What does AI actually do to the people building the physical world around you?

In this compilation of Why Design, ten hardware practitioners share the belief that sits at the heart of good AI use: it is a tool for someone who already knows what good looks like, not a replacement for knowing.

Rather than accepting AI hype at face value, these guests interrogate it from inside real hardware work: robotics, medical devices, industrial design, investment, and recruitment. That interrogation led to some of the most honest, occasionally contradictory, moments the show has recorded on the subject.

This conversation is not about whether AI is good or bad. It is about who stays in the loop when it is used well, and what gets lost when nobody does.

Don't just listen. Go beyond the podcast.

Join the Why Design community → teamkodu.com/whydesign

What You'll Learn

Why an SOSV investor believes AI is not like any other tool, and what a hammer teaches you that AI does not

How a field robotics team turned every operational anomaly into training data instead of a disaster

Why one engineer's rule is: if you can write the problem in a prompt, you have already solved it

What a real executive AI mandate looks like from inside a design team, with no training and no plan

Why large language models are structurally built to find the average, and what that costs inclusive design

Why the expert-in-the-loop principle is the one rule almost every guest agrees on

Memorable Quotes

"It fundamentally makes you a little bit dumber when you use it most of the time."

"My problem is it's scary believable rather than scary accurate."

"If I can write it in a prompt, I've solved the problem."

"That's a terrible creative brief."

"LLMs are just massive heuristic engines. They're like a pachinko machine on the scale of the solar system."

Resources & Links

🎧 Listen on Spotify, Apple Podcasts, YouTube & Amazon → whydesign.club

👥 Join the Why Design community → teamkodu.com/whydesign

📸 Follow @why.design_ on Instagram

🎥 Watch full episodes → YouTube.com/@whydesignpod

🔗 Follow Chris Whyte → linkedin.com/in/mrchriswhyte

About the Episode

Why Design is powered by Kodu, a specialist recruitment partner for the hardware and physical product development industry.

Through honest conversations with designers, engineers and creative leaders, we explore not just what they build, but why they build it: the beliefs, decisions and responsibility behind meaningful work.

About Kodu

Why Design is produced by Kodu, a recruitment partner for ambitious hardware brands, design consultancies and product-led start-ups.

We help founders and leadership teams hire exceptional talent across industrial design, mechanical engineering and product leadership, bringing structure and clarity to one of the hardest parts of scaling.

🔗 Learn more → teamkodu.com

This episode is sponsored by Vax and TTi Floorcare. Kodu, is running the hiring. Recruiting for Vax for more than ten years, placing over fifty people including a VP of Engineering and a Head of Design. We know how they build. Right now they’re rebuilding their UK engineering and innovation function from the ground up: advanced development, design engineering and industrial design, all at once. We haven’t seen them move at this pace in a decade of working with them.

The practitioners on this show are building hardware.

2

Robots, medical devices, lab automation, electric vehicles, consumer products.

3

They are using AI.

4

Some of it is genuinely changing what is possible.

5

Some of it is not working the way anyone says it is, and most of them have a sharper read

on both sides than anything you'll find in the mainstream conversation.

6

This is the Y design AI compilation.

7

AI fundamentally makes you a little bit dumber when you use it most of the time.

8

And other tools do not.

9

If I pick up a hammer, there are many learning opportunities with a hammer, especially if

I'm unskilled with a hammer, right?

10

And you can say that AI might have some learning opportunities.

11

The problem is you don't feel any pain except for occasional rage.

12

Where AI genuinely changes hardware work.

13

Set aside the hype for a moment.

14

Three guests on moments where AI actually moved the needle on a physical product in ways

that would not have been possible in any other way.

15

Well, it's something very, very close to our hearts, because um our early prototypes, and

I I think this might be the same for many businesses.

16

It's our early prototype was very hard coded.

17

So you take it into the the operational environment, and every time it hits an anomaly or

an edge case.

18

You've got to write another piece of code and you've got to reconfigure the stack.

19

And so every every error is is uh a real disaster.

20

And um as you move to learning systems and generative AI, um, then and I think this is a

journey that we're on.

21

every edge case becomes an opportunity, a learning opportunity, right?

22

So you you get to the point where

23

You're deploying a pretty dumb product and it it's learning and you can put AI across the

whole stack.

24

So we've got robots operating in in the fields and we can put it we can we can put it we

can put AI into the perception stack and detecting raspberries and ripeness, or we could

25

put it into the how it moves and how it navigates around obstacles.

26

We can put it in many different l layers and and then

27

E everything is an opportunity.

28

Every time you pick and you hit a an a new a new variety or a new um environmental

condition, you learn, you share it with your the rest of the robots and it improves.

29

And uh that's what I find really exciting about where where AI is going at the moment.

30

You know, I gotta tell you I trusted the people that were on the decisions of what we

needed.

31

So the first thing I did was a crash course and

32

I would write down everything they'd say I'd say, Hey, we need, you know, a Raspberry Pi

board.

33

We need this, you know, uh we need S DK hard, you know, all software written and all this

stuff.

34

We need, you know, all these, you know, accesses and and they would give me this

description and the first thing I would do is jump online and go, Okay, what does that

35

mean?

36

Right.

37

And I would study what does S DK stand for?

38

So I could put it in my terms to say, Okay, next time I'm on a call and somebody says, I

need an S D K okay, well that means software development camp, right?

39

Well, I didn't know that six months ago.

40

But I had to learn those terms in order to talk to my engineers, talk to my, you know, my

IT guys, right?

41

So when I went to CES, I had this knowledge of something, right?

42

So I could talk about, hey, we're using Linux and Raspberry Pi and Java and it's on an S D

K and we're gonna drive you know, and I started to become the project, right?

43

So my whole entire life was, you know, not about anything like that.

44

So I really had to do a lot of crafts course studying.

45

Um

46

And and thank God for the internet.

47

I mean, I was able to to to really go through and say, okay, what does this mean and and

what's going to be the hurdle?

48

Um, you know, somebody asked me a few months ago, what do I use AI for?

49

Um, I use AI to learn, uh, to go through this and say, what are the two technicals that I

may be missing?

50

And it's very um it's very nice to have a second view and share with me this is why it's

not gonna work, or this is why it's gonna work, or this is what you need to do to make it

51

work.

52

yeah, there's there's a lot of figure outable stuff now, you know, with AI, you know, as

long as you can you know roughly where you're going and what the output needs to look

53

like.

54

Yeah.

55

I had coffee with a industrial designer earlier this week and he was telling me about a a

a project that he was doing as a buzz project and he was programming the circuit board.

56

He was doing the firmware and the software, getting all the motors to work.

57

And he never touched that in his life, but because he knew what he wanted the the thing to

do, he knew how to articulate that to the to Claude to get it to run the code on its side.

58

Well I also worked with coders and things like that.

59

Yeah.

60

And my job is to put all the pieces together to get to that.

61

Yeah.

62

And I just think time and the international experience and being able to work across

different kind of media.

63

I mean, I'm personally comfortable in the industrial design medium.

64

the research.

65

I'm also, you know, full-on photographer, soulmaker as well.

66

I'm kinda comfortable with all of those things.

67

There's a lot of stuff you've figured out.

68

Now as a director, I'm able to to be a little bit more involved in some business

decisions, both tactical and strategic.

69

I'm very excited about that.

70

Um particularly strategic part obviously because that's my one of my strengths.

71

and

72

I'm also particularly excited to see how the things have implemented so far and the design

team are going to to come together.

73

You know, I I I had a big vision when I started, but you put all these things in place and

then you hope that it comes together.

74

And and I think it is.

75

Um the fact that I am now spending a little bit less time, I hope, um to to enjoy seeing

my team taking a

76

you know, stepping up and and being able to to create really exciting uh products and

innovations that we can be really proud to present to the world in the coming years.

77

So those are the the two main things.

78

It's like the scenes I've planted, how are they gonna uh turn up and the things I'm

learning that are ahead of me and

79

How I can learn but also influence and that excites me a lot.

80

The honest critique.

81

All three of those examples are real, but the same guests and others are equally direct

about where the technology falls short and why the easy comparisons to other tools do not

82

quite hold up.

83

I've not found LLMs particularly transformative because a lot of what I need to do, I need

to do the process.

84

To solve the problem.

85

Mm-hmm.

86

If I can write it in a prompt, I've solved the problem.

87

But it's harder to do it in text for me than it is just to go and build it.

88

You know, good designers are just people who can articulate why that quality's good to

someone who maybe can't can kind of sense it but can't see it in definitive terms.

89

I just don't see how an L L M's going to do that.

90

I I worry on being no, that AI I'll never take my job.

91

AI can't find a bloody invoice in my inbox at the moment.

92

We're we're still finding our feet with it in terms of where it fits and why and yeah,

like you said, how you how you control it to make sure you get the output and steer it in

93

the process.

94

But yeah, I it it's like two things at the same time for me.

95

It's like near term, like you said, how you leverage it from like a prompting generation

supporting process point of view.

96

And then longer term, like it's that's this anticipation that, yeah, say it becomes a

gentic and

97

can work 247 and like run tasks without cont continuous input or like supervision, what

that looks like is like a whole I think at that point it also it becomes like almost

98

collaboration potentially as well.

99

It's hard to see and like hard to like know, but like if it genuinely gets to like AGI

level intelligence, then then yeah, there's no reason why I wouldn't become more of a

100

collaborator.

101

And that's gonna be a whole other thing.

102

And again, I just make I just feel quite thankful

103

we're not a huge company 'cause I think it's gonna be massively disruptful t basically the

bigger you are.

104

Uh and I and I do believe that like design will exist beyond where AI's going.

105

Like I just don't think we'll it'll change.

106

But I I'd be more worried if I was in like digital design or UX UI.

107

And I think you look back over here like the physicality and meaning we put into objects

has existed since we've existed.

108

And I don't think we're suddenly gonna

109

not care about human meaning and how that's represented in three-dimensional form.

110

So yeah, personally not worried, but I would probably uh arguably be more worried if I was

more digitally leading.

111

But my my problem is it's scary believable rather than scary accurate.

112

Like that that's the thing that is totally annoying and also the reason why it's

ultimately pretty much a dead end.

113

Um and and we are we are gonna have to do a lot to improve AI uh for more general use.

114

Um it turns out though that people are enormously bullshit tolerant, which is unfortunate,

right?

115

Because as Yval Harari would say, it's really cheap to make up bullshit.

116

It's really expensive to discern truth.

117

And and AI is really expensive way of

118

of getting a sort of seventy thirty bullshit truth mix.

119

No, I mean listen, controversial is the I mean, listen, what what we need to be doing is

we need to make sure that we don't leave it up to AI to do solve all of our problems.

120

So I think we're gonna come, you know, the the younger generation shouldn't just uh phone

it in as we say.

121

That's probably an an out of state reference as it is, you know?

122

Or chat GPT it or you know.

123

They shouldn't they should make sure that they um they do the work because that's what

design is about.

124

And if you're using a large language model, no matter how good it is, it's gonna find the

average.

125

It's gonna find the safe thing.

126

So in fact, I can tell you specifically a uh reference to something that we just had a a

really interesting conversation.

127

So we talked about the founders of IDO, the current managing director of IDO.

128

Mike, uh I met in Athens, Greece, of course of all places at a uh forum that I was at and

I was speaking at called the House of Beautiful Business.

129

So uh and we were the topic was AI.

130

And so um the conversation was exactly that.

131

Because of large language models are gonna it's all about the data, right?

132

They're gonna find sort of the average of averages.

133

So if you look at it as a s as a curve, so a little bit of geeky mathematics at this

point, right?

134

So it's it's about the AI is gonna be in the middle, you know.

135

The bell curve is here, it's gonna probably be within that space.

136

But as designers, what we wanna be working for and designing for are the folks outside.

137

So the outliers.

138

So what so if we're talking about inclusive design, it's never the people in the middle,

it's about the outliers.

139

Right.

140

And so that's what we want to be working for in terms of solutions as well, right?

141

So that's where AI can does not go to do a good job right now, right?

142

Because they can't.

143

They can't.

144

It's they're not it's not in the model.

145

Yeah.

146

It's not there.

147

So as designers, we need to make you know, we need to make sure that the younger

generation doesn't phone it in, doesn't prompt it, but actually does the work themselves.

148

You know?

149

So

150

That's what I would think we should do.

151

So in terms of a design, the other thing that I think we should do is do a better job of

collaboration.

152

Yes, we're doing better than we did, but we still got a long ways to go.

153

You know, LLMs are just massive heuristic engines.

154

They they they just predict the next token with some possibil probability.

155

They're like a pachinko machine on the scale of the solar system.

156

If you've ever seen those Japanese gambling machines with the little balls that come down

around the pins.

157

Well imagine that the size of a solar system and you're kind of at the right scale for

just how many pathways the little balls can go through in a large language model.

158

Well, it's not only bonkers, it's impossible to figure out how any particular conclusion

was reached.

159

Like it works well enough that humans are kind of

160

Pleased by this toy?

161

This episode is sponsored by Vax and TTI Floorcare.

162

And Kodu, my company, is running the hiring I'm about to describe.

163

So you know exactly what this is.

164

This is why I wanted to make it anyway.

165

I've recruited for Vax for more than 10 years.

166

Over 50 people, including a VP of engineering and a head of design.

167

I know that building well.

168

Right now they are rebuilding their UK engineering and innovation function from the ground

up.

169

Advanced development, design engineering and industrial design all at once.

170

I've not seen them move at this pace in a decade of working with them.

171

So I'm going to tell you what is actually happening and you can decide.

172

The bad mandate problem.

173

Problem is not just that AI has real limitations.

174

Is that executives are driving AI mandates through organizations without a clear brief.

175

Two guesses on what that looks like from inside a team.

176

Yeah, or but in within this year we're gonna get rid of 8,000 um team members and you're

gonna choose 15% of your your team who's gonna be let go and move a place with AI.

177

It's like, well, that's a terrible creative brief, you know.

178

And again, coming back to design thinking, like what are the constraints?

179

I have to lay off fifteen percent of my team, that's the that's the constraint, but for

what purpose?

180

Like what what what are the tools?

181

So that's that's really what I think this is challenging design leaders and managers to

think about.

182

But actually everything else is constantly being updated.

183

Or someone is just slapping AI onto something badly considered.

184

You're using up someone's change tolerance and educational sort of bandwidth on stuff that

you had no plan to do.

185

So actually just trying to protect the core of what we're doing.

186

Um, which hasn't really changed that much, but still be able to communicate with each

other, file stuff away in the same way.

187

It's really basic stuff.

188

But it seems to get harder and harder with all these new tools in a weird way.

189

I'm worried I'm just turning into a grumpy old man.

190

AI in hiring.

191

This is a topic close to home.

192

The same AI wave is reshaping how people are hired and how they apply.

193

And in the hardware sector specifically, the tools have a problem.

194

It goes way beyond bad C Vs.

195

I imagine AI is making things worse as well.

196

I've seen your AI job descriptions and then people using AI to apply and write their C Vs.

197

It's just like a AI's um, you know, it's a great great tool.

198

Um

199

You know, I've I use it every day.

200

I've used it to prepare my notes for for our conversation today.

201

I use it to summarise um, you know, uh profiles of candidates that I then go and talk to

my clients about.

202

But it's it's a tool and it's a it's a guide.

203

I still deliver it kinda as me.

204

And when I I I get emails from people at all levels and messages and it sticks out like a

sore thumb when they've used AI to craft it because there's there's tropes, there's um

205

things like

206

Your um your work as a as a founder at Kodu is is exemplary and it's all just like no one

talks like that.

207

No And there's double well the the one the classic one is always like it's not just this,

it it's this.

208

Like that would I want to decide.

209

It's like the the Mark Spencer's advertis, it's always like, you know, this isn't just MS,

this is blah blah blah.

210

Um that's just it all the time.

211

But I I also found as well, like I I'm the same, like I you know, I've written a few blog

posts for automata.

212

And I write them myself first before then passing them through.

213

And then I even end up having to like edit what it does, even though I'm very quite quite

strict on what I tell it to do and not do.

214

'Cause you still want it to sound differentiated.

215

I I've I I struggle when I go on LinkedIn now 'cause all the posts are in by AI and they

all read the same and sound the same and nothing stands out.

216

But then also I find if you then become too reliant on it, you lose the sharpness and

you're n then to be able to absolutely write or or think creatively.

217

Um

218

Yeah, we're gonna put ever tangent there.

219

But yeah, I think the experts in the loop view.

220

So when does AI work?

221

The consensus, spoken or implied, across every guest on this topic is the same.

222

You need someone in the room who knows what the answer should look like.

223

Then it then it's a tool.

224

Because there's at least one other human with subject matter expertise in the loop.

225

Yeah, then it's then it's a potent tool.

226

i uh it should come with a safety warning, don't go in alone.

227

'Cause 'cause loneliness, by the way, is one of the great killers.

228

You know, it it it's up there with not washing your hands to kill you.

229

like loneliness will do more harm to you than than a high fat diet and a a drinking

problem put together.

230

Across the board and probably more senior, I think it wouldn't hurt if you were like more

of a specialist than a generalist.

231

And that might be different in different businesses, but I feel like AI is gonna become

that like

232

generalist layer that's connecting almost like the connective tissue between everyone.

233

But it's not I think one thing it's definitely going to struggle with nearer term is like

the physicality of product development.

234

Like if you're great at hands on model making, prototyping, testing, or the end of the

process mechanics, like it's not gonna tap into that near term.

235

I think we the if it to answer in like how we're shaping up is like I do have a bit of

like a AI strategy

236

that that kind of is known between the team here where we kind of accept that like from an

intelligence level it's going to become a layer, it's going to become part of our process

237

of where we check ideas against or run ideas through and it and it will integrate with

everything.

238

But the way I'm setting it up, and I think it's fortunate we're a smaller team, is I'm

almost different people are going to act as different heads of different things.

239

Again, where they have a specialism, whether that's visualization or DFM

240

creative judgment sort of design language, uh community events, like we'll leverage it,

but there will be different owners that like will almost own well, I perceive eventually

241

will be the normalization of like agents that work for them.

242

So I think in a in a in a junior level, I think that means you're gonna have to be way

more comfortable with making decisions and delegating, which sounds weird to do when

243

you're like new to something, but I think

244

With AI, that's gonna have that's gonna be important.

245

So I think having like self-confidence that you can give direction and also make the right

decisions and show that you're capable of making the right decisions will become more

246

important.

247

That's not really been a junior skill before.

248

Um and then another reflection on that particularly from a design point of view is that I

find with a lot of juniors you have to like if you're designing for a certain company and

249

a certain design language and identity.

250

It might differ from what you actually like.

251

And you saw it takes a while to get people to be able to switch off what they like at a

moment of designing something so that they can kind of evaluate what they're designing for

252

that brand or against that identity independently of what they like.

253

And it takes like it can take years for someone to sort of make that journey where they're

like they're judging an idea as an idea and not holding on to it as their idea and pushing

254

it because it's their idea.

255

It's like a it's like

256

diminishing that like ego sense of self with the idea.

257

And I think that if you can show that as a junior, that'll be a good skill set.

258

Because again, if you're going to have to give direction and make decisions, being able to

take yourself out of it and be able to see the and understand the bigger picture, there's

259

a certain like maturity to that.

260

They're all quite difficult things because I think if anything, it you can already see

this in how like what's happening with

261

where jobs are shifting towards senior and away from junior, that they all tend to be like

psychological traits that lean towards people that have that have a certain maturity, that

262

have had a certain amount of life experience and then that benefits that.

263

But at the same time, you can have someone who's young, who's had a lot of life experience

and they've learned a lot and they can be mature for their years.

264

So I think that that maturity probably be become more important rather than less

important, I think.

265

Yeah, I think it's really about like how do you I this particular course is like for

design leaders is really to teach designers how to make confident decisions about AI.

266

So like it it you know, it's like your current leading designers or design initiatives,

you're tired of all this hype, you want honest practical guidance, you need frameworks

267

that can help you evaluate all those different opportunities, not just tutorials.

268

And like how do you make smart decisions about AI when you're not an engineer?

269

Right.

270

When none of us can predict the future.

271

Um, and so how do you how do you learn this?

272

And and that's really, you know, how do you how do you know what's hype versus what's

real?

273

How do you know what cost structures are?

274

You know, when should you say yes?

275

When should you just say no?

276

Like these are some of the things that I think are really the most important.

277

Uh it's funny.

278

I I'm a really strong believer of um being in.

279

So so um maybe I got the like the

280

post-COVID piece of being like, my God, working from home is a terrible idea.

281

But in in products in particular, so I think you're definitely right.

282

The the tech roles, the um kind of ancillary roles, a lot of them can be done remotely.

283

That's absolutely brilliant.

284

It makes such like massive sense.

285

I think one of the key learnings for me as a first time founder and a first time founder

in mental devices is this piece of building products.

286

And having your hands on the products as much as possible and having as many of the

engineers knowing, you know, what goes where, which glue was used.

287

Was it UV, was it epoxy, was it was it and and so if you lose that piece and only have one

or two people that can do that, you're very vulnerable actually.

288

So I think I I'm I totally agree with you.

289

There's there's a real benefit to it.

290

but it is

291

And I I actually think it's why medical devices is getting harder and harder, because more

and more people are looking at roles that can be remote and and how they can learn that.

292

But in Ireland actually, Anna, this is a very large segue, but the um applications for

healthcare based um courses in university have gone up by threefold or something.

293

This this because

294

I I think it's it's obviously kids seeing that like AI is taking a lot of grad jobs.

295

So now I think you're gonna see people kind of moving back to physical world jobs.

296

So that's ten guests across ten episodes.

297

All of them are working in or around physical products.

298

All of them are linked below.

299

If any of these moments resonated, the full episodes go much, much deeper.

300

Thank you for listening.

301

Thanks for listening to Why Design.

302

If this episode gave you some new ideas, share it with someone in your team who'd find it

useful.

303

And follow the show so you don't miss what's next.

304

And if you're building a product team or looking for your next challenge in design or

engineering, that's exactly what we do at Codu.

305

We help companies hire smarter and faster, and help great talent find work that actually

fits.

306

I'm Chris White, connect with me on LinkedIn and let's keep raising the bar for design

leadership.

307

Until next time.

308

Stay curious, stay creative and keep asking why design.