Why Platform Engineering is the New Bedrock for the Agentic | The Platform Engineering Show Ep 10

Hey everyone, it's Alan Shimmel from Techstrong. Welcome back to another platform engineering show. org community.

Luke and my friend. It's been too long. Too long.

It's been too long. We've been busy though. Yes.

How, how have you been? How are you doing? Well, you've been, I've been good.

I've been, you know, I've been running around. I, I was, I was in a, you know, don't, don't cry for me, right. I had to go to Napa Valley tour conference, drink some wine, do some videos.

I was in Houston for a, a risk operations conference. I was in Austin. I, I've been off, I've been traveling a bit, you know, busy and, and AIing the whole time.

Luca, this AI thing is just, yeah. Crazy. Yeah.

What about you, my friend? Where are you, VIN? Yeah, same, same busy.

Q4 is here is upon us. Uh, all the events, all the things. Um, we, we hosted a platform called Paris together with, uh, we scale our partners and friends, um, last week.

And it was awesome. It was, it went really, really well. Um, there was way too much eating and drinking involved, uh, on my end.

Always embarrassed. It's been full week there. Yeah.

Yeah. But it was awesome. And actually, look, uh, the, the, the CEO of we scale, uh, he told me like, oh, there's a little surprise when we closed the event.

So him and I did the, the closing panel and then there was, uh, literall champagne and full gra for everybody for like an hour and a half was, was pretty epic. Wow. Oh, that's epic.

Epic. Yeah. Alright.

Besides the food and drink, Luca, what, what was the learning? What was the, the, you know, curriculum over there? Yeah, well, so we, we kind of kept the structure similar.

Uh, if you remember, if people remember to the ones that, that we did in London and New York. So we kind of had this like main track. That was the main stage.

Um, we had Kelsey Tau, we had Gregor Opa, we had incredible speakers. We had a lot of, also, also like real case studies from enterprise practitioners, from very large French companies, for example. So it was super cool.

The, the content was awesome. And then we had this combination on the side of trainings, round tables, workshops, um, that also worked really, really well. Uh, so the content program was awesome.

Um, in terms of, you know, like the main themes throughout, it's kinda like the usual suspect. You mentioned AI before. Um, there was a lot of chatter about kind of the intersection of AI and platform engineering and how do we think about that?

And I think similar to what we saw in London and New York in June this year, there was really, I think it people were really highlighting this kind of dual role of platform engineers as both consumers of ai, of course, uh, like everyone else. Um, but also as key enabler, um, of ai. Yeah.

Um, and in fact we were discussing, um, there was this, um, uh, let me actually pull up real quick because there was this, uh, uh, a recent report by, um, garner, um, and 2025 strategic transient platform engineering. And they were predicting that by 2028, 60% of developer interactions with platform engineering services will be via conversational AI and AI agents. Also, 60% of internal developer platform will include AI agents.

And then 30% of internal developer platform users will have given agency to AI agents for specific steps. So really, and, and, and so you, that was kind of the, the first bit, right? Of like, you're really seeing AI being infused across platforms all over the place.

Obviously these are predictions, you know, 2028. But I mean, another thing that we discussed at the conference was this original prediction. I dunno if you remember, but we, we discussed a couple of times, um, I think this was back in 2023 when Garner said, by 2026, 80% of our, of enterprise organizations will have some sort of platform engineering, um, thing in place.

And I remember at the time when it came out, it seemed very optimistic, very aggressive. Uh, but I think we're in track. In fact, I think Dora, the new Dora report came out Yes.

Just a couple weeks ago saying like, well, actually we're seeing 90% of enterprises have some sort of platform already in place. So we even beat the original forecast of mm-hmm. Garner.

Now, I think like that forecast was a little bit more solid in terms of foundation that this like AI infused, um, RNs, like there's not that many data points that they can draw this from, but still, I think, you know, very, very interesting. And so that's one part, right? That's the kind of the, the platforms that are being infused by ai.

But then obviously there's the, there's the other part which is platforms for AI and platform engineers as being key enablers for AI adoption across, um, the enterprise. And, you know, we discussed a lot this in famous MIT study, right? Of the, like, 95% of pilots are failing in the enterprise, whatever.

I've seen actually some people debating that and how true that stu like how, you know, double blind or whatever that study is. Um, but, but regardless, I think we all know at least anecdotally, right, that there is this gap right now between very impressive, you know, day one or day zero demos, right? Of what you can do and then the actual impact that you can drive on an enterprise scale day two, day 100, which we know also from CIS admin, from platform engineering, you know, is the day two, day 100 stuff that actually really delivers the value, right?

Um, and so, and so that's where we're seeing, um, a lot of the responsibility and the ownership and the load already shifting over onto the platform team. Um, and it's interesting, we recently did, we recently published our state of AI and platform engineering report in the community. Um, and we, uh, asked, you know, hundreds of teams there and, and I think it's 36, 30 7% of all the AI workloads agents workflows already owned by the platform engineering team up from, you know, like five, 10% just a year ago, right?

So I think you can already see this like shift and, you know, and I think even if you look at like, um, sort of ad hoc AI and AI ops teams, they're still relying, uh, on the platform at the end of the day. So really you could actually argue already 50% plus of all AI workloads, uh, uh, and, and, and workflows are already kind of supported by the platform engineering team. And I think that's only growing.

So, um, that's, you know, where I think we keep going back to the fact that, you know, there's been a lot of trends that we've discussed and, uh, you know, especially you've covered in the last 10, 15 years in the enterprise that, um, I think have been overshadowed in the last 18 to 24 months by AI in the enterprise. Um, and I think a platform engineering holds strong as actually one of the few ones that not only hasn't been overshadowed, but I would argue it's been turbocharged by ai. Um, because people are realizing like, Hey, if I wanna move past this initial, uh, you know, quick cool demo and, and go, you know, to something that actually lets me deliver on the promise of AI in the enterprise, I need some well-designed platform underneath it.

Yeah. You know, look, I, I'm listening to you. com about AI's influence on platform engineering with some real things you should, you could do to make, make AI your friend not your enemy.

Right? You Yeah. For, and and this goes beyond platform engineering.

I think it goes everywhere that AI and AI is everywhere. You could, you could embrace it and, and make, help it make you better, help make what you do better, or you could resist it and you'll get steamrolled, right? You, you, you, there is no, you can't run away from it.

You can't hide from it. That being said, you know, we had Nathan Harvey from, uh, Google, uh, at the DevOps experience virtual event, I think you presented. Yeah, yeah, of course.

Uh, Nathan, Nathan, Nathan gave, you know the report, the Dora report? Yeah. Well, he's in the Dora working group.

Yeah, yeah, yeah. Yeah. He, you know, he's, he's kind of the, a lot of the face for it.

So, but here's the thing. I'll go back to early DevOps. There was this huge debate, ah, DevOps is Bull Bull DevOps doesn't work.

It's, it's just a marketing term, right? And it's a lot of kumbaya about cultural stuff, but it really doesn't. And what we saw in DevOps is when individuals or in or small teams, you know, an individual started using Puppet or Chef or Ansible, they started using Jenkins or, or you know, some other CICD tool and, and they brought it to their small team.

It worked. Yeah. People were like, yeah, this is real.

It's great. Yeah. And then you go to a large enterprise and you had a little small team here using Jenkins and this little small team here used something else, and this little small team there, something else, and all these little small teams, it was working.

It wasn't until you brought it all together and said, okay, now we're gonna roll this out enterprise wide. That you started saying, wait, wait, we can't use five different CICD tools. We can't use Puppet chef and Ansible.

We've got to, or, or Terraform or, you know, whatever. We've gotta standardize on one tool across the enterprise. Right?

And that really is what gives rise to platform engineering. People say, wait a second, we gotta standardize on a platform, not 12 platforms. Right?

Each little team can't go off. And, and so we had a very similar kind of thing in the DevOps space where yeah, everyone's using it, but does it really work? 'cause everyone's using their own flavor or whatever you want to call it.

But you know, you, you cited the Gartner and I, I've also looked at that Gartner report very, you know, very encouraging. Let me tell you another report that I've seen, though, 90%, 90% of developers are using AI to help them code, whether it's in the IDP or in Git or what have you. 40% don't trust it.

40% don't trust it. 65% say it absolutely, uh, introduces instabilities into their code base. Yeah.

Vulnerabilities. Yeah. Yeah.

So think about, well, not just vulnera either vulnerabilities or other instabilities, right? Right. Mm-hmm.

But, but let's just think about the logic here. We know 40% of you don't trust it. We know 65, 2 thirds of you think it introduces instability, but yet 90% of you are using it.

Yeah. Like what, where that, that's a disconnect. It's like, yeah.

You know, just the heck with it. I don't care if I don't trust it, I don't care if it introduces. Yeah.

Well, because it still accelerates productivity so much, right? That and if you don't do it, you're the only one falling behind. 'cause you don't do it.

Well, it's FOMO that you just said. Yes. Fomo.

There's a lot of fomo. Yeah. Because you say it accelerates.

Yeah. I've seen other studies where it says it decreases developer. Yeah.

Or you think it accelerates what you think it does, but it actually decreases developer productivity by 19%. Now, whether you believe that or not. Yeah.

'cause there's a million, you know, there's lies. Damn lies. And then there's metrics.

Yeah. So, you know, but, but I think we're still at that kind of prove it stage right. With, with this, and, and here's my other, and I'll just say this quickly and throw it back to you.

Do we need a new platform for the AI age? Or can we ai empower the platforms we've been building, right? So in words, can we build on what we've already got or do we gotta just like, I've heard people say we need an AI stack.

Yeah. Right. We we're not going to use the old stack.

I'm not a big fan of throwing out things that work. Yeah. Right.

Yeah. Me too. Me too.

I think, and by the way, just to add to what you were saying, I think like this, this just really speaks to, um, just the enormous, you know, kind of revolutionary thing that we're living through right now, because Yeah. You know, also, like I, I was, you know, people keep saying like, okay, you know, all this ai, there's, there's, we're like overinvesting in ai, you know, we're the, the spend of the Googles, the open eyes, whatever, it's crazy. Um, and like yes.

You know, those are massive numbers, right? And, and especially if you compare it to the revenue captured, it's, it, you know, there's a big miss right now. It's Finger port is Somalia.

Have you seen that one yet? That Yeah, yeah, yeah, yeah, yeah. It's single Port Somalia.

Exactly. And so, a big mismatch, however, you know, I was listening to this podcast the other day, it was very interesting. 'cause there were, um, there were actually, you know, differentiating between like, yeah, it's true.

The revenue capture today, you know, there's a big mismatch. But if you think about the value created today, um, you know, you could argue actually that there is like a, a, a match, right? Because I mean, everybody's using it.

Like, it's, it's insane, right? Like the, the penetration of usage and like, I just see myself both on my personal life and my work life, just the AI usage that then I'm just paying like, what, 20 bucks per month for like, it's crazy. Like the, the value it's created for me is like many, many, many multiples of 20 bucks per month.

Um, you know, so, so, but, but there's still no like clear mechanism, right? Um, like Google a little bit when they figured out, okay, well we can do the sponsor Blue links, right? And it is like bang, you know, and there's like the man, the money printer started.

So I think we're also in that phase. That is, is very, very interesting. And I think because let me, hold on, let me throw something outta at you.

Yeah. You're paying 'em 20 bucks a month, how much is it costing them? Yeah, exactly.

Probably. Exactly. This is the thing.

They're not capturing, you don't make it up in value, but they're not capturing the value that they created for me the right way. Right. Because it's like, even just like, you know, we're just talking offline about like me buying a house, like the, the amount of money I saved in, in like consulting experts, um, just by like going back and forth on like deep research stuff with Chad GPT.

It's crazy, you know? Yeah. And like, of course, in the end, you still need an expert, but like, instead of like waiting for like three weeks and like doing a call every 10 days because these people are busy, you know, I just did one call just cost me like 500 euros.

That's it. You know? Yeah.

So it's like, it, it's, it's insane. So that saved me probably like two, three grand, and they captured none of that. Right?

Like, so it's, it's very interesting. Um, but yeah. Anyway, so going back to, uh, to your thing, to your question, I, I agree.

Like, I'm also not a fan of, of, um, you know, throwing everything away. And, and I think that was exactly a big topic or conversation in Paris last week, is like, look, at the end of the day, um, in fact, I, I'd argue it's, it's really about using the same golden pats, the same kind of paved roads that we've already been building. Um, uh, and, and, and in fact, this is the ultimate stress test, right?

Because it's literally like, if you think about a road, right? For example, it's like, well, you could have like a road that's like not really well maintained or not really well built, right? And it's like, okay, well you can have maybe one car driving over it, two cars, three cars, right?

But at some point you're like, it's causing problems, right? Whereas, um, and you know, and like essentially going from like X amount of developers to X amount of developer plus y agents, plus, you know, Zed AI enabled the developers, right? It's like you're going from like, you know, 10 cars to now like a million trucks and they're all like running, you know, they're all driving like super fast on the same road.

Yep. That's not gonna work, right? But to think that's, that's, that's actually why I think AI and platform engineering, um, you know, I think, you know, there were a lot of people also being like, oh, if it wasn't for ai, platform engineering would be the number one trying enterprise today.

And again, I think this, even this just exposition is not the right framing. The right framing is like, what can AI plus platform engineering unlock? Um, and I think it's huge.

And, and it's actually just leveraging the existing stack, to your point. It's leveraging the existing, um, design golden paths, but just is gonna push them further in terms of better design, um, and better constraints, uh, and, and better, um, design thinking around what's the right level of obstruction versus the right level of context that you wanna provide to different users. Yes.

You're expanding the number and type of users, right? Like we said, we go from just app devs essentially to ML ops engineer, data scientists, agents, you know, agent enabled people. But ultimately it's, they're all driving on the same highway.

It is just, it's gotta be a better highway. Uh, absolutely. So here, here's the lesson.

I I, you know, we talk about this a lot at Textron on techron gang, on podcasts like this. We write about it A mistake some company or some organizations are making are, they're looking at what they have now and saying, okay, by using ai, I can make this better. But what they're not taking into account is that by using ai, right?

So instead of let's use your highways, instead of saying, look, I can make a two-lane road, a three-lane road. Yes. And I could smooth it out, so I'm gonna have less traffic, I'll be able to go faster on that road.

Great use of ai. But what they don't take into account is that by everyone else using ai, I now need a five lane road. Yeah.

Right? Or a six lane road. So it's like you, you can't use AI sometimes I think just for incremental to improve your current situation, you've gotta think about what's this gonna be like in an AI enabled world where it's not just the platform, but the developer, the security, different hacker, everyone is using ai, and what does that do?

I need a super highway now, not just the three lane road. Exactly. And so let me build that super highway and that, like you said, that starts bringing in who else is gonna be driving on that road, right?

And so there's all these other players now that are, that are in the mix that we gotta account for. And that's what AI is doing. It's like, it's not just Turbocharging platform engineering, it's turbocharging the whole, the whole game.

Everything. Everything, everything. Yeah.

Interesting stuff. Hey, is, is any of the Paris in person stuff recorded available on the website? It is, it is.

It's gonna, um, I am not a hundred percent sure yet when it's gonna be edited and live, but I think it's coming very, very soon. Either this week probably, or latest next week. Um, so then we can link it in the episode notes, um, and, and share with people.

Um, but yeah. Um, and then onto Perfect. The next one you're gonna be at, uh, Q Con, right?

I'm gonna, we're gonna be broadcasting live at Q Con on the show floor. Uh, uh, my friend Steve Foskett in the tech field day are doing the tech field day there. I'm trying to remember the date off the top of my head.

I wanna say it's like November 10th, something like that. 11th week. Yeah, November 10th is, is the, well, November 10th is the co-located events the day before.

Right. That's also when Hazel Cube happens. Um, and, and so that's when we'll party.

And then on the 11th is, is when the conference kicks off. Yep. It will be there.

I think I get in the ninth actually. So, well, you are coming to as a cube right here. I'll be at House of Cube for sure then.

And then, uh, nowhere there, I guess, I don't know if I leave Thursday night Friday because for us I'm on Thursday night. Yeah. We, we have actually, well, we have a short flight for me.

I mean, it, it's nice Atlanta down here. I'm actually going down to, we're hosting this, uh, platform day, um, in Sao Pao in Brazil the week after. So I'm going down there.

Really? Which I don't know, it's like so, so European of me. I was like, oh, you know, once you're there, it's gonna be very close.

It's like, no, that's a far ride my friend. Yeah, that's a far ride. That's probably from Atlanta to Sao Paulo, I think is further than to Italy For you or close.

Yes. About the same, same nine and a half hours flight. So I think it's very similar.

Yeah. Uh, I was like, um, yeah, we have, uh, Brian, one of our cameramen, his wife is, she's Brazilian. She's home in Brazil right now.

So we were just talking about how long that flight is. Yeah, it's so long. Yeah.

Now, if you went from Atlanta down to Miami and stayed by us for a day or two before going to Brazil, that would be easier. We could not, we could take an hour off of that, but maybe, maybe on the way back, let's talk about that on the way back. You, let's do that.

We'll talk. All right. But CubeCon, lots of platform engine.

They're, I'm sure, are they having their platform engineering day again? Yeah, there's the engineering day that's also on the Monday throughout the day. Then people come to us Cube for the party, and then we kick it off on the, I guess it's the Tuesday, um, with the, with the main event.

Yeah. And we'll, we'll have a lot of stuff there. We'll, I mean, you're doing the, you know, the interviews, I've been doing videos all, all week on the show.

Floyd, stop by, we'll do video. Perfect. Let's do that.

We'll do workshops. We have a couple of round tables that we're hosting with ThoughtWorks and other people, so it's gonna be fun. It's pretty busy.

Good stuff. Cube con. I, I'm interested to, you know, lately the European, uh, q coupons have been bigger than the North America.

Yeah. So I'm interested to see if that holds this year. Same.

I was, I was, I was having this conversation literally with somebody, um, a couple weeks ago. 'cause she was telling me that Well, it didn't used to be like that. It used to be No pre COVID.

The US was bigger pre COVID the US. Yeah. Like, uh, the one they did in San Diego was big.

Yeah. And the one in Seattle, but dang there. And COVID.

Well, I also think it is like, you know, San Diego, like, probably wasn't people Diego? No, it wasn't bad. A lot like Detroit and Chicago and Winter is like a different vibe.

Look, I don't know what they were thinking between you and me, but, you know, look, they make a, they make a cube con. We go, Atlanta is, I don't know if how much time you spend in Atlanta. Atlanta is in nice DT It's a, it's a nice town.

Uh, they got good stuff there, you know. I mean, look, it's not Paris, but, uh, it, it's a nice town. It's a, it's a nice town.

Nice people. Well, looking forward to, uh, to seeing you there. Absolutely.

Luca, we gotta pull the plug on this. It's over a half hour. Yeah.

But I will see you there. And we'll, my promise is we're doing more of these regularly. I, we, we let this go about traveling.

Yes. We can let this go. Yes.

We gotta do it from the road. Yeah. All righty.

Luca Gallente, platform engineering community here. org. com.

We'll see you next time.