The real value of AI agents in HR comes from connecting them to work across the employee lifecycle, not from deploying them on individual tasks. That’s where most large organisations are getting stuck. Working in a fully agentic way means dealing with different systems and different data sources that often have no shared foundation. The result is fragmented experiences for employees and managers, with the end-to-end potential remaining out of reach. Getting there requires some serious work in data governance, process design, and integration, the kind of foundational work that rarely gets mentioned at industry conferences.
So what needs to be in place before AI agents can work at enterprise scale?
My guest this week is Melissa Shelley Höjwall, Global HR Technology Lead at H&M Group. In our conversation, which we recorded live at HR Tech Europe, she explains what it takes to build a connected AI architecture across HR and why many companies are undermining their own progress
In the interview, we discuss:
• The approach to Agentic AI in HR at H&M
• From niche agents to connected architecture
• Process automation design and date integration
• The role of data governance
• Adoption in the enterprise
• Shadow AI and over-governance
• Why cutting jobs isn’t the way to get true value from AI
• New roles for HR professionals
• Breaking the silos in the Talent function
• What to focus on for the future
Follow this podcast on Apple Podcasts.
Follow this podcast on Spotify.
00:00
Matt Alder
We’re still tending to talk about AI agents in very compartmentalised terms when it comes to the tasks they can do. However, the true value of agentic AI will come from connecting agents together to work across the entire employee lifecycle. Most enterprises haven’t got there yet, so what’s holding them back? And what does it take to unlock that potential? Keep listening to find out Support for.
00:26
Matt Alder
This podcast comes from CoderPad. Here’s a question that should keep you up at night when your next engineering hire sits down at their computer on day one, are they the person you interviewed, or are they someone who just knew how to perform for an interview? Meta discovered a dangerous gap between interview performance and actual job performance. So did LinkedIn. So did Yahoo. They all found the same culprit. Their hiring process was testing for skills that no longer reflect how the best engineers actually work. AI hasn’t just changed how software gets built, it’s changed what it means to be good at building software. And the companies who are winning the war for technical talent? They’ve already rewritten their interview playbook. CoderPad is the platform powering that shift from AI enabled live interviews to realistic project based screens.
01:27
Matt Alder
CoderPad helps you see candidates in action, not just on paper. Meta scaled to 16,000 interviews using it Yahoo cut recruiter screening time by over 50%. MNTN is onboarding new hires in days instead of weeks this April. If you’re still banning AI interviews, you’re not just behind, you’re screening out your best Future hires. Visit CoderPad/io.podcast to talk to the team that’s CoderPad/io.podcast.
02:24
Matt Alder
Hi there. Welcome to episode 791 of Recruiting Future with me, Matt Alder. The real value of AI agents in HR comes from connecting them to work across the entire employee lifecycle, not from deploying them on individual tasks. That’s where most large organizations are getting stuck. Working in a fully agentic way means dealing with different systems and different data sources that often have no shared foundation. The result is a fragmented experience for employees and for managers, with the end to end potential remaining out of reach. Getting there requires some serious work in data governance, process design and integration, the kind of foundational work that rarely gets mentioned at industry conferences. So what needs to be in place before AI agents can work at enterprise scale? My guest this week is Melissa Shelley Höjwall, head of Global HR Digitization at H&M Group.
03:23
Matt Alder
In our conversation, which was recorded live at HR Tech Europe, she explains what it takes to build a connected AI architecture across HR and why many companies are undermining their own progress.
03:35
Matt Alder
Hi Melissa.
03:36
Matt Alder
Welcome back to the podcast.
03:38
Melissa Shelley Höjwall
Thank you, Matt. What an honor to be back.
03:39
Matt Alder
Well, it’s a pleasure to have you on the show. Let’s just start by you introducing yourself and telling everyone what you do.
03:45
Melissa Shelley Höjwall
My name is Melissa Shelley Höjwall and I am globally responsible for HR digitalization at H&M. So all of our global HR systems and HR tools are supported by my team.
03:58
Matt Alder
It’s the last day of HR Tech Europe. We’re sitting on a table near the entrance. Well, it’s entrance exit because it is really near the end. And, and it’s been an interesting couple of days. Lots of talk about agents in particular, which is not surprising because it’s all everyone seems to talk about. But what’s been your take on it? Because I think that there’s some very.
04:19
Matt Alder
Top level stuff, but not much around.
04:22
Matt Alder
The detail of how this might work in a kind of enterprise environment. What’s your sort of take on what you’re hearing?
04:27
Melissa Shelley Höjwall
Yeah, no, I agree. There are increasingly more companies that have implemented one or a couple of agents doing usually quite niche things. And I’m going to use recruitment as an example. But also we’re seeing within recruitment a crazy new amount of system vendors that have created AI native niche solutions to support parts of the recruitment process, for example. But I think that when you are working with a company of our size and most skills, even smaller companies, it’s easy to get going on creating this agent or that agent. But I think when you start looking at what you have to do to make this work in the total perspective, if we go back to our customer, the employees, the managers, to create a unified experience, you don’t want to just throw out tons of agents. How are they connected?
05:24
Melissa Shelley Höjwall
How are they creating a cohesive experience? How do I even know now what chatbot or agent to go into? Right. So what I’m really seeing, and this is also being raised from a more theoretical standpoint, even though I know it’s very based in practice from of course, Josh Burshan, one of the great thought leaders, is that we, I think the companies taking the next step in this journey are no longer just thinking, okay, this cool agent, that cool agent, but they are thinking about how do we create a cohesive architecture where this all makes sense. And I’m going to exemplify with the recruitment case in practice.
06:03
Melissa Shelley Höjwall
Now we have like a couple of other great companies that have been here, like Pandora for example, we’re on the main stage talking about similar journeys that we have also been on and how they are using niche players and recruitment to really modernize and improve parts of the end to end flow. Things like volume assessment and interview scheduling support and so on. But also when I asked this very of course successful HRO who’s been doing great things at Pandora, and I asked, okay, so how are you linking that in the end to end flow then to simplifying for store managers, how to bring that data then forward once you’ve chosen a candidate to bring that into registration in the HR systems and of course then payroll systems and all of that. Is that something you guys have looked into?
07:05
Melissa Shelley Höjwall
And he said no, but that is definitely a really interesting next step for us.
07:11
Melissa Shelley Höjwall
Now if we want to then think agentic through the whole workflow end to end, it’s not going to be enough with one agent or maybe even two agents because first of all you need to make sure that the underlying data that you’re using end to end in a recruitment flow from the start of creating a requisition to hire that includes data on the role or job that you’re using, then you know, making that into, you know, a job requisition in your recruitment system, for example, or if it’s part of a larger erp and then you know you’re going to get start getting data on the candidate and then you want to move that into an HR system and then a payroll system and all of that at the same time as you need data to use these niched chats for, sorry, these niched AI agents for volume assessment, et cetera, you need to create a stable end to end data flow and you are going to have to do the hard work of, for example, in a global organization of aligning those kind of data inputs so that you get this stable data foundation where apples are compared to apples and pears compared to pears.
08:27
Melissa Shelley Höjwall
Pairs are compared to pairs and only then will the individual niched AI agents actually have reliable data that they can do their work on. So let’s say that we have this perfect data layer with excellent data and I think not all companies are there yet. And this is an area where we are putting a lot of focus in addition to cool things in AI because we see that as a very good foundation and we need to align not only data globally, but we need to align how data gets inputted in an aligned way. But let’s say we have that, but then we have still multiple. And we will have multiple agents to handle end to end HR processes, even if we redesign them completely.
09:18
Melissa Shelley Höjwall
So I think what is new and what most companies have not yet succeeded with is the rise of what Josh Burshin calls the super agent. The kind of orchestrating agent that holds together the other agents and can create this kind of more cohesive experience. Not only like a cohesive experience for the customer or the end user, but also one that can actually get agents to communicate with each other or run tasks that are where different parts of the process are done by different agents that are accessing the same data. Foundation preferably, and not multiple data foundations, even if that would be the case. So I hate to say it, Matt, but in addition to all the fun and attractive stuff, companies really need to continue investing in data governance. They need to continue working on process design and automating process design for data.
10:18
Melissa Shelley Höjwall
And they also need to work with integration. Still, I do foresee that integrations will become less and less important over time as AI agents get better and these orchestration agents become more powerful. But we are not there yet. And I don’t see that happening in the next five years. And let’s see if AI proves me wrong.
10:37
Matt Alder
I think that’s really interesting because a lot of the value of this kind of agentic approach to things, and you mentioned that kind of rethinking the whole HR process comes from this end to end working because you know, if you have that data and you can understand that you can analyze someone’s skills in the recruitment process and that kind of flows through so then you kind of don’t lose that data and that kind of helps later on in their journey. That just seems to be the whole value of it. And about sort of four or five years ago I wrote a book with Mervyn Dylan, who’s actually here, just taking a photo of us a second ago, called Exceptional Talent. And we talked about the problem being with HR is that it’s in silos.
11:17
Matt Alder
And were talking about it from the perspective of the employee journey. So for an employee, you’re just going through the experience of working somewhere, but you’re just all these silos in hr. And you’re right, it’s kind of AI, it sort of promises to break those, but it’s just not going to do it by magic or by itself, is it?
11:35
Melissa Shelley Höjwall
No, definitely not.
11:36
Matt Alder
Yeah.
11:37
Melissa Shelley Höjwall
And I think if we translate that then into like what new jobs do I see coming? I think larger HR departments and larger companies probably want to invest in an AI architect, an architect of HR agents that can work with the enterprise architects and the other solution architects in the company. Because then obviously next step is bridging it into, you know, the finance agents and the, you know, whatever other cross functions like sales agents, et cetera. So that is kind of like a new role that is definitely going to become interesting. And I think that the HR roles will change because I think that a lot of our HR process experts will more become like orchestrators of cohesive experience.
12:33
Melissa Shelley Höjwall
And I do think that we will go back to a space where very specialized HR people will become more generalized in creating orchestration between processes. Like, for example, I’m not sure that we will in five years time have pure like talent acquisition teams anymore. I think that talent acquisition and talent management will merge. I think that we’re going to see a lot more like general orchestrators creating a cohesive experience where a lot of the legwork is handled by agents.
13:08
Matt Alder
It kind of indicates just how much sort of work there is to do to realize the potential and the value that lots of people are talking about at this event. Picking up on something that were talking about the last time you were on the podcast. We were talking about adoption of new technology. And what’s kind of been really interesting in some of the reports and the stats that I’m seeing at that kind of enterprise level is employees are using AI in their personal lives and all this kind of thing, but actually it’s not being used at work so much. What’s your kind of take on that? Is it just part of this big sort of data problem? Is it that people are afraid of kind of bringing this to work? Is it that organizations can’t move at the speed of the new technology?
13:57
Matt Alder
What’s kind of happening out there in terms of just sort of general adoption of even the sort of the most basic ways of using these tools.
14:04
Melissa Shelley Höjwall
I’m going to be a bit provocative now and I’m generalizing heavily. And, and this is not about H&M as an example, for sure. This is more a general take that I’m seeing when I’m talking to different people working in large corporations. I think a lot of corporations are making the mistake of trying to govern AI to the degree where they’re actually hindering innovation in their own companies. So I see a lot of large companies that still haven’t like, given every single, like at least office work access to A safe environment where they can increase their productivity with the help of AI. Now, I think that is a huge hinder for people actually doing this in a good way. And what happens in those companies, even though we pretend like that’s not the case, is that leads to shadow AI.
14:53
Melissa Shelley Höjwall
I don’t think that there are a lot of people today in office that are not working with AI even for work. It’s just that they might not have any tools that are accessed, or their companies have forbidden the use of AI, haven’t created helpful strategies, et cetera. So I think we can assume that a lot of people are using AI for work. Is that being done safely? Probably not, but that’s because they haven’t been given the right prerequisites.
15:18
Melissa Shelley Höjwall
Now, if you then look at this from a broader perspective, I think that at the same time, another very scary but also interesting phenomenon is happening, and that is that even in companies where you don’t have the right, where companies have not given the right prerequisites for people to be innovative with AI, all of a sudden management or management consultants supporting management come in and decide that, hey, you know, we’re going to reduce this department by 20%. And then they quote AI, and then they, you know, go out and make people redundant. And so honestly, I foresee that there will be still movements where adoption will be forced. In the simple case that if you cut a department without having done your AI work first quoting AI, that’s going to force people to have to use AI to succeed.
16:23
Melissa Shelley Höjwall
Now, is that what I think is good? Is that what I recommend? Absolutely not. I would like to give a really good example from the Nordics, a company that I admire for many reasons that have taken a completely different stance on this. And that’s ikea.
16:38
Matt Alder
Yes, yes. Tell us about that. It’s a really interesting case study.
16:43
Melissa Shelley Höjwall
Yeah. IKEA did what many companies have done. They used AI to revolutionize their customer service. So AI agents basically made customer service agents obsolete. But instead of just letting these very knowledgeable people leave the company, these excellent customer service agents, IKEA asked the customers, okay, what things would you have wanted service or support from customer service with if they had time that you don’t, you know, that we aren’t being able to give you now with this agent service. And the customers were talking a lot about that. They wanted more support with interior design choices of, you know, furniture and stuff. So what did IKEA do?
17:32
Melissa Shelley Höjwall
They upskilled these fantastic customer service colleagues that had all of this knowledge about customers and about the product that were proud ambassadors of the company and they increased their sales because the upskilling or, sorry, the upselling of this just, you know, made it exponential. So I hope that more companies will follow, but I think it’s going to be a combination of forced and more positive types of adoption. So I don’t think it’s going to be a change management company question. I think that, that AI will become the norm quicker than later. But I just hope that companies do this wisely.
18:11
Matt Alder
Final question for you. You’ve very much kind of been involved in this event.
18:15
Matt Alder
I know you’ve spoken to lots of.
18:16
Matt Alder
People and been around and sort of listened to a lot of things that are going on. What’s the most interesting or surprising thing that you’ve heard or discovered in the last couple of days?
18:27
Melissa Shelley Höjwall
I am still surprised by how few people, and I mean the people here are still kind of more, I would say they’re representing the more forward leaning and interested people in this space. But I am still surprised at how many people here have not created their own agents, have not vibe coded, have not kind of. They might have reached the stage of being worried, but maybe are not like trying to think about, okay, what is the next step for me? How do I need to stay relevant and everything? And this is not to point fingers at anyone, but I think if I could give advice to anyone who feels worried or concerned. Go online, ask ChatGPT how do I create an agent? Or ask, you know, Claude or whatever system you’re using. Talk to someone who you know has made an agent.
19:23
Melissa Shelley Höjwall
Start thinking about what is it that I do today that gives me a lack of energy that’s like time consuming. Is there a way that I could create an agent? And it is a lot easier than you think. And I think if you haven’t even tried to make an agent, if you haven’t done this thing, I think it’s really important that you do to stay relevant.
19:48
Matt Alder
I think that’s absolutely fantastic advice because I think until you have been sort of hands on with this stuff, it’s very difficult to understand the context, understand just how it’s going to change work.
19:58
Melissa Shelley Höjwall
Yeah.
19:58
Matt Alder
Melissa, thank you very much for talking to me.
20:00
Melissa Shelley Höjwall
Thank you so much. It was a pleasure.
20:03
Matt Alder
My thanks to Melissa. You can follow this podcast on Apple Podcasts on Spotify or wherever you listen to your podcasts. You can search all the past episodes at recruitingfuture.com on that site. You can also subscribe to our weekly newsletter Recruiting future feast and get the inside track on 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.
