ditch the bulky paperwork and interview process and use Wonolo to find quality workers fast and get work done even faster. With flexible workers and no platform fees, you can save on operating costs, meet demand, and maximize earnings with ease. Wonolo is available in over 100 markets, including Chicago, Dallas, Atlanta, New York and Seattle. Get workers who are ready to work and spend less time finding them with Wonolo. Go to www.wonolo.com/pod, that’s www.wonolo.com/pod and take the stress out of finding workers.
You mentioned that it saved you huge amounts of time in terms of supporting hiring managers. So, is it a tool that’s very much augmenting what you’re doing? Is it adding additional skills to your team? Is it replacing work that other people would do? How does it fit in from a resource perspective?
Andy: So, two sides to that, the tool itself is for the hiring managers. Now, we can work with the hiring manager to help them do that, or we give the tool to them and they start using it and crack on with it. So, at the moment we’re still in test mode. So, I don’t have the data of the next question you’re going to ask me is how many hours have you saved? We can’t give you that just yet, but we will do in the near future. So, that’s one size. The tool becomes a standalone or my recruitment team can use it and work with a hiring manager.
In terms of the skill side of things, what we’re using, so that’s slightly different, is that the AI is allowing, it’s taking some of the mundane tasks, but allowing some of the recruiters and the teams around them to use the AI to enhance their skills and knowledge. So, if they do have skills gaps, we’re using AI to plug those skills gaps. You talked about skills-based hiring earlier and the trend at the moment towards people talking about what skills people have. What we’re finding is if you teach people how to use generative AI, and then you teach them how to prompt, and you teach them how to prompt properly and with the different ways of prompting, it can supplement skills they don’t have and make them much more an effective person. And that’s what’s really interesting.
So, when we have skill shortages, do we actually need to have all those skills would be my question? And could we not augment those with your 20%, 30% of AI to supplement what that person’s core skills are? Because they don’t need to know some of this stuff. We don’t need now to have people who are excellent on excel anymore. Why would you do that when you’ve got a tool that can write any formula in the world, you don’t need to spend three hours studying excel in the evenings and going crazy. That’s the one that makes me more excited, the fact I don’t ever have to use a formula ever again. I can ask excel to do it for me, but that’s an example.
And I think within TA at the moment, we’ve got such a focus on skills, but we’re maybe missing a trick. Can we not think about augmenting them? You and I talked about this at the CIPD conference. There is still a fear from organizations that are reticent to want to dive into this. It’s as if they’re waiting for the global permission to say go and then go play. Don’t wait. Just get in there and do it. If you’re using public data, which we use all our publicly available data, we’re not giving anything away. Just see how it can work within your process, which is what we’re doing. We’re testing the waters and pushing as far as we can.
Matt: Yeah, I think that’s really interesting because there’s definitely calls that people working in TA functions need to become much more specialized in certain areas. But I think it’s really interesting in terms of looking at that from an AI perspective and proactively saying, “Well, actually, do you know what AI is going to do this bit so we don’t have to have that kind of skill?” I think that’s really interesting.
Andy: I just think people need to be braver with their thinking math to be honest, I think they stop waiting for things to happen and go test it. Go play with it. I think, people, if you go past the superficial, make me a recipe of what’s in my cupboard type prompting, which is fine, and other things, and start getting into the deep, detailed, multiple linked prompting that you can do now. The power of generative AI is super scary, but also, it’s just so brilliant from our perspective because it can do so many things, but it’s bravery and persistence and a bit of learning. And once you’ve learned how to embrace it, you’ll find yourself using it for everything. I’m sure you do. I do.
Matt: Yeah. No, I mean, it’s revolutionary already and you sense that there is so much more to come. So, to ask you a question about the market for TA technology, because here’s a system that you have effectively built yourself. You haven’t needed an outside vendor to provide any expertise or any software to do this other than the low-cost tools that you’ve been using. Every single vendor in the marketplace at the moment is launching AI into their technology. How do you navigate that market when you’re looking externally for products and suppliers? What are you looking for? And where’s that balance between what you can build yourselves and what you need to bring in from an external provider?
Andy: Yeah, I think it depends on the size of your company. Obviously, we’re enterprise level for us and I think for us within TA, we’re building our own tools to roll that out enterprise wide across all of HMRC and have potentially tens of thousands of users. We would probably need to think about the scalability of an external provider to help us. I think what is out there, and Peter Gold and I have had this conversation last week, funnily enough, exact same conversation. There’s a lot of smoke and mirrors. AI is the buzzword. So, everyone’s putting AI in their sales literature. Does it actually do anything with AI? Probably not, but they’re using it to sell it and get them in the door. And when they get them in the door, it’s like, oh, it’s on our roadmap to integrate this.
So, there are very few that are actually, I think, doing stuff for real. I think the easiest thing from our perspective, and I had a demo this morning of a product, didn’t quite pitch me as AI, but it was in the conversation. But when you get to the nub of it, oh, no, we don’t use AI for that. We don’t do that. Well, that’s what you told me. And I think really you can get to the point very, very quickly with some vendors and catch them out, because I do think some are playing the game, but there are some ones that are genuine. So, to your point around, where’s the market with AI? In some respects, simple is better at the moment. There are some fundamental things we could do with fixing. One of them is the selection and sifting process.
So, what happens as a hiring manager? If I get 20,000 applications, what do I do with them? Currently, we would have to go through them one by one. So, that’s one of the problems we’re looking to solve. That’s a massive problem in recruitment, especially as candidates are getting onto AI now as well. So, they’re having CVs written for them and they’re automatically sending them to every job.
Matt: Yeah, no, that was going to be my next point. I think that is potentially forcing people to embrace this, even if they don’t want to. I mean, I’ve had a number of conversations with graduate recruiters in the last two or three months and they’re all experiencing record levels of applications being driven by the use of AI.
Andy: Yeah, we had our grad campaign this year. We average usually about 13,000, 14,000 applications a year for our tax grads. We had 22,000 this year. That’s a big difference to manage. So, I think this is an industry problem we’re going to have, but I don’t think, and that’s what I’m not seeing with all these AI tools. They’re all trying to– It’ll be interesting to see in a year’s time how many of these tools that we see on Matt Wolfe’s channel, for example, are even existing. It’s the same thing in the social world when it happened. They just come out of anywhere and build everything. But are they actually building tools that we want or building tools that they think they can make?
And I think that’s the other thing as well, is they’ll make a tool that they think is useful, don’t necessarily engage with the industry they’re selling it for, and then maybe try and wedge it in. But actually, if they come and ask us what the tools we want to be solved, then you can write a job descriptor, you can use it now. As an organization, if you are not using AI, then you should be testing it.
You should be looking at Claude Gemini and ChatGPT and looking at how you can better construct your job AdWords, how you can write job descriptions, how you can do other things, because you’ve got to think however many millions of people around the world are using AI at any one particular moment and how much data that is collecting on different job descriptions, different job titles, different technologies. It’s such a great tool to look at how you can construct the main facets of what we do in talent acquisition.
Matt: No, absolutely and next question. You’ve covered some of this already, but I just want to pull it together into one summary, if you like, which is, what advice would you give to the TA leaders out there about building a strategy for AI?
Andy: Well, first of all, be brave. Don’t wait for other people. Don’t sit there and wait for your competitors to do it before you. And everyone’s saying this AI is here now. It’s not in 10 years’ time, not even next year. If your candidates are using it, you need to use it. So, find a way of starting to take it into your teams and just use the basic tools that are out there and look at what you’re producing and start doing some comparisons. We did the same thing. We just took a whole lot of job descriptions and compared them with the AI versions that we had created and they were light years ahead, the AI ones, in terms of up-to-date current, latest skills, latest tech in there, the right wording to attract people.
And I think that’s the first thing. Be brave. Just dive in and start using it. Be careful on the data. Don’t use private data. You don’t have to use, use the AdWords that you’ve pasted on your career site, for example, so there’s nothing being given away and you’ll start to see potential. If you have a sourcing team, you should be using AI to be building your Boolean strings. It’s fantastic at saving the time of your sources to build the search strings, for example, on any platform, whether it’s LinkedIn or other platforms. So, there’s some really quick wins you can do in TA without having to get into some of the, should we say, the ethical HR challenges of personal data in AI.
Matt: And I suppose that brings me on to my final question. We’ve been talking very much about the here and now and what’s possible, what does the future look like, how is this going to affect TA in the long term? If we’re having this conversation in another eight years’ time, in terms of a gap between interviews, what do you think would happen? And if you can’t think what that might be, what do you hope would happen?
Andy: Well, we’ve been here before, haven’t we? about 10 years ago with social. So, I think it has to improve the technology. The technology has to improve and AI has to come into our ATS platforms, our CRMs, and make that much more usable process and intuitive process. So, I think that’s given. How quickly that happens, I don’t know. Then when it comes to people in an organization, if I think of my talent function or any other talent function in two or three years’ time, by then the likes of copilot and others will be embedded in your process. It should be. And if it’s embedded in your process, embedded in everything. So, in terms of how you’re using it. To me, AI then becomes your basic tool whether it’s writing your emails, replying to your emails, doing presentations, doing your spreadsheets, and allowing the talent of the individuals in talent acquisition, for example, to flourish and use their brains creatively, engaging candidates, doing better, producing better candidate experience and spending the time getting the right people rather than the logistics and the infrastructure around it to facilitate those people coming through your door.
And I think if we can give better recruiters more time with candidates from both sides, that’s a bit utopian because someone will say, “You don’t need to talk to anyone,” but you do. And I think if you want to engage the right people, that’s where I hope it’s going to go.
Matt: Andy, thank you very much for talking to me.
Andy: Absolute pleasure, Matt.
Matt: My thanks to Andy. You can follow this podcast on Apple Podcasts, on Spotify or via your podcasting app of choice. Please also subscribe to our YouTube channel by going to mattalder.tv. You can search all the past episodes @recruitingfuture.com. On that site, you can also subscribe to our monthly newsletter, Recruiting Future Feast and get the inside track about everything that’s coming up on the show. Thanks very much for listening. I’ll be back next time and I hope you’ll join me.
[music]
