[00:00:02:02 - 00:00:23:01]
Hi there, I'm Ben Pearce and welcome to the Tech World Human Skills Podcast. Every episode we talk through how to thrive in the tech world, not just survive. Now, if you want me to work with your team, just give me a shout. I love to help teams be more influential, memorable and successful with their stakeholders. Head over to www.techworldhumanskills.com to book a chat.
[00:01:22:10 - 00:01:29:07]
Thank you. I'm just casually applauding myself there, but thank you for having me.
[00:01:40:20 - 00:02:56:04]
Yeah, sure. So Steve Chan, always obsessed with technology. So been on a lifelong journey, kind of learning, using, adopting technology. Did a computer science degree naturally, spent a lot of my career in government in consulting. So doing digital transformation type projects, building capability for other organizations, centering around AI or AI at the time, things like natural language processing, image recognition, etc. As you mentioned, I spent the last five years just left Microsoft most recently as an AI solutions engineer. So part of Microsoft during that huge LLM boom, where we saw large language models, multimodal models come onto the market, explode in terms of things like chat, GPT, copilot, etc. And I've now left to the dark, scary route of forming my own organization with another person. So I'm co-founder of a company called Stealth Labs, where we're looking to really get out there and help organizations adopt, use AI, and again, build bespoke capabilities using AI tools and techniques that are out there.
[00:03:06:23 - 00:03:09:10]
I think we just missed each other, didn't we?
[00:04:19:01 - 00:04:36:07]
Yeah, so we've actually launched two things. So Stealth Labs is our consultancy vehicle. And that is very much about human interaction. So getting out there, doing the business change and the business transformation of building capability and getting that out to organizations,
[00:04:37:07 - 00:04:55:06]
ideally across the globe, we'll start small, just UK for now. We have actually also built a product, which we're calling Miki, which looks to take some of the, as you described, the more process-driven
[00:04:56:08 - 00:05:11:09]
outputs of consulting, the more research-type tasks, the document creation, the, dare I say, more boring-type, back-office-type things, and automate those using AI agents.
[00:06:17:01 - 00:06:59:12]
Yes, it's a good question. And I think consultancy means a different thing to different organisations and different people that you speak to. And there's obviously different bits of consulting. We're focusing more on the technical consulting offering and that sort of things. But generally, you'd hire a consultant or you'd hire a consulting organisation when you need to adopt change within your organisation or you've recognised a problem and you as the underlying user or the end organisation don't know how to solve that. So you're bringing in external expertise into your organisation to try and help you work towards a solution, solve a problem.
[00:07:00:19 - 00:07:13:00]
And that may be because you don't know how to do it yourself. It might be because you just don't have the time, the resources to dedicate internally to solve something and therefore you want to bring someone externally on board.
[00:07:14:04 - 00:07:40:21]
In terms of breaking that down and what that looks like, you can start to get quite, what's the word? It's a bit of algorithmic, but that's probably a bit too specific. But you can get quite process driven, I think, in terms of what that looks like. You define and understand the business problem. You do that by interviewing the users, asking them what their issues are, what the problem is that they're trying to solve and really what those pain points are.
[00:07:42:00 - 00:07:43:22]
From that, you might do some market research.
[00:07:45:00 - 00:08:08:22]
You might go and look at what's already out there. Are there off-the-shelf solutions that you can implement? Is there something in the open source community that you can fork and build? Does this need to be bespoke? So you do a bit of ideation. You think about what that solution might be. You might play that back to the users in language that they understand to get the go ahead and say, "Yeah, that sounds like the right kind of solution we want."
[00:08:10:08 - 00:08:33:16]
You then probably need to do some kind of high-level architecture designs or low-level architecture designs. So for your listeners who've never come across that term before, really just drawing out the components of a system on a piece of paper, making it abundantly clear what it is you're going to be delivering, what resources you're going to be using. Are you going to be using the cloud? Are you going to be using on-prem?
[00:08:34:22 - 00:08:39:07]
Are you going to be building your own models? What databases, et cetera, are part of this solution?
[00:08:40:22 - 00:09:26:07]
You probably want to take that through a security review, through a stakeholder review. Go on and build that solution. And then you start thinking about business change and how you bring that new solution to users. So how do you train them? How do you onboard them? How do you bring a new piece of technology to a workforce and enable them to start with success? So giving them the right tools, the right training to be able to understand what that tool is, what that new application or new service does, and allowing them to really hit the ground running. And I think when you break it down like that, yes, there is a human element. And I don't think right now anyone is saying that any of these services are going to go away. But a lot of that is process and research driven.
[00:09:27:15 - 00:09:50:09]
Unlike any algorithm with enough good source data, with a history of this is what a high level design looks like, this is what an ideation document looks like, this is how we might interview users, these are the sorts of questions we might ask. You can start to kind of see in your mind how you maybe start automating part of that.
[00:09:53:20 - 00:10:16:09]
Yeah, so it's still very human driven. But the idea is previously where you'd need a huge team, a consulting team to do different functions. Instead of scaling with people, you can have a smaller team that scales with AI agents.
[00:10:17:15 - 00:10:31:14]
So what I mean by that is building the documentation based on the human input, doing the research, the market research, the industry research automatically, and then presenting things back to that consulting team.
[00:11:01:03 - 00:11:11:10]
Good question. I think again, we're talking about document generation research and outputs based on inputs that humans are still providing.
[00:11:12:11 - 00:11:21:13]
So I don't think we're saying right now that everything is going to be automated, although later on I think we will talk about a potential world and what that looks like.
[00:11:22:14 - 00:11:54:05]
But if your listeners and viewers can imagine how people have adopted chat GPT, and really that's enabled individuals to be able to research and ask questions on a level that they haven't been able to do before, either because that was completely time consuming, having to search through standard search engines, look through pages and pages of results, scout through web pages, et cetera, or maybe because they didn't actually know what they wanted to ask and they needed something that was going to bring suggestions to them.
[00:11:55:07 - 00:12:02:00]
If we apply that to the consultancy space, then yes, having humans on the front line is still powerful.
[00:12:03:08 - 00:12:40:10]
And at the end of the day, humans still want to interact and talk to humans, but taking things like notes or transcripts from user interviews and turning that into an output that is presentable and can be signed off within an organization, or ideating with an AI system where maybe you've got some ideas, but you need that augmentation of what else is out there. And very rapidly being able to research pretty much the entire web very quickly allows for some very high end and powerful results.
[00:13:13:12 - 00:13:22:00]
Exactly. So the human is still driving the decisions. This is what we want to implement. We understand the organization. We understand the users.
[00:13:23:18 - 00:13:27:20]
You could argue that there are hundreds, thousands of solutions to potentially any problem.
[00:13:28:21 - 00:13:52:02]
So bringing that down is down to the human knowledge and understanding of the organization that they're working with. But in terms of building those high level designs, low level designs, building applications, building training plans, training packages, business change plans for a human to then go and implement, we can very easily and very quickly now generate those using AI agents.
[00:13:52:02 - 00:14:22:15]
(Silence)
[00:14:24:14 - 00:15:25:18]
It's interesting because I think the complexity and the knowledge within the human brain, of course, is great. I'm a bit biased myself being a human. But there is a finite limit to any individual's knowledge. I think, you know, I always hear the story of when chat GPT first came out, I think they were running GPT 3.5 or GPT 4 with that application. Everyone was like, "This is amazing. It's going to be the end of the world," etc., etc. Three to six months later, everyone kind of got over it. New versions have come out and different tools have come out, and people have had the same reaction to that tooling. I think what we're seeing right now is, yes, those integration points are going to be part of the human mind. That deep, intricate understanding of an external organization, where the pain points are, where things tend to fall over, is going to come from that human consulting angle.
[00:15:26:18 - 00:16:07:18]
But as time goes on, who knows? We're inevitably training these models with more and more information. And yes, they may have a generic view, as in they may have a wide-ranging view, as opposed to a view just of your organization or the organization that you're working with. But eventually, those data points will become enough for AI to be able to think about, "In these sorts of industries, I tend to see these issues," or, "I've been trained on these issues, and therefore to stop these issues coming up or to prevent these issues, I'm going to design these things into the integration that I suggest."
[00:16:51:18 - 00:16:54:07]
Yep.
[00:17:54:21 - 00:18:15:02]
I think you've hit the nail on the head. And the slight elephant in the room there is, consultancy is actually quite inefficient, I would say, when it comes from a pure business perspective. So I like you say, if you think about the big players, obviously you won't name any of them for the sake of saving you from being sued on the podcast.
[00:18:16:18 - 00:18:29:19]
But generally, the way consultancies work is they may have an overhead team, but the majority of people will be on what they call fees. So those people are sold out to organizations to do a role.
[00:18:31:05 - 00:18:59:06]
You then start getting, as you mentioned, overwhelming pressure from those running the business because they're looking at profit margins, et cetera, to keep people on fees 100% of the time. So as a consultant, and I'm sure you remember this, you end up working for one organization, but you actually spend all of your time in another organization doing a job within that organization that you've been almost sold out to.
[00:19:00:19 - 00:19:03:22]
And then that begs the question of how do you train, how do you upskill?
[00:19:05:01 - 00:19:21:04]
How do you, from a business perspective, ensure that you can answer new requirements, you can expand. To date, consultancies have expanded with people, with individuals, which then puts them around that cycle of introducing those problems again.
[00:19:22:05 - 00:19:49:23]
With AI, the best way that I would frame it is what we have right now is junior level consultants or junior level consulting type outputs that never have to sleep, they never take a break. You don't need to fund their upskilling. LLMs are trained on pretty much the entirety of the open internet. They can reach out to external sources, they can look through any data that you provide within your organization.
[00:19:51:04 - 00:19:54:17]
So you don't have to pay for onboarding, you don't have to pay for training.
[00:19:56:02 - 00:19:59:18]
They don't take time off, they don't take breaks, they can work 24-7.
[00:20:00:18 - 00:20:08:06]
And I'm aware that I'm describing a bit of a dystopian future there, and we'll get to that.
[00:20:09:10 - 00:20:23:08]
But I think it's, I mean, well, you tell me, but in my mind, I think it's plain C from a pure business perspective of where that would then allow you to have, let's have some specialized, perhaps more expensive consultants,
[00:20:24:09 - 00:20:41:04]
but a lower number and a number that we can easily manage and augment them with these AI agents, as opposed to going out and hiring tens of thousands of people and having to worry of where you're selling them, where they're going, how do we train them, how do we upskill them, etc.
[00:22:23:19 - 00:25:45:20]
Yeah, I think the first thing to think about, and I wish some of these AI offerings existed back when I was at university, is that for the first time we've probably got a very easy way for anyone on the planet that's got access to an internet connection to be able to upskill in any topic and to be able to use artificial intelligence to further themselves without having to, dare I say, and I say this as someone with a degree, but go through expensive university courses or pay for external education. There's almost a free source there of unlimited education. You can talk to AI today either by text or with voice and ask for bespoke training courses that cater to you, that speak to you in a way that you understand and you learn. I think one of the things that we have realized as a society recently is that maybe the way that we've always educated people in the past doesn't work for every single person, and those people that potentially fail within the British school system, etc., they're not necessarily, for lack of better term, unintelligent. They just haven't been taught in a way that resonates with them or that they can receive that information. I think the first thing to come home is that AI presents a very big opportunity for everyone, potentially across the planet, to be able to learn and upskill. In terms of where it's going, I guess this is all opinion-based and this is where we get into, maybe you go back into your PS Morgan-esque style critiques, but personally, what I think we will see is more and more smaller organizations where people have gone through education, they've educated themselves, they've picked a niche, and maybe they're going out on their own or with their own small organization and educating, doing transformation, helping organizations with that niche. That niche may be artificial intelligence, it may be things like cybersecurity, which is still a big component of what happens out there. Securing organizations, IT and systems is a big thing that consultants do today. I think what we will potentially see less of, I'm talking 5, 10, 15 plus years, are those huge consulting firms who hire tens, if not hundreds of thousands of people and effectively ship them out to external organizations. Because one, that power of learning also goes to those organizations too. If you're an organization that hires consulting today, you can probably get a very base-level understanding of any topic that you want to think about through AI without having to hire a consultant. That's something that's already happening.
[00:25:47:23 - 00:26:23:02]
What I think we'll see is consultancy be called upon for very niche-specific topics or subjects or areas. I've got a very good understanding, for example, in how to implement cybersecurity within the French government in the south of France. Okay, great. That's a skill that maybe people will look to consultants to use. I think we'll see a lot more of those niche-type consultancies as opposed to their larger ones.
[00:28:04:07 - 00:28:30:16]
Exactly, exactly. Building on that example, I think that there's loads of examples of things like that today that people are putting into practice and people are seeing. From my own perspective of co-founding a new startup, maybe go 10, 15 years ago, maybe even more recently, five years ago, we'd probably be looking at onboarding external advice to help us with things like
[00:28:31:23 - 00:28:53:14]
the statements that you've just described. Or how do I pay tax? How does the company's house website work? What is this CH12345 form that I'm expected to fill out? That kind of knowledge, I guess, was ring-fenced within a bunch of specialists who had done it time and time before.
[00:28:56:03 - 00:29:00:19]
Don't get me wrong, for the sake of the record, we have obviously sought the right advice.
[00:29:01:21 - 00:31:04:18]
It is very easy to go and use AI to generate documents, the kind of documents that you need to file to ask those, dare I say, stupid questions that you really don't want to admit that you don't know the answer to. The AI isn't going to judge, it isn't going to be like, "Ben, that's a stupid question." Of course you need a bottom slavery statement. Why are you even asking me this question? You don't get that kind of response. If you apply that to the consulting world, if I pick a specific example, some customers out there, some enterprises, some organizations, they just want to define their requirements. The answer to that previously has been, "Well, have a team of business analysts, have a team of business consultants." They'll go out there, they'll interview your users. Again, they'll identify those pain points. They'll ask the right sort of questions to draw things out of their users. But if you think of even today, I can have a conversation with chat GPT that flows pretty well. I can start going down a route where I talk about, "Here's a pain point that I've got in my life. Can you help me develop it? Can you help me articulate what that problem is?" It will go backwards and forwards. It will feel like it's bringing something with me. It will ask the right sorts of questions to be able to capture those things. If I just say, "Capture the output of this in a standard business requirements document, BRD," which would be a standard consulting output, I will get something that is good enough. It's not going to be right now the super polished version that I don't need to go and review and check over and maybe go, "Hang on a minute, that bit looks a bit wrong. Let me change that." But it's going to be good enough to get me to a point where I've now got something that articulates my requirements and I can take that to the next stage, whatever that next stage may be.
[00:31:16:15 - 00:31:28:15]
(Silence)
[00:31:33:15 - 00:31:56:15]
It's a good question. We brainstormed about this a couple of weeks ago when we had the prequel. I think the uncomfortable truth in general in many industries is that many jobs are potentially going to disappear. I want to say that because I don't feel like enough people are saying that.
[00:31:58:01 - 00:32:18:22]
If you look at the headlines today, what you'll see is that this speeds people up. This helps people achieve more, but not that AI can potentially replace people and jobs. There's good reason for that. We don't want to cause widespread panic. But I think ultimately
[00:32:20:09 - 00:32:24:14]
in the capitalist society that we live in, and this is where I get slightly political,
[00:32:26:05 - 00:33:16:09]
every business, every organization is always looking to cut their costs. We've seen that in big tech in the last five plus years. If you look over to what's happening in the US, there's cuts all the time. Well, there's cuts all the time worldwide as organizations become more and more efficient. I think, again, the best way to handle that today is to learn how to use AI, to learn how to use these tools alongside what the thing is that you want to do. If you want to be a software engineer, for example, understanding how tools like Cloud Code or GitHub Copilot can help you become a better developer, I think will become very attractive to employers in the future. I do think there's potentially going to be a point where
[00:33:18:07 - 00:33:57:03]
I don't want this to happen, and this is going to sound slightly harsh, but you're either on the bus or you're not. As in when development organizations, when startups, when all engineering houses realize that when they used to have a team of 15, they now only need a team of seven. If those seven people efficiently use AI tools, then the role is going to go down the route of, you can have this job, but you need to know and learn and be able to adapt to using this tool alongside your role. So I think today the best thing to do, again, is to utilize the learning experience,
[00:33:58:03 - 00:34:07:15]
work out how your job can be augmented or how your role can be augmented with what is coming on the horizon and prepare and adapt yourself for that.
[00:35:19:14 - 00:38:17:13]
I think so, I think so, but it's difficult and I don't want to come across as all doom and gloom and bear in mind to your listeners and viewers, I'm in that position as well, right? So, you know, as we've covered, I've just launched a consultancy. So, if anything, that's why this is top of mind. But if you think about how, like we call it human computer interaction has evolved over time. Right now, when people are interacting with AI, they're either typing something into a prompt box and by AI, I mean that the LLMs are chat GPT type applications of the world. They're asking a question by typing something out or they're clicking on a button, probably a microphone button and they're speaking to it in a similar way that we're having this conversation. And you can look at that and go, well, people are always going to want to talk to humans. So, of course, that human role is always going to be there, but this is evolving at impressive a crazy rate. And the next thing is maybe we have virtual avatars, maybe virtual avatars that look lifelike and that can dial into Zoom calls, team calls, et cetera. Microsoft Teams already has a virtual avatar function, but that just replaces you and your video feed. If that then becomes a vehicle for an AI model to be able to speak to you and talk back to you, then that's kind of a next stage in the evolution of how you interact with computers in general. Give it five, 10 years. We've already seen, I think there's quite a comical video of a robot that's recently pledged to launch that isn't looking quite well. There's a video of something loading a dishwasher and it can't work out how to kind of close the dishwasher type thing. But again, give that 10, 15 years, who knows, maybe we'll have autonomous androids. I mean, Elon Musk keeps promising this Tesla bot. So maybe that will become a thing. And I think as more R&D, as more research, as more and more goes into this, the concept of speaking to a human versus speaking to a AI or a machine or a robot or whatever you want to call it are going to become quite blurred and very similar ways of interfacing with having a conversation, shall we say. I mean, imagine the kind of AI robot type situation.
[00:38:37:16 - 00:39:23:05]
I think the key takeaway is let's not panic. AI presents quite a huge opportunity to many people to be able to upskill, to be able to learn and to be able to do things that previously they may have had to hire external expertise or different organizations and to be able to go out there and achieve big things by themselves. So take that, use AI to learn, use it to upskill, think about how in your day job you can adapt, think about what's coming, where it might go, stay at the bleeding edge and just continuously try and learn. I think that will be my key takeaway here.
[00:40:21:01 - 00:40:44:09]
Best way would be to go to stevechan.ai. That's my kind of personal website. There will be a bunch of things on there. There is a speaking section of which this podcast will be linked. I'll put together some thoughts based on what we've spoken about today, maybe some articles that reference some of the things we've spoken about. And there's some easy ways of getting in contact with me and bucking some time in my calendar there.
[00:40:56:07 - 00:40:56:13]
Thank you
[00:40:58:19 - 00:41:18:22]
Don't go yet. Remember to subscribe to the podcast and rate the show. It really helps us grow and book new great guests. And remember, if the podcast isn't enough for you and you want weekly micro learning delivered straight to your inbox, sign up to the TechWorld Human Skills Weekly. Head over to www.techworldhumanskills.com to sign up.
