From AI Hype to Audit-Ready HR

00:00:05 Sarah Kelsey

Welcome to the Deep Dive, the show where we explore the hottest topics in HR, assessment, and the world of work.

00:00:11 Sarah Kelsey

As we dive into the world of AI and HR, the general consensus seems to be that while most HR teams are wanting to use AI, very few feel confident actually implementing it.

00:00:22 Sarah Kelsey

This episode will move us from why to how, a practical guide in building explainable AI processes across the entire talent life cycle.

00:00:30 Sarah Kelsey

To help us map this out today, we have the expert who puts the AI in HR, Martin Redstone.

00:00:36 Sarah Kelsey

Martin is the founder of Eunomia HR and a regular advisor to HR teams on governance, skills, and implementation.

00:00:44 Sarah Kelsey

He has decades of experience in the space and many useful resources, which we will link below.

00:00:49 Sarah Kelsey

We feel extremely lucky to have him on the podcast today.

00:00:52 Sarah Kelsey

Martin, welcome.

00:00:53 Sarah Kelsey

How are you?

00:00:54 Martyn Redstone

Thank you very much for having me.

00:00:55 Martyn Redstone

I’m fantastic.

00:00:55 Martyn Redstone

Thank you.

00:00:56 Sarah Kelsey

Awesome.

00:00:57 Sarah Kelsey

So Martin, you’re the founder of Eunomia HR.

00:00:59 Sarah Kelsey

Could you tell us a little bit more about that and your work in the space.

00:01:02 Martyn Redstone

Sure.

00:01:03 Martyn Redstone

So you know your HR is the next stage in the evolution of what I’ve been doing for several years, which is helping HR leaders and recruitment leaders implement AI effectively and responsibly.

00:01:17 Martyn Redstone

And we’re now taking that responsible very, very seriously.

00:01:21 Martyn Redstone

So you know your HR is the rebirth of the AI

00:01:25 Martyn Redstone

advisory service that I provide now with a specific focus on ethical AI, good governance, and responsible implementation.

00:01:36 Sarah Kelsey

Martin, can I ask you a little bit about how you actually got into this world?

00:01:39 Martyn Redstone

Yeah, absolutely.

00:01:40 Martyn Redstone

So I won’t bother taking up the entire podcast with my background, but I’ve been in this space for about 20 years now.

00:01:47 Martyn Redstone

And for the last nine years, well, nine years ago, I started working

00:01:51 Martyn Redstone

I’ve been working with the kind of technology side of it for the last 15, 16 years of that 20 years.

00:01:56 Martyn Redstone

Nine years ago, got really heavily involved in the world of AI through chatbots.

00:02:01 Martyn Redstone

And I was really the first person in the market who was talking about the benefit of automation, chatbots, conversational AI.

00:02:11 Martyn Redstone

And that kind of carried on really until the explosion of generative AI happened with the launch of ChatGPT.

00:02:19 Martyn Redstone

And everyone was really looking for somebody who already understood it, already got it, got it, and already really understood the technology.

00:02:27 Martyn Redstone

I’d been using language models in my work for a couple of years before ChatGPT.

00:02:31 Martyn Redstone

So

00:02:32 Martyn Redstone

really, I was in a prime position ultimately to do what I was already doing, but more so.

00:02:39 Martyn Redstone

And a lot of the work that I was already doing involved good governance and good foundations for responsible AI implementation.

00:02:48 Martyn Redstone

And so what I found, especially over the last couple of years, is that there’s been some shortcuts taken in the way that people were implementing AI and using AI.

00:02:56 Martyn Redstone

So I decided that I wanted to be the responsible adult in the room and help the industry

00:03:02 Martyn Redstone

Implement effectively, but implement responsibly.

00:03:05 Sarah Kelsey

Amazing.

00:03:05 Sarah Kelsey

Makes so much sense.

00:03:06 Sarah Kelsey

And we know that you also have some really useful resources for people, which we’re going to link in the show notes as well.

00:03:11 Sarah Kelsey

So don’t forget to show those out.

00:03:14 Sarah Kelsey

But I think a great place to start is to ask you where most HR leaders go wrong when trying to implement AI.

00:03:21 Martyn Redstone

Yeah, it’s really interesting because there’s been a couple of reports released last week.

00:03:28 Martyn Redstone

The first one was

00:03:29 Martyn Redstone

the report from Anthropic around how businesses are using Claude, their large language model-based chatbot.

00:03:37 Martyn Redstone

And then we also had OpenAI launch, release a report on how people are using ChatGPT.

00:03:42 Martyn Redstone

And what there is, there’s a huge divide between how people are using it compared to how businesses expect it to be used.

00:03:50 Martyn Redstone

And so I think the biggest mistake that’s happening right now is organisations, and we had it in the advent of social media, and organisations not knowing how to handle people’s use of social media in the workplace for workplace-based tasks.

00:04:07 Martyn Redstone

So LinkedIn in our industry, the big one.

00:04:10 Martyn Redstone

And at first people were saying you can’t use LinkedIn or if you do use LinkedIn, it’s our data and we own your account.

00:04:17 Martyn Redstone

Nobody really knew what was going on, but they couldn’t bridge the gap between something that was particularly personal, but a workplace tool.

00:04:24 Martyn Redstone

And so I think that’s what’s going on right now in the world of HR is this really, really difficult challenge, which is HR leaders who don’t know how to bridge the gap between

00:04:33 Martyn Redstone

utilizing AI at work, but a very personal tool like something like ChatGPT.

00:04:38 Sarah Kelsey

Yeah, I can imagine it would be really confusing for a lot of leaders who are hearing all of the buzz about it and want to make sure they’re staying ahead of the curve and don’t want to fall behind.

00:04:48 Sarah Kelsey

But also, especially in large enterprises, comes a lot of risk.

00:04:52 Sarah Kelsey

regulation and I guess a thoughtfulness around how we implement these tools.

00:04:57 Sarah Kelsey

We actually released a report recently on AI and talent acquisition specifically.

00:05:02 Sarah Kelsey

And we uncovered a stat where 93% of crows are intending to implement AI in some way.

00:05:09 Sarah Kelsey

However, only 1% feel fully ready to do so.

00:05:13 Sarah Kelsey

And I think that speaks to that point exactly.

00:05:15 Sarah Kelsey

Does that resonate from a lot of the conversations you’ve been having recently?

00:05:18 Martyn Redstone

Yeah, absolutely.

00:05:19 Martyn Redstone

I think that there’s this kind of top-down pressure happening.

00:05:23 Martyn Redstone

in most organizations where people are being put under pressure to implement AI, do more with this, all those kind of things.

00:05:30 Martyn Redstone

And I think that there’s a massive gap in the data that Crows have in order to implement AI.

00:05:36 Martyn Redstone

Not only the enablement piece around training, upskilling and what have you, but also when you look at somebody’s job description or the actual job that they do, because there’s sometimes quite a lot of difference.

00:05:48 Martyn Redstone

and you work out, three things, what’s automatable, what’s augmentable, and what’s purely human.

00:05:55 Martyn Redstone

And then you can start from there.

00:05:56 Martyn Redstone

And the same thing across not only the work that people do, but the services that departments, divisions actually deliver to their users, whether that’s internal or external.

00:06:08 Martyn Redstone

So I think that there’s a huge breakdown when it comes to the data that Crows require in order to

00:06:15 Martyn Redstone

know where to start?

00:06:16 Sarah Kelsey

Yeah, it’s interesting.

00:06:17 Sarah Kelsey

And I suppose another question for them is who do they go to or where do they go in order to figure out where to start as well?

00:06:25 Sarah Kelsey

When you hear the term AI by design in HR, does that fall into that?

00:06:31 Sarah Kelsey

Are you thinking around how do we fill these knowledge gaps and how do we actually implement AI thoughtfully?

00:06:37 Martyn Redstone

Yeah, absolutely.

00:06:38 Martyn Redstone

I think that AI by design is quite a new phrase that people are talking about.

00:06:41 Martyn Redstone

And I don’t think that that’s a fair

00:06:44 Martyn Redstone

phrase to use.

00:06:45 Martyn Redstone

I’ll be honest.

00:06:46 Martyn Redstone

I think it’s quite a confusing phrase for a lot of people because for me, the term infer is starting again.

00:06:54 Martyn Redstone

And I don’t think that’s the case.

00:06:55 Martyn Redstone

For me, I’m a big fan of service design.

00:06:58 Martyn Redstone

And for me, AI is an enabler and a part of the service that you might deliver.

00:07:04 Martyn Redstone

But it’s not the be all and end all.

00:07:07 Martyn Redstone

And I think that’s ultimately where the pressure’s coming from, is the crows are

00:07:12 Martyn Redstone

almost thinking, we have to kind of scrap everything and start again with an AI-first design, and that’s not always the case.

00:07:19 Martyn Redstone

I think organizations need to move away from trying to become AI-first and try and become just AI-enabled.

00:07:30 Sarah Kelsey

Really interesting, because I suppose it would be easy to get wrapped up in all of the hype as well.

00:07:35 Sarah Kelsey

What are some common questions that Crows or TA and TD leaders are asking themselves right now that maybe don’t need to hold as much weight as they should in the space?

00:07:45 Martyn Redstone

I think the biggest question that I hear and I get asked is, how do we implement AI agents?

00:07:53 Martyn Redstone

And it’s just this term that you mentioned about kind of the hype.

00:07:59 Martyn Redstone

my counterparts across the pond in the US, last week there was the massive HR tech conference out there.

00:08:06 Martyn Redstone

And some people told me it was more like the AI agent, AI agent conference, because that’s what everyone was talking about.

00:08:12 Martyn Redstone

And I think that, again,

00:08:16 Martyn Redstone

HR leaders, crows, they’re really getting themselves head up in the hype.

00:08:21 Martyn Redstone

And they kind of need to step back and say, okay, well, do we need AI?

00:08:25 Martyn Redstone

Do we need automation?

00:08:27 Martyn Redstone

Do we need a bit of both?

00:08:28 Martyn Redstone

Do we just simply need better processes, better internal communication, all those kind of things?

00:08:33 Martyn Redstone

And so the biggest question I get asked is usually something to do with FOMO, ultimately.

00:08:38 Martyn Redstone

People feeling as though they’re missing out on something because they see across

00:08:43 Martyn Redstone

their peers across LinkedIn, across all those kind of things.

00:08:47 Martyn Redstone

All these kind of fancy terms that are coming out and hype terms that are coming out that just make people feel more under pressure and feeling as though they might be missing out or missing the bus.

00:08:59 Sarah Kelsey

Do you think AI agents and chatbots are overhyped?

00:09:02 Martyn Redstone

I love the fact that you put them both in the same sentence because most people call a chatbot a

00:09:09 Martyn Redstone

an AI agent.

00:09:10 Martyn Redstone

And yeah, absolutely.

00:09:12 Martyn Redstone

you have to remember an AI agent by its very definition is something that has agency, which is something that doesn’t need any human input, can get on, make decisions, decides what tools, what data to use, can go back up on itself, fix itself, et cetera, et cetera.

00:09:27 Martyn Redstone

And in our world, that’s not happening.

00:09:29 Martyn Redstone

Most what people term AI agents are just chatbots or automation with a bit of lipstick on it.

00:09:36 Martyn Redstone

And ultimately, because our world of HR is very human-led and some of the processes and decisions that we make significantly impact somebody’s life, I don’t think we’re going to be seeing true AI agents in our space for a very long time because by the very definition, they don’t need any kind of human input.

00:10:00 Sarah Kelsey

And do you think there are processes that should never be automated?

00:10:03 Martyn Redstone

I think that there has to be a very specific definition made around what kind of automation is going on for some of those processes.

00:10:12 Martyn Redstone

So what we tend to see is 2 different types of automation.

00:10:16 Martyn Redstone

We’ve got automated decision making and automated decision execution.

00:10:21 Martyn Redstone

I think that there’s a lot of processes in our space that cannot have automated decision making.

00:10:27 Martyn Redstone

And that’s why employment focused AI is in the high risk element of the EU AI Act, because when things are making, when AI is involved in making a decision that has a material impact on somebody’s life, employment, it’s one of the top three stresses in somebody’s life next to moving home and having children.

00:10:49 Martyn Redstone

So why on earth would we want a machine to make an automated decision?

00:10:54 Martyn Redstone

Automated decision execution, absolutely.

00:10:58 Martyn Redstone

That’s what we would call automation, ultimately.

00:11:01 Martyn Redstone

We are asking a machine to do things based on how we would do it, and we’re checking that off as humans and making sure that’s working properly.

00:11:09 Martyn Redstone

So yeah, so I think to answer your question, there’s a lot that we can do for automated decision execution, but decision making around

00:11:18 Martyn Redstone

anything that involves somebody’s employment, whether that be hiring, firing, promotion, anything involved in that, we can’t ask a machine to automate that fully.

00:11:30 Sarah Kelsey

So decision execution seems lower risk than decision making.

00:11:36 Sarah Kelsey

And how would a team know when they’re ready?

00:11:39 Sarah Kelsey

Is that something they would start with decision

00:11:43 Sarah Kelsey

execution and work up from there and to using bigger tools or implementing them more and more?

00:11:48 Sarah Kelsey

How would a team know that they’re ready to or ready to use AI within their processes?

00:11:54 Martyn Redstone

When I’m working with my clients, we have a framework which is crawl, walk, and run.

00:11:59 Martyn Redstone

Ultimately, you can’t run before you can walk and you can’t walk before you can crawl.

00:12:03 Martyn Redstone

So when we’re looking at the entirety of the processes and understanding what is automatable,

00:12:12 Martyn Redstone

We are, we’re looking at the easy wins, the low hanging fruit, the low risk stuff.

00:12:16 Martyn Redstone

So things that don’t involve any form of decision making, things like FAQ chatbots.

00:12:22 Martyn Redstone

So, you know, from an HR perspective, you know, HR teams are inundated with queries around what’s the policy for this?

00:12:29 Martyn Redstone

What’s the process for that?

00:12:30 Martyn Redstone

Where do I find this?

00:12:31 Martyn Redstone

Where do I find that?

00:12:32 Martyn Redstone

Hugely,

00:12:34 Martyn Redstone

low risk automation can go on there, but actually highly valuable as well.

00:12:39 Martyn Redstone

Because the amount of time that HR organizations take up in just answering those questions can be wiped out immediately.

00:12:49 Martyn Redstone

So highly valuable, but highly low risk.

00:12:51 Martyn Redstone

So that’s the cruel piece.

00:12:52 Martyn Redstone

Once you’re really starting to see the impact of that, that’s embedded.

00:12:57 Martyn Redstone

You can then start moving yourself towards walking, which is something probably a little bit more high risk.

00:13:02 Martyn Redstone

something that’s a little bit more wider, impactful across the team or across the organization.

00:13:08 Martyn Redstone

and it’s just slowly, slowly.

00:13:10 Martyn Redstone

you don’t have to do the whole big bang thing.

00:13:12 Martyn Redstone

going back to your previous question around big mistakes, that’s one of them.

00:13:17 Martyn Redstone

people…

00:13:19 Martyn Redstone

try to go all in on big bang when it comes to implementing AI and automation.

00:13:23 Martyn Redstone

It’s the wrong way to do it.

00:13:25 Martyn Redstone

Find something low risk, find a really good use case, do that first, and just start building upon that.

00:13:31 Sarah Kelsey

So it sounds like doing an assessment of where those high value, low risk tasks that might be a good way to work out the easy ones for AI implementation.

00:13:41 Martyn Redstone

Absolutely, absolutely.

00:13:42 Martyn Redstone

So there’s a big chunk of work that has to be done right at the beginning, which is

00:13:46 Martyn Redstone

Everything from time and motion studies to interviewing both the team members, the users of that service, the management, et cetera, et cetera.

00:13:57 Martyn Redstone

And it’s a long, it can be not a long drawn out process, but it has to be done properly.

00:14:02 Martyn Redstone

And from there, you get the automation opportunities.

00:14:05 Martyn Redstone

You also get some fantastic pieces of work that can go towards board level evidence for getting investment and all those kind of things, but it has to be done properly.

00:14:14 Martyn Redstone

It can’t be done on a whim.

00:14:16 Martyn Redstone

So you do have to do that piece of work first.

00:14:18 Sarah Kelsey

Yeah, okay, makes complete sense.

00:14:20 Sarah Kelsey

And so you’ve done some work and research around the effectiveness of conversational AI and chatbots.

00:14:26 Sarah Kelsey

Could you tell us a little bit about that work?

00:14:28 Martyn Redstone

Yeah, so I was really concerned because there’s a lot of new vendors coming out in our space, especially in the recruitment and talent acquisition space where, and also, you know, a lot of things being built internally where people are thinking,

00:14:42 Martyn Redstone

great.

00:14:43 Martyn Redstone

now we’ve got access to these powerful AI models like, GPT and Claude and Gemini and what have you.

00:14:51 Martyn Redstone

We can look at parts of our process that are really blocking and we can automate them.

00:14:58 Martyn Redstone

We can fire in some powerful AI.

00:15:00 Martyn Redstone

One of the biggest challenges in the recruitment process nowadays is just the inundation of applications.

00:15:06 Martyn Redstone

So people are now looking and going, great, I’ve got a problem.

00:15:09 Martyn Redstone

I’m going to fix it.

00:15:10 Martyn Redstone

And they’re asking,

00:15:11 Martyn Redstone

and they’re building tools that are using large language models to do CV screening, ultimately.

00:15:19 Martyn Redstone

And I’ve always had an issue with that because I think it’s the wrong way of doing things.

00:15:23 Martyn Redstone

We’re asking…

00:15:25 Martyn Redstone

generative AI to do discriminative AI work.

00:15:28 Martyn Redstone

And the two just don’t mix as far as I’m concerned.

00:15:30 Martyn Redstone

So rather than just getting on my soapbox and wanting to tell the world off about doing this, I thought, well, let’s do some research.

00:15:38 Martyn Redstone

So over a course of a few weeks, I every day used three AI models, ChatGPT, Google, Gemini, and Grok from XAI.

00:15:52 Martyn Redstone

the same 109 CBs against the same job description with the same prompt across all three.

00:15:58 Martyn Redstone

So that was 109 times 3 every day for three weeks.

00:16:02 Martyn Redstone

And the results coming out of there were really scary, really scary.

00:16:09 Martyn Redstone

Some of the top level results were that the rank drift that was happening.

00:16:15 Martyn Redstone

So on average,

00:16:18 Martyn Redstone

If we were looking at, just pulling out the top 10, that would change every day.

00:16:22 Martyn Redstone

And on average, each placement was changing by plus or minus 2 1/2 places.

00:16:28 Martyn Redstone

At the extreme, one CV that came in at #1 disappeared out of the top 10 the next day.

00:16:36 Martyn Redstone

So that was the first kind of major worry is that there was no repeatability, which again is just dreadful from a screening perspective.

00:16:45 Martyn Redstone

The second major one was that, and there’s been some research coming out in the last couple of weeks that has really confirmed this as well, which is I found a very odd pattern, which was that the large language model would work through the CVs and it would only, it would stop once it had its top 10.

00:17:03 Martyn Redstone

So it wouldn’t look at all 109.

00:17:05 Martyn Redstone

So on, not even on average across the board, 55% of CVs weren’t even looked at.

00:17:12 Martyn Redstone

And there was another one, which is that the reasoning behind the decision was very boilerplate.

00:17:17 Martyn Redstone

And we know from a lot of other research that the way that large language models do their reasoning is actually post-event, because ultimately they are text generation models.

00:17:27 Martyn Redstone

So you ask it why it did something, and it will make up something based on your question that seems quite okay, but actually isn’t the reason why it did it.

00:17:36 Martyn Redstone

So there’s lots and lots of

00:17:39 Martyn Redstone

Big red flags, and I would say that anybody that’s using a large language model based process or product that is doing discriminative work, so screening CVs, assessing people, don’t do it because that’s not what it’s designed for.

00:17:57 Sarah Kelsey

No massive red flags and pretty much just wipes that out as a viable option.

00:18:02 Martyn Redstone

Absolutely.

00:18:02 Sarah Kelsey

At all.

00:18:03 Martyn Redstone

Yeah.

00:18:04 Sarah Kelsey

Wow.

00:18:04 Sarah Kelsey

And were there any insights from that you think are applicable from a talent development perspective as well?

00:18:11 Martyn Redstone

Yeah, I think, like I said, I think that any solution that you’re either building or buying that involves assessing people, and as I said, that’s doing discriminative work, you can’t build that

00:18:26 Martyn Redstone

on the basis of a large language model.

00:18:28 Martyn Redstone

So if you’re building that internally, you can’t, you need to ask yourself questions.

00:18:34 Martyn Redstone

You need to do those things like, you know, if I’m assessing somebody, whether that’s for joining, leaving, or developing, you know, do those tests.

00:18:43 Martyn Redstone

You know, I’m throwing this data through.

00:18:45 Martyn Redstone

When I do it tomorrow, is it the same outcome?

00:18:48 Martyn Redstone

Et cetera, et cetera, et cetera.

00:18:50 Martyn Redstone

Do those tests internally if you want to test it.

00:18:53 Martyn Redstone

And ask the question of your vendors as well.

00:18:56 Martyn Redstone

are you using large language models?

00:18:58 Martyn Redstone

Have you done tests around rank drift, repeatability, stability, et cetera, et cetera?

00:19:05 Martyn Redstone

Ask the important questions.

00:19:06 Martyn Redstone

And that’s the bit that I try and do with the work that I’m doing at EnoMia HR, which is…

00:19:11 Martyn Redstone

Just getting people to understand they need to ask the right questions in order to protect themselves and the users of their systems.

00:19:18 Sarah Kelsey

Yeah, it sounds like that testing and experimentation piece is also really important as well to make sure that you’re not just using a tool for the sake of the hype or because of the FOMO that you mentioned.

00:19:29 Sarah Kelsey

Exactly.

00:19:30 Sarah Kelsey

But because you really know and can prove that it works.

00:19:34 Martyn Redstone

Yep, exactly.

00:19:36 Sarah Kelsey

And a common problem we see in relation to this with a lot of our enterprise clients is this tug of war between speed and implementation.

00:19:44 Sarah Kelsey

And I think that this is an issue that comes up for many different reasons, not just the implementation of AI.

00:19:51 Sarah Kelsey

But where smaller firms can kind of be a lot more agile and maybe implement and discard really quickly, that’s harder for bigger organizations.

00:20:01 Sarah Kelsey

So from your perspective, do you see

00:20:06 Sarah Kelsey

How do you see bigger organizations being able to keep up without breaking the rules?

00:20:11 Martyn Redstone

That’s a great question.

00:20:13 Martyn Redstone

Anybody that knows me knows that I love an analogy.

00:20:16 Martyn Redstone

And anybody that knows me knows this analogy inside and out now because it’s something I repeat time and time again.

00:20:24 Martyn Redstone

It’s called guardrails, you know, and the whole point of guardrails, we think about it, you know, on a motorway, a highway, you know, right in the middle of the road, there’s these metal rails that go along.

00:20:36 Martyn Redstone

And we call them guardrails.

00:20:38 Martyn Redstone

They’re not there to slow the traffic down.

00:20:40 Martyn Redstone

They’re there to help people drive at speed safely.

00:20:45 Martyn Redstone

Because if they weren’t there, and they normally are right next to the fast lane, if they weren’t there, then people in the fast lane would be crawling along slowly, scared they’re going to veer off into the oncoming traffic and cause a horrific crash.

00:20:55 Martyn Redstone

Now, the same applies to setting guardrails in an organization when it comes to the adoption of AI.

00:21:02 Martyn Redstone

Before you do anything else,

00:21:04 Martyn Redstone

Decide on your guardrails, decide on your acceptable use policy, decide on your vendor strategy, decide on your privacy policies, decide on your transparency policies, et cetera, et cetera.

00:21:15 Martyn Redstone

Set those all first, because when you set those guardrails, people then know how to innovate and they can do it at speed.

00:21:23 Martyn Redstone

So the whole point of guardrails is not to get in the way of innovation, it’s to allow people

00:21:28 Martyn Redstone

to know exactly what they can do, how they can do it, and get on with it and do it quicker.

00:21:33 Martyn Redstone

And that’s again, one of the biggest mistakes that I see is guardrails being set retrospectively.

00:21:39 Martyn Redstone

And that ends up actually putting a kibosh on a lot of projects that people have just got on with because they didn’t know the rules.

00:21:44 Martyn Redstone

They didn’t know what they can and can’t do.

00:21:47 Martyn Redstone

And when they start setting guardrails retrospectively, it pushes back innovation.

00:21:52 Martyn Redstone

So do it first and allow people then to innovate at speed safely.

00:21:57 Sarah Kelsey

It’s such a great

00:21:58 Sarah Kelsey

analogy.

00:21:59 Sarah Kelsey

It really makes a lot of sense.

00:22:00 Sarah Kelsey

Why do you think that companies are often setting these guardrails retrospectively?

00:22:06 Martyn Redstone

Because like I said right at the beginning, what we’re seeing nowadays is a lot of people shortcutting.

00:22:11 Martyn Redstone

They follow along the hype.

00:22:14 Martyn Redstone

They’ve tinkered with ChatGPT.

00:22:15 Martyn Redstone

They now think that they’re an expert.

00:22:17 Martyn Redstone

They might bring somebody in from external who hasn’t really had much experience in

00:22:24 Martyn Redstone

delivering AI-based transformation projects or any form of technology-based transformation projects.

00:22:29 Martyn Redstone

And so, unfortunately, the basics are skipped or forgotten or not even thought about.

00:22:35 Martyn Redstone

They dive straight in wanting to make a huge impact really quickly.

00:22:39 Martyn Redstone

And that’s what’s happening.

00:22:41 Martyn Redstone

And that’s why a lot of organizations now are talking to me about, unfortunately, having to retrospectively set those guardrails.

00:22:49 Martyn Redstone

But any organization that’s already starting to think about it and not actioned it yet, get on with the guardrails.

00:22:56 Sarah Kelsey

Yeah, it seems like a great way to also get buy-in from other people to show that this is something that will be useful and safe to implement.

00:23:04 Martyn Redstone

Exactly.

00:23:04 Sarah Kelsey

And do you think bigger organizations have…

00:23:07 Sarah Kelsey

any kind of built-in advantage that maybe a smaller firm might not have when it comes to AI or implementation?

00:23:14 Martyn Redstone

I think so, if done properly.

00:23:17 Martyn Redstone

So larger enterprise level organizations, they have a legal team, a procurement team, an IT team, et cetera, et cetera.

00:23:25 Martyn Redstone

But actually what I find is that Crows and HR in general, they have to be the bastions of AI transformation.

00:23:33 Martyn Redstone

They have to be the leaders of AI transformation because

00:23:38 Martyn Redstone

Every form of AI transformation touches people.

00:23:42 Martyn Redstone

And we see it in the news, we’ve got reduction in force due to AI.

00:23:45 Martyn Redstone

We’ve got, we’re hiring a different type of person due to AI.

00:23:49 Martyn Redstone

We’re changing job descriptions due to AI.

00:23:52 Martyn Redstone

These are all very much centered around people.

00:23:56 Martyn Redstone

So HR needs to take the leadership, but in taking the leadership, they need to bring all those people together in order to help set the direction, set the guardrails, and own the strategy.

00:24:07 Sarah Kelsey

And I’m excited to talk about how company structures are changing a little bit as a result of AI in a second, but I guess something a bit more fun first.

00:24:14 Sarah Kelsey

If you could change any classic enterprise policy to help unblock the use of AI in a company, what would it be and why?

00:24:21 Martyn Redstone

That’s a really, really good question.

00:24:24 Martyn Redstone

I think that

00:24:26 Martyn Redstone

AI isn’t being blocked, I don’t think, because of a classic policy.

00:24:32 Martyn Redstone

I think it’s being blocked because of a lack of policy.

00:24:35 Martyn Redstone

So I’m just going to turn that on its head a little bit and go back to the whole guardrails thing because

00:24:40 Martyn Redstone

When we see reports coming in, research coming in about why people aren’t adopting AI, it’s because mainly a concern around ethics, legal, et cetera, et cetera.

00:24:53 Martyn Redstone

Not technology, because technology is there.

00:24:55 Martyn Redstone

It’s about the safety and the ethics behind it.

00:24:58 Martyn Redstone

So I actually think it’s not a

00:25:02 Martyn Redstone

traditional policy that’s stopping AI, it’s a lack of new policy.

00:25:06 Sarah Kelsey

Yeah, it’s an incredibly valid perspective.

00:25:08 Sarah Kelsey

I think when there is a lot of risk involved with implementing new things, it’s really about finding the way to do it, to find the guardrails to implement as opposed to what can we move out of the way to get going.

00:25:23 Sarah Kelsey

Very interesting.

00:25:24 Sarah Kelsey

And so if I’m a crow or a head of TA at the moment, what’s one move this quarter that will change

00:25:32 Sarah Kelsey

the pace for me.

00:25:35 Martyn Redstone

I’m going to keep going back to it.

00:25:36 Martyn Redstone

If you haven’t done it, get your guardrails in place because you don’t need to actually be working with AI in order to set the foundations.

00:25:44 Martyn Redstone

You know, acceptable use policies, transparency policies, procurement policies, you know, policies aligned to legislation, aligned to ISO frameworks, aligned to list frameworks, you know, making sure that

00:25:57 Martyn Redstone

you are setting yourself up for success in the next step, whether that be next year, next quarter, next half.

00:26:05 Martyn Redstone

Start now by just setting those policies.

00:26:07 Martyn Redstone

And like you said earlier, you do those and it creates a fantastic internal conversation as well.

00:26:14 Sarah Kelsey

Okay, and if we could simplify that down to the number one guardrail, what’s the number one guardrail that needs to be implemented?

00:26:20 Martyn Redstone

An AI use policy.

00:26:21 Sarah Kelsey

Okay.

00:26:22 Martyn Redstone

Yeah, start with that, and that will also bridge the gap between shadow AI and what’s going on internally as well.

00:26:30 Sarah Kelsey

Okay, great.

00:26:30 Sarah Kelsey

That’s incredibly practical and useful.

00:26:32 Sarah Kelsey

So moving on to, I suppose, how…

00:26:35 Sarah Kelsey

structurally companies are changing as a result of AI and the implementation of AI.

00:26:40 Sarah Kelsey

So Moderna, a pharmaceutical and biotech company, they merged their people and tech functions earlier this year.

00:26:47 Sarah Kelsey

They created a new role of chief people and digital technology officer.

00:26:51 Sarah Kelsey

What do you think of that?

00:26:52 Sarah Kelsey

And do you think we’re going to see more of this coming over time?

00:26:54 Martyn Redstone

I think it’s fantastic.

00:26:56 Martyn Redstone

You know, Jensen, you know, NVIDIA CEO, founder has been in the news recently because he’s been over here in the UK, you know,

00:27:03 Martyn Redstone

chucking a load of money AI here.

00:27:04 Martyn Redstone

But last year, he famously said that IT is going to be the new HR.

00:27:13 Martyn Redstone

And everyone sat back and went, oh, don’t be so silly.

00:27:15 Martyn Redstone

But actually, we’re seeing that now.

00:27:16 Martyn Redstone

And I go back to my previous point that what we’re seeing now is that every AI focused transformation that’s happening in a business has a human element to it as well.

00:27:26 Martyn Redstone

And so I think that

00:27:29 Martyn Redstone

seeing what’s happening at Moderna, I absolutely think that’s going to happen more and more when people understand that IT isn’t systems in silo, it’s systems that impact people.

00:27:41 Sarah Kelsey

Have you seen any other examples of that recently that come to mind?

00:27:45 Martyn Redstone

Yeah, so we’ve seen, especially in the HR world, we’ve seen people and technology strategists coming on board in HR.

00:27:52 Martyn Redstone

We’ve seen Dropbox creating a role like that.

00:27:55 Martyn Redstone

We’ve seen Disney creating roles like that.

00:27:57 Martyn Redstone

So kind of big

00:27:59 Martyn Redstone

big enterprise businesses, big global businesses are now starting to understand that in order to see a success in the future, we need to marry up people and technology.

00:28:09 Martyn Redstone

And I absolutely see that in the future.

00:28:11 Martyn Redstone

I think that…

00:28:12 Martyn Redstone

the role, the future role of TA isn’t going to be talent acquisition.

00:28:17 Martyn Redstone

I think it’s going to be resource acquisition and resource allocation.

00:28:21 Martyn Redstone

So I think that eventually, and we talked about it earlier in terms of the data need around jobs, you know, what part of a job is human only, what’s automatable and what’s augmentable.

00:28:30 Martyn Redstone

You know, I think that, you know, we’re going to have a role in the future where

00:28:35 Martyn Redstone

we’re going to look at a job and we’re going to say, okay, well, we know that a job is a group of tasks that are done at different skill levels, at different competency levels.

00:28:45 Martyn Redstone

So we’re going to work out what’s best to do that task.

00:28:48 Martyn Redstone

Is it going to be a human or is it going to be a machine?

00:28:50 Martyn Redstone

So I think the future of HR and TA is going to be all around

00:28:54 Martyn Redstone

human and machine teaming.

00:28:56 Martyn Redstone

And so we’re definitely going to see a lot more of that in the future.

00:28:59 Sarah Kelsey

It’s quite an exciting time, really, isn’t it?

00:29:01 Martyn Redstone

Absolutely.

00:29:01 Sarah Kelsey

I mean, understandably a lot of fear for people as well.

00:29:05 Sarah Kelsey

However, I think it’s quite cool to see functions in a company merging that maybe we never would have anticipated.

00:29:13 Martyn Redstone

No.

00:29:13 Sarah Kelsey

I think with Moderna, there was an example with this new digital technology officer who, or the new title, was working with the chief operations officer around

00:29:24 Sarah Kelsey

things did need to be automated and which was a human job, et cetera, et cetera.

00:29:29 Sarah Kelsey

I find that such an interesting conversation that prior to probably wasn’t happening as much and not in the same way anyway.

00:29:35 Martyn Redstone

No, and it’s going to change the face of work.

00:29:38 Martyn Redstone

You know, we know that we’re going through an industrial revolution and industrial revolutions always change the face of work, you know, whether that be going from a horse-drawn plow to a steam-drawn plow, whether that be, you know, people in a factory compared to robots in a factory.

00:29:52 Martyn Redstone

You know, when we see

00:29:54 Martyn Redstone

an industrial revolution happening, we know it changes the face of work.

00:29:58 Martyn Redstone

And so that’s what we’re in right now.

00:30:00 Martyn Redstone

We’re at the very, very early stages of it and it takes almost a generational shift to change that world of work.

00:30:07 Martyn Redstone

So all I would say is ride the wave.

00:30:10 Martyn Redstone

You know, it’s happening.

00:30:11 Martyn Redstone

There’s nothing that people can do to stop it.

00:30:12 Martyn Redstone

I know there’s been kind of almost protest movements happening.

00:30:16 Martyn Redstone

anti-AI protest movements happening all over the world.

00:30:19 Martyn Redstone

And we’ll probably see more of that, but it is happening, and we need to ride the wave, otherwise we’d still be living in caves if we didn’t allow innovation to happen.

00:30:29 Martyn Redstone

So my call out to people is understand what the future might hold for you and work out what you need to do to reskill, to upskill and get ready for a new future of work.

00:30:41 Sarah Kelsey

And so switching gears a little bit now, we hear a lot about AI transforming hiring in particular.

00:30:47 Sarah Kelsey

And something we’re quite curious about more and more is how this impacts talent development.

00:30:54 Sarah Kelsey

And I think your comments around resource allocation and taking someone that might be in one role already existing in the business to another, I think that speaks to that as well.

00:31:01 Sarah Kelsey

And it’s quite an interesting space that we’ll see develop over time.

00:31:05 Sarah Kelsey

What percentage of development processes do you think are or will be automated?

00:31:11 Martyn Redstone

I think that there’s a lot of the process that can be automated.

00:31:14 Martyn Redstone

We’ve already seen it, even a couple of years ago, BCG were saying that, their internal LNG bit at BCG University, their content creation, went from weeks to hours using AI.

00:31:29 Martyn Redstone

And I think that for me, that’s one of, that’s going to be one of the key transformations around development is going to be the rapid way that not only can you create development content,

00:31:40 Martyn Redstone

but also that you can serve up something very personalized towards what the person does or even what they want to do in the future.

00:31:47 Martyn Redstone

So I think personalized development is going to be the huge thing, especially as we move towards a world of work which is skills-based.

00:31:54 Martyn Redstone

So I think a lot of development will be automated.

00:31:58 Martyn Redstone

And I’ll go back to my previous point, which is that when it starts stepping into the realm of decision-making on somebody’s employment, that’s when you need to kind of step back a bit.

00:32:07 Martyn Redstone

But anything that involves

00:32:09 Martyn Redstone

giving people further and more opportunity to develop themselves as individuals.

00:32:13 Martyn Redstone

Absolutely.

00:32:14 Sarah Kelsey

Okay.

00:32:15 Sarah Kelsey

And bringing that into context, if you were, say, the chief people and digital technology officer at Moderna right now, what would be some of the first few questions you would be asking when it comes to the implementation of AI in their development programs?

00:32:28 Martyn Redstone

I would be going back to the data around what people do for work, what people do in their jobs, what skills they need for those.

00:32:36 Martyn Redstone

And again, you know, what part of that is augmentable?

00:32:40 Martyn Redstone

and then helping people work better with machines in order to augment themselves better.

00:32:45 Sarah Kelsey

Okay, great.

00:32:46 Sarah Kelsey

That’s really interesting.

00:32:47 Sarah Kelsey

A huge consideration for our clients, and I know for a lot of the people you have conversations with as well, is the accountability and explainability piece around decisions or the implementation of AI when making any kind of decision or executing anything as well.

00:33:04 Sarah Kelsey

What is the minimum that HR leaders need in the audit trail so that they can justify and defend decisions?

00:33:11 Martyn Redstone

Yeah, so first of all, they need to show that they’ve been auditing for things like bias and what have you on a regular basis.

00:33:18 Martyn Redstone

That’s the first piece of data they need.

00:33:21 Martyn Redstone

That gives you a little bit of transparency, a little bit of explainability.

00:33:25 Martyn Redstone

However, for me, it comes down to a simple rule.

00:33:28 Martyn Redstone

Can you explain why this machine made that decision?

00:33:32 Martyn Redstone

in a natural language way that somebody with no technology background could understand.

00:33:37 Martyn Redstone

It’s as simple as that for me.

00:33:39 Martyn Redstone

That is the core concept of transparency and explainability.

00:33:44 Sarah Kelsey

And what powers do you think humans will keep in this area?

00:33:48 Sarah Kelsey

Like why are they still absolutely necessary to these processes?

00:33:52 Martyn Redstone

When I’m working with people on how they augment what they do or how they work with AI systems, one of the things that I try and impart is that

00:34:01 Martyn Redstone

and it came out of the research that we spoke to earlier, is that when you’re looking at large language model-based chatbots like ChatGPT and what have you, they are highly intelligent interns that do not know how to do anything.

00:34:14 Martyn Redstone

That’s how you kind of have to think about it.

00:34:16 Martyn Redstone

So I always talk to people and say, look, you know, when you’re, let’s say you’ve got an intern there, you wouldn’t just say, go away and do this and expect the best.

00:34:24 Martyn Redstone

You’d first of all, delegate that task to them properly, tell them exactly what you want, how you want it, give them some examples of work done in the past, et cetera, et cetera.

00:34:33 Martyn Redstone

But the key part from that is when they bring it back to you and say, here’s the work that I’ve done.

00:34:39 Martyn Redstone

It’s the human skill of evaluating that work, using your experience, your knowledge, and your expertise to evaluate that work, and then coach them through how to do it better.

00:34:49 Martyn Redstone

And that is a particularly human skill.

00:34:52 Martyn Redstone

that we’re going to have and we’re going to continue having, alongside things like learning velocity and what have you, so you can actually keep up with the pace of change.

00:34:59 Martyn Redstone

But I think that humans are absolutely still going to be involved, especially in our world when it comes to people.

00:35:04 Martyn Redstone

to question, to evaluate, and to point in the right direction.

00:35:09 Sarah Kelsey

That really simplifies it.

00:35:10 Sarah Kelsey

To summarize, it sounds like you could say the human is speaking to the why in every decision.

00:35:18 Martyn Redstone

Exactly.

00:35:18 Sarah Kelsey

And I suppose we add that layer of explainability.

00:35:22 Sarah Kelsey

We need to be able to explain almost better or beyond the AI as to why this decision was made.

00:35:29 Martyn Redstone

Exactly, yeah.

00:35:31 Sarah Kelsey

What’s one question that TA leaders, and I know you’ve spoken about this on other podcasts as well, which it would be interesting to hear your perspective on.

00:35:38 Sarah Kelsey

So what’s one question that TA leaders have to ask of any vendor when looking to use one?

00:35:45 Martyn Redstone

I mean, there’s a multitude of questions ultimately, but it starts with explainability and transparency.

00:35:53 Martyn Redstone

Especially if you’re bringing on board a tool that helps you make a decision on who’s going to be hired and who isn’t.

00:35:59 Martyn Redstone

We have to start there.

00:36:01 Martyn Redstone

Can you tell me in a natural language way exactly how your system makes a decision?

00:36:06 Martyn Redstone

And then we go on to the repeatability.

00:36:08 Martyn Redstone

Do you see rank drifting, all those kind of things.

00:36:12 Martyn Redstone

But ultimately, there are a load of questions you have to ask.

00:36:15 Martyn Redstone

I’ve got a fantastic free resource around that people can access.

00:36:20 Martyn Redstone

But those are the two most important ones right now.

00:36:22 Martyn Redstone

Because ultimately, as we see regulation coming online here,

00:36:29 Martyn Redstone

EU, US, et cetera, et cetera, we’re gonna see a ramp up of litigation happening and we’re already seeing it in the US.

00:36:36 Martyn Redstone

So explainability around how a decision is made, the lack of bias, et cetera, et cetera, needs to be asked first and foremost.

00:36:45 Sarah Kelsey

Yeah, amazing.

00:36:46 Sarah Kelsey

We’ll put that resource in the show notes for everyone to refer to as well.

00:36:50 Sarah Kelsey

If an auditor walks into a company tomorrow, what sort of thing is going to be a red flag to them?

00:36:55 Sarah Kelsey

What are things that they’re going to be questioning straight away, do you think?

00:36:59 Martyn Redstone

So it depends what they’re looking at.

00:37:01 Martyn Redstone

But ultimately, in things that we’ve covered already on the conversation, they’re going to be looking at whether bias auditing has been done.

00:37:08 Martyn Redstone

They’re going to be looking at explainability.

00:37:11 Martyn Redstone

They’re going to be looking at things like human kill switches around decision making.

00:37:16 Martyn Redstone

So it depends on the regulation, the jurisdiction, the use case, but they are going to be looking that to make sure that the organizations haven’t just blindly implemented AI.

00:37:27 Martyn Redstone

without doing it responsibly and ethically.

00:37:29 Sarah Kelsey

Yeah, this is quite interesting.

00:37:30 Sarah Kelsey

So you were recently on an episode of Matt Alder’s podcast, who we’ve also had on the podcast before, and another incredible thought leader like yourself.

00:37:38 Sarah Kelsey

So you mentioned on that podcast episode that generative AI can create a false confidence and maybe enabling shortcuts as well, which is something you’ve also spoken to in this episode.

00:37:48 Sarah Kelsey

This is where concerns around compliance and governance tend to be most valid.

00:37:52 Sarah Kelsey

Could you tell us a little bit more about that?

00:37:54 Martyn Redstone

Yeah, so

00:37:57 Martyn Redstone

what happened with the advent of ChatGPT was a very powerful AI model, a large language model, had a chatbot put on front of it.

00:38:05 Martyn Redstone

And people started using this chatbot and getting magic out of it.

00:38:11 Martyn Redstone

And it immediately created a false confidence.

00:38:13 Martyn Redstone

You know, I had people saying to me, you know, I know how to program AI.

00:38:17 Martyn Redstone

And I’m sitting there going, you know how to program AI?

00:38:19 Martyn Redstone

Are you a Python coder?

00:38:21 Martyn Redstone

Yeah, how do you do that?

00:38:22 Martyn Redstone

No, I, you know, I talked to ChatGPT and like, well, that’s not programming AI.

00:38:27 Martyn Redstone

And it gives you this kind of, people have been getting this false confidence in their abilities and their understanding.

00:38:34 Martyn Redstone

And unfortunately, you know, I kind of used a term, didn’t use it on Matt’s podcast, but I, again, another analogy, you know, when, you know, thousands of years ago when we sat in our tribes and,

00:38:45 Martyn Redstone

and they had shamans and witch doctors and they were able to make, fantastic smoking illusions.

00:38:50 Martyn Redstone

And they’d tell somebody that it was a deity, that it was a god.

00:38:53 Martyn Redstone

And the people who were less understanding of the world of chemicals and what have you believed it and thought it was God.

00:39:01 Martyn Redstone

And they believed the magic.

00:39:02 Martyn Redstone

And it’s the same in Vegas with the shows and everything.

00:39:04 Martyn Redstone

So ultimately it’s created an illusion of competency.

00:39:10 Martyn Redstone

where people now get great answers out of AI.

00:39:14 Martyn Redstone

They don’t second guess them.

00:39:16 Martyn Redstone

They don’t understand that large language models are text generation.

00:39:21 Martyn Redstone

And so they do hallucinate.

00:39:23 Martyn Redstone

They do make things up.

00:39:24 Martyn Redstone

And so it’s created this kind of false confidence in people that they know AI and they get AI.

00:39:31 Martyn Redstone

And I hear it all the time.

00:39:32 Martyn Redstone

And I just think people just need to temper that a little bit and understand that all they’re dealing with is a chatbot on the front of 1

00:39:40 Martyn Redstone

AI model.

00:39:41 Martyn Redstone

And AI isn’t a large language model.

00:39:43 Martyn Redstone

AI isn’t ChatGPT.

00:39:44 Martyn Redstone

That’s one AI model within a sea of millions.

00:39:49 Sarah Kelsey

It’s quite an interesting.

00:39:50 Sarah Kelsey

space in that sense where it almost feels quite polarizing where some people are so fearful and put off it just at the thought of having to potentially implement AI and other people grasp that but then take it to the point where they’re now a programmer and an expert in it.

00:40:09 Sarah Kelsey

doesn’t seem like there are many people who feel sort of

00:40:12 Sarah Kelsey

In the middle, I’m sure there are, but there are so many people that are either, really worried by it or completely excited by it as well.

00:40:19 Martyn Redstone

Yeah, I see that a lot.

00:40:20 Martyn Redstone

if I’m doing a workshop, at the start of a process, I’ll always take a polling of the room.

00:40:26 Martyn Redstone

where do you think you are from basic to intermediate to expert?

00:40:29 Martyn Redstone

And you don’t tend to get many people saying that they’re intermediate.

00:40:33 Martyn Redstone

So I totally and utterly agree with your statement.

00:40:36 Martyn Redstone

The people who are expert by the end of a couple of workshops usually say, well, I probably think I’m a little bit lower than that now.

00:40:42 Martyn Redstone

Oh, absolutely.

00:40:43 Martyn Redstone

Yeah, absolutely.

00:40:45 Martyn Redstone

But it’s all to do with that kind of forced confidence piece.

00:40:48 Martyn Redstone

So I think so, yeah.

00:40:49 Martyn Redstone

And you’ve got people that say, I’ve tried it, didn’t like it, comes out with generic answers, yeah, those kind of things.

00:40:56 Martyn Redstone

But again, one of the things that I keep pushing back on people is that when you talk about AI, you’re not talking about AI, you’re talking about a large language model, generative AI.

00:41:05 Martyn Redstone

And so people who now call themselves, you know, AI experts, they might be very, very good at prompting a large language model, but that’s it.

00:41:14 Martyn Redstone

I wouldn’t really call them an AI expert.

00:41:15 Sarah Kelsey

Yeah, it’s almost one of those things as well where the more the less Absolutely, absolutely.

00:41:21 Sarah Kelsey

So hopefully we can create some common themes and simple ways to get started.

00:41:27 Sarah Kelsey

Another thing that there is a lot of noise around is these AI laws.

00:41:31 Sarah Kelsey

So help us cut through the noise here.

00:41:33 Sarah Kelsey

What actually is landing in the next 12 to 24 months that our listeners need to be aware of?

00:41:39 Martyn Redstone

So less than 12 months time, so next August, August the 2nd,

00:41:43 Martyn Redstone

The high risk part of the EU AI Act goes live.

00:41:47 Martyn Redstone

And what that means is anything to do with life-changing decision making, so employment, healthcare, et cetera, et cetera, has to now be compliant to those laws.

00:42:00 Martyn Redstone

And we’ve covered quite a lot of it, you know, explainability, transparency, bias audits, making sure that you’ve got the right documentation, annex 4 documentation.

00:42:08 Martyn Redstone

So to cut through the noise, you know, you’ve got less than 12 months,

00:42:12 Martyn Redstone

unless you’re listening to this in 12 months time, then you haven’t got any time at all.

00:42:16 Martyn Redstone

But August the 2nd, 2026 is when you need to be ready to go, because there are going to be people waiting to jump up and down from a litigation perspective.

00:42:27 Martyn Redstone

So that’s in Europe.

00:42:28 Martyn Redstone

And what that means, and it’s something to remember, because I have this conversation with especially some of my peers in the US, is that it doesn’t mean that if you’re based in Europe, it applies to you.

00:42:39 Martyn Redstone

It means that if you are,

00:42:41 Martyn Redstone

If your AI is in the European market, it applies to you.

00:42:45 Martyn Redstone

It is a market access based, supply chain based piece of litigate, a piece of legislation.

00:42:54 Martyn Redstone

So if you are putting AI into the market, if you’re supplying it to people in the market, if you’re putting it out for public to use in the market in the Europe, it’s very similar to GDPR.

00:43:04 Martyn Redstone

You need to be compliant.

00:43:05 Martyn Redstone

So that’s here.

00:43:07 Martyn Redstone

UK, Next 12,

00:43:10 Martyn Redstone

12, 24 months, we’ll just start seeing a movement of some kind of AI legislation coming through Parliament here.

00:43:20 Martyn Redstone

We’ve already seen, you know, already seen the start of that going through the House of Lords and Lord Holmes putting out a piece of work on that.

00:43:28 Martyn Redstone

So I expect to see more around that over here in the UK.

00:43:32 Martyn Redstone

In the US, probably more state-based law coming out.

00:43:35 Martyn Redstone

We’re seeing that a lot more now.

00:43:37 Martyn Redstone

and probably another change of mind around federal law as well.

00:43:40 Martyn Redstone

We see that quite a lot.

00:43:42 Martyn Redstone

So it’s a bit messy over there in the US.

00:43:43 Martyn Redstone

And again, across kind of other jurisdictions as well, we’re seeing a lot more movement in South America, a lot more movement in Asia and the Middle East around upcoming legislation.

00:43:55 Martyn Redstone

So I think next 12, 24 months is going to be a really exciting time with a lot of legislation.

00:43:59 Martyn Redstone

Stressful, but exciting because there’s not going to be a standardization.

00:44:02 Sarah Kelsey

Yeah, time to get those guardrails.

00:44:04 Sarah Kelsey

Absolutely.

00:44:04 Martyn Redstone

Yeah, absolutely.

00:44:06 Martyn Redstone

Absolutely.

00:44:06 Sarah Kelsey

And it seems, that the UK hasn’t gone to some of your points quite as far as an act yet.

00:44:13 Sarah Kelsey

Instead, we’ve got the ICO guidance and principles-based approach.

00:44:16 Sarah Kelsey

So would you recommend, even if we might not have products in market in the EU, that we sort of look to that as a gold standard for now?

00:44:24 Martyn Redstone

Absolutely.

00:44:25 Martyn Redstone

Yeah.

00:44:25 Martyn Redstone

I think that whilst a lot of people think it’s quite draconian, and it is, you know, I’m personally quite opposed to

00:44:32 Martyn Redstone

the act because I think it’s overly legislative.

00:44:37 Martyn Redstone

However, it’s there for a reason.

00:44:39 Martyn Redstone

And even with my kind of basic research that I’ve done shows that it’s not trustworthy using AI, certain AI models to make life-changing decisions.

00:44:49 Martyn Redstone

And so it has a purpose.

00:44:52 Martyn Redstone

And absolutely, I think that it is the best practice gold standard that you can aim towards.

00:44:59 Martyn Redstone

And if you cover yourself from the EU AI Act,

00:45:01 Martyn Redstone

to be covered everywhere else in the world.

00:45:03 Sarah Kelsey

Okay, cool.

00:45:04 Sarah Kelsey

That sets some really useful places to start.

00:45:07 Sarah Kelsey

And so do you expect that the UK will follow suit and introduce something stronger than just this guidance?

00:45:12 Martyn Redstone

I don’t know.

00:45:14 Martyn Redstone

Pre-last year’s general election, the previous government were very pro-innovation and they were very much, you know, when they were looking at bringing in some AI-based legislation, it was very pro-innovation.

00:45:30 Martyn Redstone

With the current government, I don’t know.

00:45:32 Martyn Redstone

I think what we’ll see is something leaning more towards heavier regulation.

00:45:37 Martyn Redstone

You know, I’m hoping that with the recent visit from, you know, tech CEOs from the US putting in a lot of investment into the UK, I’m hoping that it alleviates any future over-regulation of AI.

00:45:54 Martyn Redstone

But it’s still very open.

00:45:55 Martyn Redstone

I’m hoping that it kind of sits somewhere in between Europe and the US from a legislation to an innovation perspective.

00:46:03 Sarah Kelsey

Okay, excellent.

00:46:04 Sarah Kelsey

Makes complete sense.

00:46:05 Sarah Kelsey

Thank you so much, Martin, for your thoughts on everything there.

00:46:08 Sarah Kelsey

It’s really given us some solid places to begin.

00:46:12 Sarah Kelsey

Before we finish today though, we’ve introduced a new quick fire round to the deep dive.

00:46:17 Sarah Kelsey

And we’d love to wrap up with some key takeaways for our lovely listeners.

00:46:20 Sarah Kelsey

Is that okay with you?

00:46:21 Martyn Redstone

Yeah, absolutely.

00:46:22 Sarah Kelsey

Okay, let’s get into it.

00:46:24 Sarah Kelsey

What’s the biggest myth about AI in HR?

00:46:28 Martyn Redstone

Is the AI in HR shouldn’t happen because people like dealing with people.

00:46:33 Martyn Redstone

I think that’s the biggest myth.

00:46:34 Martyn Redstone

People don’t, people like dealing with things that provide them with immediacy, convenience and agency, which is a machine.

00:46:40 Sarah Kelsey

Are organizations ready for what’s coming in the next decade or so?

00:46:44 Martyn Redstone

No, And I’ll hark back to what I said.

00:46:47 Martyn Redstone

there’s a lot of work that we need to do around collecting data, understanding the world of work, understanding how to break down jobs into tasks.

00:46:56 Martyn Redstone

You know, what tasks can be done by human, what tasks can be done by machine, and what tasks can be done by human and machine team.

00:47:01 Martyn Redstone

And I think until we get that ready, that data ready, then no, we’re not ready.

00:47:05 Sarah Kelsey

Yeah.

00:47:07 Sarah Kelsey

Adding on to that, from your perspective, it sounds like, especially for larger enterprises, a way to sit ahead of the curve at this point is actually to start implementing those guardrails, because…

00:47:18 Sarah Kelsey

No one else seems to be doing that from your perspective.

00:47:20 Sarah Kelsey

And if they are, it’s retrospectively.

00:47:22 Martyn Redstone

Exactly.

00:47:23 Martyn Redstone

And that’s the first point of readiness from an organization.

00:47:26 Martyn Redstone

But I think to answer your question, we’re nowhere near ready for what’s coming in the next decade.

00:47:30 Martyn Redstone

No.

00:47:30 Sarah Kelsey

Okay, very interesting.

00:47:32 Sarah Kelsey

What percent of HR processes should be automated and why or why not?

00:47:37 Martyn Redstone

Percentage is a difficult one.

00:47:39 Martyn Redstone

I think anywhere between 40 to 80%, depending on the organization.

00:47:43 Martyn Redstone

But I think anything that involves repeatable

00:47:47 Martyn Redstone

mundane robotic processes, service desk automation, transfer automation, holiday request automation, all those kind of things.

00:47:56 Martyn Redstone

I think the world is your oyster when it comes to automating in HR, but doing it responsibly and ethically.

00:48:05 Sarah Kelsey

Makes total sense.

00:48:07 Sarah Kelsey

What’s one process you’d never automate?

00:48:09 Martyn Redstone

I think going back to what I said, I wouldn’t automate the final decision making in a process that involves

00:48:16 Martyn Redstone

deciding on somebody’s employment.

00:48:18 Sarah Kelsey

Are there any others that you would never automate?

00:48:20 Martyn Redstone

I think that there’s certain instances where you might have something automated, but you should hand off to a human pretty quickly.

00:48:27 Martyn Redstone

So, you know, in the drastic situation where, you know, somebody is contacting an organization to let them know that a relative of theirs who’s an employee of the organization has had a terrible accident, you know, potentially has died, you know, I think that if you’ve got an automation around

00:48:46 Martyn Redstone

around notification, you probably need to have a safeguard in place that goes over to a human pretty quickly when it comes to that kind of thing.

00:48:55 Sarah Kelsey

And complete the sentence for me.

00:48:57 Sarah Kelsey

Chatbots fail when?

00:48:59 Martyn Redstone

Not designed properly.

00:49:00 Sarah Kelsey

Number 6 is what percentage of companies would be in breach of compliance regulation if the rules came into force in the EU and US tomorrow.

00:49:09 Martyn Redstone

I have the vast majority of them.

00:49:11 Martyn Redstone

I mean, I dread to think, but I think we’re talking

00:49:15 Martyn Redstone

In the 90s percentage, absolutely.

00:49:17 Sarah Kelsey

Wow, really.

00:49:19 Sarah Kelsey

It’s, yeah, it’s so interesting, isn’t it?

00:49:21 Sarah Kelsey

Because so many people, maybe it seems like people are suffering with that analysis paralysis thing as well, needing to implement, but not knowing how, and so not doing anything.

00:49:31 Martyn Redstone

Exactly, yeah.

00:49:33 Sarah Kelsey

The last one, what’s one feature you’d add to every ATS if time and budget wasn’t an issue?

00:49:39 Martyn Redstone

I think the one feature I would add to ATSs is

00:49:44 Martyn Redstone

candidates self-serve.

00:49:47 Martyn Redstone

I always hark back to e-commerce principles when it comes to thinking about the candidate experience.

00:49:53 Martyn Redstone

And one of the great things that I love about buying online is that at any point I know where my order is.

00:50:00 Martyn Redstone

I know whether it’s still in the warehouse.

00:50:01 Martyn Redstone

I know when it’s being delivered to me.

00:50:03 Martyn Redstone

I get fantastic updates by text message.

00:50:06 Martyn Redstone

I don’t see why we can’t do that.

00:50:07 Sarah Kelsey

Yeah, I love that idea.

00:50:09 Sarah Kelsey

It’s really, really interesting.

00:50:11 Sarah Kelsey

Well, thank you so much for your insight and your time today, Martin.

00:50:13 Sarah Kelsey

My pleasure.

00:50:14 Sarah Kelsey

It’s been so great to have you here.

00:50:16 Sarah Kelsey

Our listeners are very lucky.

00:50:17 Sarah Kelsey

We feel very lucky to have your insights.

00:50:18 Martyn Redstone

Thank you.

00:50:19 Sarah Kelsey

We’ll leave all of your details down below, but for the listeners right now, where can they find you if they want to hear more?

00:50:24 Martyn Redstone

LinkedIn is always the best place to find me.

00:50:26 Martyn Redstone

Martin with a Y, Martin Redstone.

00:50:28 Martyn Redstone

If you can’t find me there, then I’ve disappeared off the face of the earth, ultimately.

00:50:32 Martyn Redstone

So come and follow me.

00:50:34 Martyn Redstone

Come and connect with me on LinkedIn.

00:50:46 Sarah Kelsey

If you found today’s conversation valuable, you can find related articles in the show notes below or previous episodes wherever you get your podcasts.

00:50:53 Sarah Kelsey

Like, follow, and subscribe, and we’ll be back soon with more great industry insight on the Deep Dive.