Vox

Leitura de 49 min

EP75: Building an AI-Native Engineering Team with Nic Neate

Summary

In this episode, host Ben Pearson speaks with Nic Neate, CTO of Nimbus, about the practical realities of building an AI-native engineering team. Nic shares his unique perspective, having previously lost his job at Microsoft due to AI investment shifts and now leading a team through this transformation.

The conversation defines what an AI-native team means, moving beyond simple AI assistance to ‘vibe coding’ and agentic systems where AI handles core coding tasks. Nic explains this is a spectrum, from human-written code to AI assistants, supervised agentic coding, and potentially fully autonomous AI, with the need for human guardrails and accountability.

They explore why organizations should pursue this path: competitive necessity and the ability to do more of what their business does best. However, Nic emphasizes this requires a fundamental skill shift—away from deep focus coding and detailed code review toward context engineering, rapid decision-making, and initiative-taking. The discussion covers practical challenges including team member adaptation, mental fatigue from constant context switching (illustrated through Jevons Paradox), and the evolving role of senior versus junior engineers.

Nic provides concrete advice: prioritize investment in your AI toolchain (suggesting 20% of engineering effort), completely overhaul recruitment processes away from traditional coding challenges, and focus on training for the new required skill sets rather than the old career paths. The episode concludes with Nic’s key takeaways about the urgency of adaptation and the importance of creating fulfilling roles in this new landscape.

Recommendations

Concepts

  • Jevons Paradox — An economic principle where increased efficiency leads to increased total demand. Nic uses it as an analogy for why AI-powered engineers become more efficient but also more stressed and busy—like steam trains using less coal per journey but the railway boom increasing total coal demand.

People

  • Mark Russinovich & Scott Hanselman — Mentioned in relation to a Microsoft article about early-career developers in the AI era. Nic expressed skepticism about their approach given actual layoffs in big tech.

Tools

  • Cursor — Mentioned as an example of an IDE with AI capabilities that enables vibe coding experiences.
  • GitHub Copilot — Referenced as an AI coding assistant that can be enabled in IDEs to transform the coding experience.

Topic Timeline

  • 00:01:04 — Introduction to AI-native engineering teams — Ben introduces the episode topic: building AI-native engineering teams rather than just using AI at the edges. He welcomes Nic Neate, CTO of Nimbus, who has led his department on this transformation journey for the past year and a half. Nic shares his personal background, including losing his job at Microsoft when they shifted investment to AI.
  • 00:04:16 — Defining AI-native and vibe coding — Nic defines what an AI-native engineering team means, distinguishing it from superficial AI tagging. He introduces ‘vibe coding’—using AI as an assistant where humans prompt and direct while AI does the actual coding work. Nic describes this as a transformative, ‘divine powers’ experience and explains AI-native means riding the wave of constantly improving AI tooling while maintaining human control and accountability.
  • 00:08:44 — Why pursue AI-native transformation — Ben asks why organizations should build AI-native teams. Nic compares it to a logistics company sticking with horse and cart instead of adopting lorries—it’s about competitive necessity and doing more of what your business does best. While AI-native improves all measured outputs (productivity, speed, quality), it requires a complete transformation in skills and ways of working, not just flicking a switch.
  • 00:12:14 — Experiments and lessons at Nimbus — Nic shares practical experiments at Nimbus, including having product managers and data engineers try vibe coding to create products. While these produced impressive demos, they revealed challenges when integrating into production tech stacks and maintaining code. These experiments highlight that AI tools don’t automatically produce perfect, production-ready software—there’s still significant work required.
  • 00:15:41 — New skills for AI-native engineers — The discussion shifts to the fundamental skill changes required. Deep focus coding and detailed code review—previously prized abilities—are no longer key since AI does these better and faster. The new essential skills are: 1) Context engineering (providing AI with the right business, product, and architectural context), 2) Rapid decision-making (handling AI’s faster output cadence), and 3) Initiative and leadership (driving direction and unblocking AI agents).
  • 00:28:28 — Challenges: team adaptation and context switching — Ben raises challenges including team members potentially not wanting this different job (comparing to weavers versus machine operators in the Industrial Revolution). They discuss mental fatigue from constant context switching as AI agents complete tasks in minutes rather than days. Nic introduces Jevons Paradox—where increased efficiency (like AI-powered engineers) leads to increased demand and stress—and emphasizes the need to develop context-switching as a skill.
  • 00:34:46 — The junior engineer pipeline problem — Ben questions what happens to junior engineers since the new AI-native roles seem more senior. Nic references a recent Microsoft article about investing in early-career developers but expresses skepticism given actual layoffs in big tech. He argues we need to train people for the new skill set directly, not take them through the old journey—comparing it to not needing to train someone to groom horses to become a lorry mechanic.
  • 00:37:38 — Concrete actions and advice — Nic provides actionable advice: prioritize investment in your AI toolchain (suggesting 20% of engineering effort), completely overhaul recruitment processes away from LeetCode challenges, and focus on training for the new required skills. He emphasizes that if you’re not already changing processes, skill sets, and toolchains, you won’t be competitive in 6-12 months.
  • 00:41:10 — Key takeaways and conclusion — Nic’s key takeaways: 1) You must change your process, skill set, and toolchain now to keep up competitively, and 2) Focus on creating fulfilling roles for your people through training and understanding motivation. Ben reflects on how software engineering is fundamentally transformed by AI unlike many other functions. Nic notes this transformation will eventually extend to all business functions.

Episode Info

  • Podcast: Tech World Human Skills
  • Author: Ben Pearce
  • Category: Technology Business Careers Education Self-Improvement
  • Published: 2026-03-04T05:30:00Z
  • Duration: 00:45:11

References

Podcast Info

Transcript

[00:00:00] Hi there, I’m Ben Pearson. Welcome to the Tech World Human Skills podcast. Every episode

[00:00:06] we talk through how to thrive in the tech world, not just survive. Now, if you want

[00:00:11] me to work with your team, just give me a shout. I love to help teams be more influential,

[00:00:16] memorable and successful with their stakeholders. Head over to www.techworldhumanskills.com to

[00:00:22] book a chat. Hey folks, and welcome to the Tech World Human Skills podcast. Well, this

[00:00:33] is a very special episode today. Why? Well, this episode is a recording of a fireside

[00:00:40] chat from the cloud and AI infrastructure stage at Tech Show London. Well, well, sort

[00:00:46] of. The tech setup was a bit hard to do it live, and we can make this slide.

[00:00:52] It’s slightly longer on the podcast. So, so it’s another version of the conversation

[00:00:56] that happened at Tech Show London on the stage. Now, we’re talking about building out an AI

[00:01:04] native engineering team. Software engineering is definitely one of the earliest industries

[00:01:11] to be massively changed by generative AI. So rather than talk about how AI has helped

[00:01:18] around the edges of a team, we’re talking about how AI has helped around the edges of

[00:01:22] a team that’s fundamentally changed. Now, our guest today has led an engineering department

[00:01:28] for the last year and a half on that very journey. He’s got some amazing lived experience

[00:01:35] and insight to share with us. So please welcome to the show, CTO of Nimbus, Nick Neat. Nick,

[00:01:44] it is brilliant to have you with us. Ben, it’s an honour. Thanks for having me. It’s

[00:01:50] great to see you again. And yeah, I can’t wait to…

[00:01:52] To explore this, because as you say, it’s a really, really important and really relevant

[00:01:57] conversation right now.

[00:01:59] It sure is. And literally, as we’ve been prepping for both the Tech Show London recording, which

[00:02:06] is actually happening next week, we’re recording this. But as we’ve been prepping for that and

[00:02:10] prepping for this, article after article after article has been hitting our news feeds talking

[00:02:16] about this very topic. And we’ve been like pinging each other on WhatsApp going, have

[00:02:20] you seen this article? Have you seen this article?

[00:02:22] So it feels like one of those topics that is just the top of everybody’s mind at the

[00:02:28] moment, top of everybody’s news feed. And so it’s great to have a conversation about

[00:02:32] it. But before we get into the detail, just for all our lovely listeners that have never

[00:02:37] heard of you before, Nick, or met you before, could you tell us a little bit about your

[00:02:42] background and what you’re doing?

[00:02:43] Where have you been? Yeah, I’m Nick. And I’d start by saying I’m probably…

[00:02:52] One of the first people who’s actually kind of lost their job to AI as a software engineer.

[00:02:57] I was at Microsoft a couple of years ago, and my whole department got sacked off, the reason

[00:03:04] being that Microsoft wanted to invest in AI instead. And therefore, we were no longer

[00:03:09] the strategic priority. So I’ve kind of lived that experience of kind of lost that income

[00:03:17] and that purpose to AI.

[00:03:21] But then gone into this…

[00:03:22] This job with Nimbus, the CTO, where I’m now leading a product team, leading an engineering

[00:03:27] team, and riding that wave and just going on that transformation. And in a slightly

[00:03:33] different way, because we’re not getting rid of our team. We’re sort of in the opposite

[00:03:38] of what we’re doing in our approach to AI native and how to do that well. So that’s

[00:03:43] what I want to talk about.

[00:03:46] Well, I’m really excited to have this proper sit down, this proper conversation. I think

[00:03:51] the first thing…

[00:03:52] The first thing I’d like to talk about is to maybe define what an AI native engineering

[00:03:58] team. I mean, everybody is AI everything at the moment. And often that’s vaporware.

[00:04:06] You know, we’re just tagging the word AI on what we’ve always done before. But I don’t

[00:04:10] think that’s the case with you. So could you tell me what you mean by an AI native engineering

[00:04:15] team?

[00:04:16] Yeah, and I think that’s really important because I want to be real about it and not

[00:04:20] just kind of sugarcoating everything.

[00:04:22] And telling you that you can do anything you want with AI. So let’s start with what is

[00:04:28] vibe coding. Because that’s kind of also kind of a key buzzword and something which is transforming

[00:04:34] software teams. And so by coding, I mean using AI to write code where it’s an assistant to

[00:04:44] a human, but it goes beyond that. A human is there driving it. They’re prompting it.

[00:04:48] They’re sat there with cursor, with VS code, whatever it is.

[00:04:52] The AI is doing all of the actual coding work and splitting things out. Yeah, I mean, have you

[00:04:58] got an experience of vibe coding, Ben? Is that something that you’ve tried yet?

[00:05:03] Not really, not in a vibe coding. So I have prompted and got code created for me that I’ve

[00:05:09] then read and gone, yep, that does what I want. And I’ve then deployed that. But I would not say

[00:05:16] that was before the term vibe coding has come along. So I’m going to say, I’m going to say,

[00:05:19] I don’t really, I don’t really know. So, so no.

[00:05:22] I think to kind of understand what’s happening, I would strongly recommend that anyone who’s got

[00:05:28] any kind of engineering background, even if they haven’t done it for ages, or even if they’re much

[00:05:33] too senior to actually look at the code anymore, have a go. It’s dead easy just to, if you’ve got

[00:05:40] an IDE, enable the AI part of it, copilot, cursor, whatever it is, and just see what happens.

[00:05:47] Because it’s an incredible transformative experience when you do that. If you think

[00:05:52] about the first time you used ChatDVT, you know, years ago, back in the day, and you’re like,

[00:05:56] hold on, this is like a person who knows stuff. And it’s great ideas, it’s talking to me.

[00:06:01] Yeah.

[00:06:01] Multiply that by 100 for the experience of, it’s, I’ve just created this web app from scratch,

[00:06:07] and I didn’t even know how to do it myself. And it’s there, and it works. So that’s,

[00:06:12] that’s vibe coding. And it’s, it’s, it’s like having divine powers. It’s like, you know, you,

[00:06:16] you speak into your computer, and the, out of the swirling chaos of zeros and ones in the memory,

[00:06:22] these, these, these, these, these, these, these, these, these, these, these, these, these, these, these,

[00:06:22] incredible creations just, just spring into being. So that’s, that’s kind of a starting point

[00:06:28] for AI Native. And maybe, maybe six months ago, I’d have been saying, you know, that’s,

[00:06:32] that’s AI Native. But the thing about AI Native is, obviously, it’s, it’s a journey,

[00:06:36] it’s a spectrum. And the tooling, the capabilities keep on changing and getting better. So there’s no,

[00:06:42] there’s no single goalpost that you can say, this is AI Native, we’ve now done it.

[00:06:47] We’re getting on a journey from, originally, humans did, did it all. They, they wrote every

[00:06:52] line of code by, by hand. Then there was AR assistants who started, you know, giving

[00:06:58] suggestions of what the next line might be. And then they started writing the, the whole thing.

[00:07:03] And that’s, you know, we’re in the bat vibe, vibe coding phase. Then it’s kind of moved on to,

[00:07:07] to supervised, where you’re, you maybe you’ve got an agentic coding system, you’re,

[00:07:12] you’re telling it, this is the requirement, and it just goes away, and works on that in the

[00:07:17] background, and then comes back when it’s finished and says, here’s your PR, review that. If you’re

[00:07:21] happy, then, then we’re good to go. And then maybe beyond that, it gets on to the AI working

[00:07:26] completely independently. But that’s, that sounds, sounds dangerous. So we need, we need some

[00:07:33] cautions over whether they actually, how far on this journey we want to go, and what guardrails

[00:07:37] need in place for, for safety and responsible use of AI within it as well. So, so when you’re

[00:07:43] saying AI Native, what you’re saying is your main production code, the stuff that you do, that your

[00:07:48] team do for a living, is in, created with vibe coding and AI tools.

[00:07:56] And vibe, vibe coding, agentic coding, it’s, it’s, it’s, so the things right now, but it’s,

[00:08:03] AI Native means you’re riding the wave of this AI tooling, which is, is constantly coming through

[00:08:08] and supercharging your, your delivery velocity. But, but also dealing with the reality of that,

[00:08:14] because it’s not like, it just, it just happens and everything’s suddenly rosy. So there’s, there’s all

[00:08:18] kinds of issues you work through. So you’re, you’re maximising the agency of the AI tools,

[00:08:24] you’re bringing in the, the new systems, the new capabilities as, as they appear.

[00:08:29] But also you’re retaining that, that human control, that human accountability

[00:08:33] as to, as to what gets, gets delivered and actually makes it into the code base.

[00:08:37] Okay. So that’s a good overview of what we’re talking about. So, so my next question

[00:08:44] therefore, and we’ve, you’ve already started to talk about it, but, but why would you

[00:08:48] wanna do this, you know, there’s a lot of shiny stuff out there. You know, AI slop

[00:08:52] you could talk about as well. You know, there’s lots of shiny stuff with AI, but why would you

[00:08:58] want to create an AI native engineering team? So, so, so to me, that’s almost, not even a

[00:09:04] question because, okay, that’s like, that’s like saying I’ve got this incredible logistics business

[00:09:11] and it’s super efficient, highly professional, well-organised team who do it all with horse and

[00:09:18] cart and do that do that perfectly why i don’t want to use a lorry and yeah so there’s there’s

[00:09:25] kind of a there’s a carrot side of it which is whatever whatever your business is i assume your

[00:09:30] business is doing something awesome in the world and you would love for it to do more of it and

[00:09:35] being a native enables you to do so much more of it and there’s there’s a stick as well which is

[00:09:40] i assume your business also has competitors and you want to continue to exist as a business

[00:09:44] and riding this wave and keeping up with the the ai native aspect of your business is is crucial to

[00:09:52] to doing that so that in a nutshell that’s why and and so if we were to break that apart for the

[00:09:57] say we’re that we’re in the business so i know we’ve all got businesses but we’re in the business

[00:10:03] of creating software say for that business whatever it might be so is ai native gonna make

[00:10:09] your team more productive is it going to raise the speed of what you’re doing is it gonna

[00:10:14] raise the quality of what you what’s the things that it’s gonna do better for you as the business

[00:10:20] of a software engineering team it’s all of those things yeah everything that is a measured output

[00:10:26] of your of your team in terms of delivery to the business or to the customer what it’s doing can

[00:10:32] be made better through being a native but you don’t just flick a switch and it is better it’s

[00:10:39] a transformation it’s a whole different skill set and way of working that you move into and it’s

[00:10:44] navigating that journey which is is the challenge for for engineering leaders and engineering teams

[00:10:49] in in today’s world okay so basically you know so far we’re saying this is a rosy future of

[00:10:56] everything that you do and create is going to be better in an ai native landscape why would you not

[00:11:05] do it that’s kind of what i’m hearing from you so far um yes and better in terms of the what is

[00:11:14] but you it’s you know you’re going to be facing challenges within that and you might be an

[00:11:19] individual who’s um you’re not so happy about the parts of your job which you don’t get to do anymore

[00:11:24] um or you might be facing particular sort of nightmares with with context switching and

[00:11:32] trying to keep up with different agents doing different things all the time so it doesn’t mean

[00:11:37] that your your mental health is is better necessarily so understanding that and navigating

[00:11:42] that is part of it we’re going to be able to do that in the future so i think it’s going to be

[00:11:44] we want to make this better for everyone yeah better for the individuals and and better for

[00:11:48] the business so i guess what you’re saying is the output of your function the output of your team

[00:11:52] is better as a result of ai but that doesn’t mean that it’s a silver bullet that it’s not

[00:11:59] difficult that there’s not transitions that there’s not things to the but ultimately you’re

[00:12:03] saying that the output is going to be a lot better of the team as a whole if you’re embracing some

[00:12:08] of these new technologies absolutely so so what sort of things have you

[00:12:14] raced and what sort of things have you done at nimbus so we we’ve tried a lot of things um it’s

[00:12:19] it’s been a hugely exciting time to be anywhere in tech and to be um at a company which is sort of

[00:12:26] small scale up um has got the agility and the flexibility just to to jump in and have a go at

[00:12:32] something see if it works um and and move on and but also in a company that’s in a competitive

[00:12:38] situation and has a necessity to innovate and to and to move forwards and and and to try and

[00:12:44] get the right things done and and and and and and and and and and and and and and and and and

[00:12:44] and ahead so i i could i could give you some examples you know we’ve we’ve tried getting um

[00:12:50] product managers to to create right code create product using using vibe coding using using

[00:12:56] lovable um is that a quicker way to get it get something viable viable to market and that’s not

[00:13:02] something we’ve been able to make make work out for us when actually the rubber hits the road and

[00:13:06] you try and take take that purely vibe coded thing and and run it in production as part of your tech

[00:13:12] stack um

[00:13:14] we’ve tried getting an enthusiastic data engineer to create a software product and you know quickly

[00:13:20] vibe coded something which is you know incredible demo um but all kinds of issues with with the

[00:13:28] code base again um when you want to do proper devops with it or you you want to maintain it

[00:13:32] and enhance it that it suddenly becomes very expensive um so that is trying trying these

[00:13:40] things is a key part of the journey you know the fun of the journey um and and and and and and and

[00:13:44] some of it works some of it not so much right so but you’ve been experimenting trying things and

[00:13:51] trying to figure out what’s working really well and it’s interesting there because like when when

[00:13:55] you started saying hey this is hey this is better you’ve got divine powers i think it’s like you’ve

[00:14:02] got divine powers that you can just type in and actually you’re saying well we tried it with a

[00:14:08] data engineer and actually it didn’t work quite right and we tried it with some product managers

[00:14:11] and it didn’t quite so it’s not quite as good as it should be but it’s not quite as good as it should be

[00:14:14] divine powers think it and it is now perfect in front of you there’s there’s still lots of things

[00:14:20] that need to happen to make good quality production ready software using yeah so that that kind of

[00:14:27] comes on to the the kind of the skill set um shift and the skills that you need in order to

[00:14:33] deliver engineering deliver software deliver your your function to the to the highest quality and

[00:14:39] and to the highest velocity making that making the best use

[00:14:44] of these tools and you know what what what does it mean to have had be a highly efficient high

[00:14:48] performing engineering team in in this new world because you know we we all we all say we all

[00:14:54] believe that we’ve got this this high performing engineering team we all tell a board that’s what

[00:14:58] we’ve got certainly um have you because the the people that make up that team the processes and

[00:15:05] the ways they work are very different now to what they were 12 months ago so unless you’ve been

[00:15:11] evolving your team and and kind of keeping pace and

[00:15:14] performing then then i’m i’m not convinced that you you still do and that’s why it’s important so

[00:15:19] should we unpack that because i think that that’s really interesting so you know if you’re if you’re

[00:15:26] thinking now about a software engineering team today versus one a year ago two years ago three

[00:15:32] years ago what are those skills of an ai native software engineer yeah because it i mean it’s

[00:15:41] going back to that kind of that example of you’ve got a software engineering team and you’ve got a

[00:15:44] transportation logistics company that’s very skilled with with horses then the the skills the

[00:15:51] people that you need to to drive lorries and coordinate lorries and be mechanics on lorries

[00:15:56] is going to be very different so we need to understand that shift um and what what the skill

[00:16:01] set is so a couple of examples with with engineering and developers one of the things that

[00:16:09] we we look for that we valued most highly a couple of years ago

[00:16:14] was the the deep focus the ability to get kind of get locked in to zone in on on a task to achieve

[00:16:21] that flow state and to be you know fully motivated and at your most productive and just churn out some

[00:16:28] some reams of code doing doing something new and incredible within a few days a few weeks

[00:16:34] and that suddenly is completely not the job anymore because however good you were at it

[00:16:41] and however like have a much funier

[00:16:44] had doing that the ai can do it better it can do it in in 10 minutes what you could do in in a week

[00:16:50] and that’s that’s exactly the thing that ai’s is brilliant at is when it’s got the context when it

[00:16:57] knows what it’s doing is just creating good quality code and the the skill set has now

[00:17:04] shifted to how do i make the ai do that rather than how how do i do it myself so that’s that’s

[00:17:10] that’s a big shift for someone who was you know incredibly prized

[00:17:14] for their just coding output their deep deep focus ability a couple of years ago code review

[00:17:21] similar okay the people who could could really do a detailed low-level code review and and find the

[00:17:28] the salient the important problems at that early stage and save tons of time down the line in

[00:17:33] in testing and you know fixing problems after they hit the field again that’s that’s no longer

[00:17:39] the key skill anymore because that again is something ai’s is very good at you don’t have

[00:17:44] to review every line of code in the way that you used to and the way you do code review is is changing

[00:17:50] so so i guess that so some examples of what it’s not so it’s no longer deep focus coding it’s no

[00:17:59] longer reading code checking code code reviews that kind of stuff so those skills that you may

[00:18:04] have had before are not what are going to make you really thrive in this new world so what are

[00:18:10] the sorts of skills that are going to make you thrive in this new world

[00:18:14] number one for me is is context engineering okay um because the the context is what turns

[00:18:21] ai from from a novelty into something which is um super powerful and and amazing um

[00:18:28] as a as a cto as an engineering leader we’ve all suddenly got the ability to spin up an engineering

[00:18:36] team which rivals microsoft’s in terms of the the number and the quality of the brains which can

[00:18:42] can create product create code and and and and and and and and and and and and and and and

[00:18:44] why aren’t we all doing that um it’s because we we don’t have the ability to do the context

[00:18:50] engineering which gives the ai the salient relevant information that it needs to write the

[00:18:57] the right code to move your product to move your function move your business business forwards so

[00:19:02] the the key the key skill as a as an engineer is in driving ai by giving it the right context

[00:19:09] um and that can be uh prompting and at a

[00:19:14] simple level how do you communicate most most clearly in a and a few sentences to the ai the

[00:19:19] the simple task that you want it to do um so that that communication and that’s kind of

[00:19:25] the english language communication is suddenly a very important skill for for engineers where you

[00:19:30] know if you’re the kind of engineer who’s actually better at communicating in c-sharp than you are in

[00:19:35] english then you’re going to struggle to to make that transition and to be a strong engineer in an

[00:19:41] native world

[00:19:42] um but it’s

[00:19:44] not it’s not just the prompting it’s the the broader context as well so how do you bring the

[00:19:48] documentation the understanding of your whole stack your architecture the bigger picture of

[00:19:53] what you’re building into the the narrow task that an ai is doing so so so skill number one

[00:20:00] was was context engineering so so that is being able to effectively describe in the prompt

[00:20:09] this is what i am trying to achieve um is that sort of

[00:20:14] so outlining the vision of where your product is heading or is that can you unpack that a bit more

[00:20:21] for me there’s um there’s the kind of the the big picture context um it needs to understand your

[00:20:27] your company and your what what your business does so that it knows the the domain it’s working in

[00:20:33] there’s the the kind of the product context there’s this is my my tech stack this is the

[00:20:39] wider architecture you might have 15 code bases that make up a broad range of things so it’s

[00:20:44] your product and your ai is working on one of them as part of a feature but it needs to understand

[00:20:50] the the bigger picture in order to make the right decisions within that within that one code base so

[00:20:55] that’s how do you encapsulate that knowledge and bring it in and then there’s this specific task

[00:21:01] and the requirements around around what that’s doing so it’s kind of there’s multiple levels

[00:21:05] of context some of those are going to be context which is sort of intrinsic to the um

[00:21:14] the company the business doesn’t change very much some of it is very specific to the task

[00:21:18] and getting that structure in so that you can easily give any given agent the right context

[00:21:24] at the right time that’s context engineering and that’s what makes ai something which which

[00:21:30] really supercharges your your output rather than just something which you know also completes some

[00:21:36] code for you and and so when you you know you’re talking about some of your failed experiment

[00:21:40] experiments where you’d maybe say taken a data engineer or

[00:21:44] a product manager and gotten to try and and vibe code and create something was that part of the

[00:21:49] challenge that they perhaps didn’t understand that full context to be able to prompt it in

[00:21:54] the right way to create something that was ultimately quite useful yeah and and those

[00:22:00] those and those were earlier stage things as well so that you know that the context windows and the

[00:22:06] um the mcp server tools that are available to to bring in cortex that has developed since then as

[00:22:12] well so i i don’t want to be down on the

[00:22:14] individuals involved in in those experiments because they were they were successful in

[00:22:17] in some ways as well they just you know they turned out not you know it’s embracing failure

[00:22:22] as a as a part of this um but yeah it’s it’s that you you do those things you you work out why they

[00:22:30] don’t work you understand okay what what’s the new context and what’s the the changing process

[00:22:35] we need to bring in to make it work and that’s kind of the the bigger picture of how context

[00:22:40] engineering works and and how you change your workflow change your your

[00:22:44] tool set in order to get um better efficiency better output from the agents and enable them

[00:22:50] to do what they do best which is that that core core coding bit okay so context engineering that

[00:22:57] was thing one that that uh the skills that new software engineer engineers need and any others

[00:23:05] absolutely um i say decision making that’s a that’s another key one and this is it

[00:23:14] kind of relates to context switching as well um because it part of part of this kind of empowering

[00:23:21] ai is also putting the right guardrails around ai and getting making decisions at the right level

[00:23:28] and making sure the human is is pulled into be the one who’s the accountable and says yes or no

[00:23:35] to those decisions and it you know used to live in this world where as as a manager as a tech

[00:23:41] lead you could you could define a task you could define a task you could define a task you could

[00:23:44] give it to a junior engineer and you could you could let them run with it for a couple of days

[00:23:48] before you you kind of you check in it depends on the individual obviously um and then you come back

[00:23:54] and you you know you you re re-priorities or adjust their direction and and set them going

[00:23:59] again and you had you had time to line up the next thing you had time to do some more kind of

[00:24:05] strategic big picture thinking before you had to kind of go back to that individual with ai

[00:24:11] it comes back to you in in five minutes ten

[00:24:14] minutes and so you’ve you’ve got the you’ve got the ability to to set lots of lots of agents off

[00:24:20] if you’re experienced with vibe coding you’ll probably find you’re often got you know multiple

[00:24:25] vs code windows up and working on one product here working another one over here and maybe another

[00:24:30] one over there and you set the agent going and it does some stuff and then you come come back to it

[00:24:35] and you’re constantly having to come back and

[00:24:38] you know get back into the headspace for for that task and make a decision

[00:24:44] do some communication get the reaction in place and then set it off again so that efficiency of

[00:24:49] of understanding uh a situation and making a decision and moving on that’s a that’s a key skill

[00:24:56] which you used to be able to do at a much more relaxed cadence but if you’re going to maximize

[00:25:02] your ai native output then you need to be much much faster so getting better at zoning in

[00:25:09] making decision moving on that’s a that’s a key skill that you

[00:25:14] can develop but is is a skill you need to need to adopt to be successful in this in this space

[00:25:21] and also you know you’re talking a lot there’s a lot of context switching there isn’t there so

[00:25:25] you know whereas we started at the beginning talking about deep focus it’s not deep focus

[00:25:28] anymore it’s about and maybe you know you’re not creating the individual lines of code you’re not

[00:25:33] really into the real weeds of everything but you need to understand the high level

[00:25:38] to then be able to make those decisions switch context think about so you’re you’re constantly at

[00:25:43] those difficult high intensity points rather than in the churn out a bit of easy stuff to

[00:25:51] you know that allows me to get some headspace exactly yeah and and that that kind of leads

[00:25:57] leads on to number three which is um being the one who’s who’s actually being the leader showing

[00:26:03] the initiative bringing the ideas and and driving driving the direction so it’s that kind of higher

[00:26:09] level leadership role and bringing that to as many ai agents as you’re able to to control and

[00:26:15] make decisions for and the the ai is not the thing that’s bringing the initiative to your your

[00:26:22] product your output that remains for now a key human skill so being the person who can who can

[00:26:29] do that bring in the ideas identify the things which are blocking that you’re slowing down the

[00:26:35] ai any any kind of bureaucracy that is getting in the way

[00:26:39] and and clearing that and being that you know talking to other teams getting consensus

[00:26:44] putting another tool in place that you know allows the ai to do something which it couldn’t do before

[00:26:50] those those are the skills that you’re bringing to overall kind of tune the machine improve the

[00:26:55] efficiency of of the um the function that you’re doing rather than doing those things individually

[00:27:02] yourself all the time okay so we’ve had three things so far remind me of the three things

[00:27:09] we’ve had um uh context engineering we’ve had decision making and a third thing was

[00:27:14] bringing the ideas and the initiative right and any others and i i think i think those are the

[00:27:23] the most important ones that that you want to focus on we we could we could go through more but

[00:27:29] um i think i think that’s plenty to to get that’s the thing and so that’s a very different job to

[00:27:36] what it was uh a year ago absolutely two years ago

[00:27:39] it’s it’s a it’s a more of a senior role it’s more of a lead role it’s it’s bringing all kinds

[00:27:44] of experience that you’ve got as as someone who’s worked on different projects got battle scars

[00:27:50] knows the things which are going to make a system work and aren’t going to work and

[00:27:55] combining that with you know delegation communication

[00:28:00] bringing teams together and unblocking and and making the making the project happen

[00:28:09] okay so so we’ve thought about what it is we thought about why it’s good we thought about

[00:28:14] the skills that you do need maybe let’s let’s think about some of the challenges and some of

[00:28:20] the objections um the first one that springs to my mind is that’s that’s a completely different

[00:28:28] set of skills and so the people that you’ve got um doing that might not want to be the people that

[00:28:36] are doing this this new thing you know and the

[00:28:39] analogy that sort of comes to my mind is back to the industrial revolution and like weavers you

[00:28:43] know weavers were there we weaving clothes and then with the industrial revolution suddenly now

[00:28:49] we need machine operators and people that wanted to weave were not then people that wanted to

[00:28:55] operate machines it feels like they’re they’re almost very different jobs indeed and that’s um

[00:29:01] that’s potentially a leadership challenge you’re going to have you’re going to have some people

[00:29:05] who are hugely enthusiastic who are early adopters

[00:29:09] who are kind of leading the way in in a new way of working who you find it kind of naturally fits

[00:29:15] a skill set in themselves which is coming to the fore and that’s that’s fantastic and you’re you’re

[00:29:20] onto a big winner and they can become the advocates and uh you know the thought leaders

[00:29:26] within your team you probably are going to have other people who who struggle um who you don’t

[00:29:34] want to adopt a change to something they were very comfortable with or who maybe they want to

[00:29:39] but it’s not their natural skill set and you know you need to work on them in terms of your your

[00:29:45] training your your education try and bring them on the journey um but also you need to you know

[00:29:50] think about your organization and what is the skill set we need in this team and and probably

[00:29:55] bring in some other people who are more aligned with the the new the new normal the new optimal

[00:30:02] for the roles that you’ve got yeah so yeah could be a massive well it is a

[00:30:09] change now the other thing that sort of come to my mind is just that fatigue of context switching so

[00:30:15] you know i’ve never been a professional developer but i do remember earlier in my career

[00:30:20] where i used to debug a lot of code when i was in a support team and i used to have windows source

[00:30:24] code on my desk and we’d be you know trying to debug and you know i’d have a queue of let’s say

[00:30:29] 20 20 cases and i’d be spending half an hour on this one and you and i go right to the weeds and

[00:30:35] then i’d have to context switch into a completely different technology to

[00:30:39] completely different customer a completely different support case that context switching

[00:30:43] was hard and then i do half an hour of that and i’ll be on to the next one that felt like when i

[00:30:49] had a real day like that like my brain was going to dribble out of my ear the mental fatigue of all

[00:30:54] of that context switching and trying to handle all of that in my brain are you seeing that kind of

[00:31:00] thing from from people that are in this sort of ai native world i yeah i can i can certainly

[00:31:05] relate to that and i can say you know we see it with a lot of people and i think it’s a great

[00:31:09] example of that and i think it’s a great example of that and i think it’s a great example of that

[00:31:09] within our within our team as well and it’s again i’d say that is that’s a skill it’s it’s a muscle

[00:31:15] that you can develop it’s um it it’s not something which the brain fundamentally can’t do it’s

[00:31:23] something which the brain often finds difficult but the more you practice it the the better it

[00:31:28] can get um you know a bit like solving quadratic equations or or whatever it is so that’s that’s

[00:31:36] something not to say means this this doesn’t work but it’s

[00:31:39] it’s it’s it’s something which you need to kind of focus on and develop individually if you’re

[00:31:44] someone who’s who’s in that kind of role and and probably the more that you use ai and the better

[00:31:50] you are with it the more that you’ll find then that’s that’s an issue it’s so that this is this

[00:31:56] is kind of it’s kind of the jevons paradox um okay but have you have you kind of looked at the

[00:32:02] jevons paradox and and it’s it’s it’s a kind of analogy that’s been used quite a lot in the

[00:32:07] ai world

[00:32:09] so so jevons was he was like an economist in the 19th century or something and he was he was

[00:32:15] looking at steam trains and he he noticed this weird phenomenon that as the steam engines got

[00:32:22] more and more efficient the total demand for coal went up and up and up so each individual

[00:32:30] journey with a steam train used less coal and yet more and more coal was was needed in the economy

[00:32:39] anyway

[00:32:39] and the reason of course was the total number of journeys increased because that efficiency

[00:32:44] made people start to see in all kinds of use cases for railways and and things you know i can now

[00:32:51] grow lettuces in california and have them in a market in in new york when they’re out of season

[00:32:57] new york because that efficiency has come so suddenly the the railway business um goes into

[00:33:04] into a massive boom if you’re an engineer who is

[00:33:08] you’re an engineer who is good at doing things and you’re good at doing things and you’re good at

[00:33:09] with ai then you are the the steam train in this analogy you’ve suddenly become hugely more

[00:33:16] efficient in in what you can do you’re able to produce much more output because you’re using ai

[00:33:22] to do it therefore the demand on you has gone through the roof and you’re incredibly busy and

[00:33:28] stressed in a way that you weren’t before when you didn’t have this wonderful tool that could

[00:33:32] do your job for you so it’s um it’s hugely kind of frustrating and counterintuitive to to be living

[00:33:38] that reality and i think that’s a really good point and i think that’s a really good point and

[00:33:39] but that’s something that um it we need to we need to get get the balance on and we need to

[00:33:46] to get the value for those individuals and get the reward but also enable them to manage their

[00:33:52] time and their context switching and to do this in a sustainable and achievable way yeah yeah no

[00:33:59] that’s interesting i’d not i’d not come across that before now the other thing or a another thing

[00:34:05] earlier on you were talking about the people who are the people who are the people who are the

[00:34:09] people orchestrating these ai agents really they’re of a more senior persuasion so what i

[00:34:16] mean by that is they understand the context they understand the context of the business

[00:34:20] the context of the architecture those sorts of things so what about the juniors where where do

[00:34:29] the juniors go or where do the seniors come from because you can’t suddenly become a senior unless

[00:34:35] you’ve got that experience of being the junior so so where’s the pipeline or

[00:34:39] where do all those junior type roles go or what happens there yeah and so there was there was an

[00:34:46] article from from microsoft recently um about their their approach to kind of early and career

[00:34:52] developers um and and how they how they see that evolving did you did you see that one and the

[00:34:59] i did this is the one i think that was written by mark razinovich and scott hanselman i saw

[00:35:04] published it and the reason i know because scott has been on this podcast before um so yes i did

[00:35:09] see that they had sort of created an academic paper on their approach to early in careers

[00:35:14] yeah and i i had huge respect for those those individuals but my reaction to that was i i don’t

[00:35:22] really buy it and do you just want to explain what was their kind of their that i’m you know

[00:35:29] i’m going to give you my interpretation of it which is we we need to invest invest in in early

[00:35:35] and career developers they will be at a loss initially because

[00:35:39] you know they’re not as efficient as senior engineers using using ai to to do the job you

[00:35:46] know that’s that that’s that’s an important part of the process because otherwise we’re going to

[00:35:52] run out of senior engineers and when when the seniors all move on we need some someone to

[00:35:59] replace them and the that that that makes sense obviously at a level the the part i struggle with

[00:36:08] is is

[00:36:09] number one i see microsoft and all the other big tech companies laying off thousands and

[00:36:14] thousands and thousands of jobs rather than investing in in new careers for early and

[00:36:20] career developers i’ve not i’ve not seen that happening in reality so i struggle to to believe

[00:36:24] that it’s it’s really the the strategy and and number two that it doesn’t acknowledge the the

[00:36:31] shift in its in its full extent in the way that it should do so the um the role of the senior

[00:36:39] is changing and the the people who are best able to you know drive the the ai function and the the

[00:36:48] agents and things at the moment are the senior engineers but they’re not doing the same job

[00:36:51] that they did six months ago a year ago and they’re they’re evolving into into a whole new

[00:36:57] job and there’s a whole new skill set as as part of that that we’ve talked about the thing that

[00:37:02] we need to train people for is is the new skill set and the new role and not take them on the on

[00:37:09] the same journey that the current senior engineers have been on yeah because you know you don’t need

[00:37:14] to train someone on how to groom a horse in order to have a brilliant you know lorry driver mechanic

[00:37:21] who can function in a modern logistics company yeah so you know time is rapidly escaping us um

[00:37:30] so if people are yet to go on this ai journey what what should people do i guess what sort of

[00:37:38] concrete actions

[00:37:39] what what would be your advice to people on what to do to make sure that they are relevant and not

[00:37:46] obsolete in this ai native team i think my my number one most important thing i’ll put this

[00:37:53] first i don’t know how much time we’ve got um but it’s you you need to put investment in your

[00:37:58] tool chain right at the top of your priority list as an engineering function um and that’s really

[00:38:05] hard to do because you are under all kinds of pressures competitively and you’re not going to

[00:38:09] be competitively you’ve got a product roadmap which is you know way behind where it should be

[00:38:13] and you’ve got all these things that you you you must deliver um and getting that investment in

[00:38:19] an internal tool chain or any kind of tech debt is always a massive challenge for an engineering team

[00:38:24] but it’s it’s it’s never been more important at microsoft we had we had a team called team

[00:38:30] tachyon um whose whose purpose was to make all the other teams more efficient and go faster they

[00:38:36] were our internal tool chain team and we called them tachyon and they were the ones that were

[00:38:39] we called them team tachyon for two reasons one is um that a tachyon is is an elementary particle

[00:38:44] which travels faster than light and so it was all about velocity and making everyone goes

[00:38:48] as fast as possible and the the other reason was that tachyons don’t really exist they’re

[00:38:54] a hypothetical particle because nothing can travel faster than light and it was really hard

[00:38:59] to make that team exist we we donated developers to it from other teams and they kept on getting

[00:39:06] plundered and moved on to other other projects which

[00:39:09] were more obviously hitting the bottom line and delivery even in in an organization that had

[00:39:15] hundreds of engineers just getting three or four to work on that that internal efficiency was was

[00:39:22] a challenge and you want to be putting maybe 20 of your engineering effort into adopting an ai native

[00:39:30] approach to your to your function right now because that is what it takes in order to move

[00:39:39] team and skill set into the native world getting the right mcp servers in place getting the right

[00:39:45] agentic coding framework in place getting people understanding how to use that and and enthusiastic

[00:39:51] and and doing it as their their natural way to approach their work that’s yeah that’s the thing

[00:39:56] everyone’s got to do so change your tool chain to embrace ai native that’s the thing that you need

[00:40:02] to do and need to do firstly with a considerable amount of effort and then there’s all that bit

[00:40:05] that we talked about people need a completely new set of skills that are a little bit different

[00:40:10] um to what people not a little bit a lot different to to what they’ve got today and and then i guess

[00:40:16] that means the people that you’re hiring also need to be different and how you recruit needs

[00:40:22] to be different exactly you’re you’re ahead of me so if you haven’t already then you’re you know

[00:40:27] your recruitment process the the kind of tests and analysis and that you put people through

[00:40:33] that that’s got to completely change so that

[00:40:35] you’re optimizing for the skill set that you that you now need which is is vibe coding is

[00:40:41] understanding how to use agents how to do context engineering and how to create output in in that

[00:40:47] new world if you’re still using leet code and encoding challenges in your in your recruitment

[00:40:52] process then you’re you’re not getting the right people on board because they’re they’re going to

[00:40:58] be massively outpaced by by anyone who’s putting ai first in the way they do their job yeah

[00:41:03] but we’re at

[00:41:05] time pretty much so um let’s wrap this let’s wrap this conversation up so

[00:41:10] what would be your key takeaways from this episode

[00:41:15] um i would say that if you’re on the fence about um how much you you should really put into ai

[00:41:27] versus just delivering what your company’s very good at doing already

[00:41:32] then you need to look very hard at that and you need

[00:41:35] to get yourself into the camp where you’re changing your process changing your skill set

[00:41:41] changing your tool chain and planning for six months 12 months time where you will not be

[00:41:48] keeping up if you aren’t doing these things already that’s that’s number one and number two

[00:41:55] i would say that you need to look at how do we maximize the the value for the people that

[00:42:04] that are working here

[00:42:05] how do we give them the the most fulfilling roles um while also bringing the ai and that’s that’s all

[00:42:15] about about the training and the understanding their motivation and having the right people in

[00:42:19] place and it’s you know it’s it’s fundamentally the most important thing for engineering leadership

[00:42:25] yeah it’s and it’s yeah i guess for me what’s you know interesting is just to hear your your story

[00:42:33] because your industry is going to be going to be going to be going to be going to be going to be going to be

[00:42:35] your profession has been so transformed by ai and so many people it is around the edges you know

[00:42:43] like you talk to a lot of people and you go what you’re using ai for oh well we got meeting

[00:42:48] transcripts and actions generated you know but it’s not fundamentally changed their core business

[00:42:55] and so to see that it’s fundamentally changing your core business and so to tell to hear about

[00:43:02] those war stories and the lessons that you’re learning and how much

[00:43:05] you’re investing into it and then the benefit you’re getting from is is really quite interesting

[00:43:10] absolutely and it’s how do you take that into the the other business functions as well

[00:43:15] because it’s not just going to be software that is is transformed in this way it’s it’s it’s

[00:43:20] everything that’s the that’s the next step that’s brilliant well thank you so much if people have

[00:43:26] really enjoyed what you’ve been saying and want to get in touch with you how can people get in

[00:43:29] touch with you nick so i’m i’m terrible with um you know

[00:43:34] you

[00:43:35] emails phone calls that that sort of thing so um apologies if you’re trying to get in touch with me

[00:43:41] and and and struggling i would say that the most effective way is probably if you get a number

[00:43:47] salesperson on a call and give them the impression that you want to spend a lot of money on some

[00:43:52] maybe some bespoke api work or something then they will inevitably pull me in

[00:43:57] but um i i don’t i don’t really want you to to waste time in our sales team

[00:44:05] linkedin is the place and that’s that’s the that’s the best way to find me and engage me absolutely

[00:44:11] and i’m going to be at tech show london so please do um please do find me there and i’d love to

[00:44:16] link up and have a coffee and i’d love to hear everyone else’s experience and all the people

[00:44:21] are thinking no you haven’t tried this or what about this i’ve yeah i’ve got to hear those

[00:44:25] stories as well please and please bring them on yeah brilliant so final thing for me to say is

[00:44:31] thank you so much nick for taking the time out your really busy schedule to to come and have

[00:44:35] this conversation and share all your experience with us it’s been a joy thank you ben always a

[00:44:40] privilege and um all the best don’t go yet remember to subscribe to the podcast and rate

[00:44:47] the show it really helps us grow and book new great guests and remember if the podcast isn’t

[00:44:53] enough for you and you want weekly micro learning delivered straight to your inbox

[00:44:57] sign up to the tech world human skills weekly head over to www.techworldhumanskills.com to sign up

[00:45:05] and we’ll see you next week

Visão de gráfico