The magentIQ Show Ep. 2 | The Operator's Edge: Why Foundation Models Need Frontline Wisdom

Co-founders Ian Barkin and David Brain have been having these conversations for fifteen years. This time, the recording light is on. Welcome to Episode 2 of The magentIQ Show, and the first installment of our founder discussions.

Wall Street just placed a very loud bet on a quiet truth: enterprise AI is hard. Anthropic has teamed with Blackstone, Goldman Sachs, Hellman and Friedman, and General Atlantic to stand up an AI services firm aimed at private equity portfolio companies. OpenAI is making the same move with even more capital behind it. The takeaway is unmistakable. Powerful models like Claude and ChatGPT are no longer the bottleneck. Deployment is. Messy workflows, missing process documentation, and the reality that ninety five percent of proofs of concept never reach production.

In this episode, Ian and David unpack:

• Why Anthropic and OpenAI are buying their way into the enterprise and what their new consultancies will inherit
• The bookends dilemma: most companies lack the process foundation to automate and the imagination to deploy AI ambitiously
• AI coding as a two to three times accelerator for senior developers and a liability when treated as a replacement
• Service as software and outcome-based pricing as the commercial shift the industry has needed for a decade
• How input-based pricing in BPO and managed services actively punishes innovation
• The boardroom hype problem and the cost of swapping "gen AI" for "agentic" without a plan to back it up

It is the kind of honest, peer-level conversation The magentIQ Show was built for. No hype, no fluff, just two co-founders who have been doing this work since before it had a buzzword.

Listen now and tell us where you land on the great enterprise AI deployment debate.

Find us in your favorite streaming service: 

iTunes: https://podcasts.apple.com/us/podcast/the-magentiq-show/id1896570951 

Spotify: https://open.spotify.com/show/033f2oKxnFr5fmcpdSHjSL 

iHeartRadio: https://iheart.com/podcast/333292440

Got a perspective worth sharing? We are always looking for guests. Reach out at info@bemagentiq.com.

Transcript

Welcome And What We’re Tracking

SPEAKER_00 0:00

Anthropic and a whole bunch of big private equity firms have decided to launch a new AI services firm. The gist of it was to help the portfolio companies of the private equity funds involved. Welcome back to the Magentic Show, the soon-to-be world's greatest podcast on work life and culture in the Agenc era. I'm Ian Barkin, and today I'm joined by my co-founder, partner in crime, and the most grounded voice in this whole hype cycle, David Brain. This week, we cover Anthropic, teaming up with Blackstone, Goldman, and half of Wall Street to launch a consulting firm. OpenAI is also making big moves to chase the enterprise, and we talk about that. Boards are swapping Gen AI for a Gentic without truly knowing what either one actually means. And somewhere in there, somebody actually has to make this stuff work. David and I explore all of this and more in this episode. So buckle up, don't forget to follow, and let's get into it. Welcome everybody. Welcome back to the Magentic Show, episode two. Thrilled to have you here. We had a great discussion last time, and this one is going to be, dare I say it, as good, maybe even better, and it is going to be an exploration as we set out to an all episodes of what will soon be the world's greatest podcast on the impact and use of AI in work life and culture, uh, the magentic show. Uh, I'm going to have a discussion with my my friend, my partner, my longtime colleague and co-founder David Brain. David, welcome to our show. Thank you very much, Ian. It's a pleasure to be here. So we we have these discussions constantly, you know, late nights, early mornings. You're you're in England, I'm in the US, so our times are always sort of a little weird and off. Constantly analyzing headlines, stories, trends, what's happening. We have been at this for a while. We've known each other for a hundred and five years. Uh, I don't check that math, but that explains it. It's yes, exactly. That that explains a lot. We look great, by the way. You look tremendous for forwarding each other. And we have been in some way, shape, or form, involved in the application of AI and automation in the world of work for most of the time that we've worked together and we've known each other. Uh, we've started things, we've invested in things, and we certainly uh spend a lot of time talking about and analyzing what hits the headlines. Frankly, because we're just trying to figure out what the heck is going on. And so thought what better way to do that than in public and have people be able to join and listen in and comment. Um, please do get in touch with us if you think our ideas are absolutely uninformed, crazy, and incorrect, or or spot on and absolutely inspired and brilliant. We'll we'll read the second uh second group much closer, but um, but please please do let us know. Um, David, we've got some headlines we're gonna we're gonna just drain. We're gonna talk about each one of them. I'd love to hear your perspective. You are always so grounded and and have such a solid perspective on it.

Anthropic And PE Launch AI Consulting

SPEAKER_00 3:20

The first one that literally everyone is frankly tweeting and posting and and blogging about it was a big one, which was uh a few days ago, um Anthropic and a whole bunch of big private equity firms, lots and lots of money, Blackstone, Hellman and Friedman, Goldman Sachs, uh, and and some other big PEs as well, um, General Atlantic and others were named in the uh in the announcement, have decided to launch a new AI services firm. Um, it sounds like a a new consulting firm. It seems like the gist of it was to uh to help the portfolio companies of the private equity funds involved. So all these PE firms buy companies, so they've got a big portfolio of companies that they own that they want to see operate better because ultimately they want to sell them for even more money. They want to help them to use, in this case, one specific set of tools, the ones that are made by Anthropic. And so they've set up this whole consulting firm to help them do that. I mean, where do we start, David? What do you think about the announcement first and foremost? Let's let's unpick this because we've got a lot of opinions.

SPEAKER_01 4:32

Absolutely. I mean, I think it's fascinating, um, and just goes to show just what a fast-moving sector this is and how quickly things will change. Uh, there's always something to talk about each week, so that makes what that's what's make gonna make this a lot of fun. Um, so this came out around the same time, and I'm not sure which one broke first, but uh perhaps you've you'll be able to tell me. But there's there's two announcements, right? We've got this one and a similar one from OpenAI uh for the consortium they're building. Um, and I'll give you, you know, I've read some read some articles, but also just give you what what I think is going on here. Um a partially formed opinion. Uh so makes it more interesting. Absolutely rip it apart. But I mean OpenAI, and I do wonder if Open and AI OpenAI was the one that sort of you know uh pushed the timeline on these releases, you know, suffering last week, a lot of news around their somewhat disappointing um figures for the year in terms of user growth and retention, and um the struggle really to stay in in that growth phase with new uh enterprise and business customers. They're they're still the market leaders, far dominant when it comes to personal individual B2C type subscriptions. But where they're struggling is getting into the enterprise space. And I think I think that's visible to us at the moment for OpenAI, but I'm sure it's a common theme that is felt by both OpenAI and Claude, perhaps less so with uh Google, just because the Gemini stack is kind of rooted in that um that enterprise uh space uh with vertex tools and and the like. Right. Um and I I see that this is a way in which these megafirms, these foundational model providers are looking at how they can break into the enterprise. And if you're able to get the, you know, rather than going and and pitching to the individual sort of enterprise owners, if you can go and uh go to the companies that own them, then um I think that's really where their strategy's at in terms of getting in front of many enterprises in one go. And also, I mean, we we've had a lot of experience when it comes to PE firms, and basically PEs will operate in one of two models, either to take a majority controlling share or a minority. And I think what's really important in pursuing this model is that this is uh this is a play for the majority-owned and controlled PE um portfolios, uh, where they're going to be able to have the biggest impact.

SPEAKER_00 7:00

Outstanding. And that's that's really interesting because because you're right. I mean, they're they're they're making they're making a lot of money, they're losing a lot of money, but you know, in the the world of Silicon Valley, you you spend a lot to make a lot and your valuations grow accordingly as you finance uh next and next waves. Um it is fascinating. OpenAI, um, as you said, market leader anthropic has sort of roared uh in the last few months up to what is it, 30 billion in revenue. So they're a beast and very successful, and a lot more people, a lot more people are exploring using their tool set. Um, but what's so fascinating about this announcement, especially for you and me specifically, is it's it's it's effectively saying that our we believe our our technology is great, nobody is deploying it, no one knows how to, and it's not being adopted at a rate that we need it to be adopted for us to be able to stand by and justify our valuations in any way, shape, or form. Now, uh it's got a lot of got a lot of opinions on this. I mean, I'll I'll literally I'll read you a paragraph from from the anthropic announcement for a moment, just because again, and just just before yeah, sorry, just before you do, I think it's I mean, just so that we're we're clear on our our reference points here.

SPEAKER_01 8:18

I think there's three key markets. You've got the sort of direct-to-consumer, albeit a lot of business customers, and that's the kind of you know, the the assistant um model that is being deployed so so well. And open AI are the leaders there. You've got um your coding uh base, which has really, you know, exploded in the last year. Um, and that's really where Claude has taken a large dominant share. A lot of talk now about under Spud with um uh OpenAI picking up um a lot of a lot of uh users with the latest model. So we'll have to see how how quickly people can navigate and port between different models. But there's there's definitely a battle going on there, and that battle is definitely within the enterprise space. But where we see this really playing out, and where Ian and I we're so focused and have been all our careers, is on the business processes that are performed within enterprises. So, how are you rolling this out within your insurance company to handle claims or applications? How are you doing this within a bank for new accounts? For any business, how are you doing this to handle uh how you're using AI to do um back office or front office functions? And that's the area where there's some stagnancy in the market. You know, they've they're not really finding enough use cases. A lot of initial pilots get started, very few of them make it into production. And I think that's the area that is really under focus now for both anthropic and open AI.

SPEAKER_00 9:46

Yeah. And and you know, all the power to them, I get it, and we get it. We get why they're doing what they're doing, is they they they they realize that while their tools have some pretty tremendous capabilities, it's the act of rubber meets road, getting it into those enterprise environments where uh you do need to understand process, and and there's a lot more to it than just plugging automation in to replace, right? You have to understand what it is that people do, you have to re-engineer heavily before you just sort of you know pave cow paths, as we've we've long said. What's fascinating to me about the announcement, too, is and and obviously, obviously, these are the world's smartest people and they've got all the resources in the world. So um there's a lot that just it's not that it isn't happening, it's just it was not shared in the announcement. Was you had a bunch of the PE folks quoted as to the importance of this and why it's gonna be big. Uh, you you don't hear from practitioners. Um, I imagine they're gonna use an anthropic's um case 1.5 billion and open AI is is like 4 billion. They're gonna use that money to go buy consultancies, they got to buy up capabilities to actually put this these teams together. Um, I don't know who those early kernels are, like who's where is the where is the starter um in in the the middle of that? And then then what they're doing is it heavily depends on heavily leans on anthropics team. They mentioned anthropics, uh what I think we're all calling now for better or worse forward-deployed engineers, which are really just when we saw this 15 years ago in in the robotic process automation space, is you can have sort of technical sales folks who show up to do rapid prototypes to just prove out the fact that this stuff works in your environment. They never ever stick around to turn this full-scale production ready and and fully up and running, right? That's just not their job. And so yet they seem to suggest that Anthropic in this case has an infinite number of these folks just waiting around. Um, the release itself, and I'll just I'll read this part and then rant a bit, and then we'll move on. Um, but the release itself says, like an atypical engagement starts with a small team working closely with the customer to understand where claud can have the biggest impact. From there, the company's engineers, alongside anthropic applied AI staff, will develop cloud-powered systems tailored to each organization's operations, right? That that explains nothing new ever, right? That like this is what we did 15 years ago, that a small team sits down next to your operators to figure out what you actually do. I mean, how else would you do it? Don't just like don't just rock up to a company and just and guess and throw something at their operations. Um, it it gives a specific example. Consider a multi-site healthcare services group, like a network of physician practices. Clinicians spend hours each day on documentation, medical coding, prior authorizations, and compliance reviews. And engagement might begin with the company's engineering team sitting down with clinicians and IT staff to build tools that fit into the workflows that staff already use. I mean, uh I mean, yes, but but no credit for saying the most obvious thing in the world, that's why this stuff's not working, because because they're not spending the time doing that, and they haven't explained how they are going to do that, and you know, and honestly, not to pat ourselves on the back in any way, shape, or form, but like it is exactly what we've been doing for 15 years. It is the last company we set up was just experts. I mean, not just doing that, but experts doing exactly that in exactly that setting, actually in healthcare, actually in prior auth, actually in documentation. Coding. Yeah. And it's funny to watch watch this the a whole group of folks sort of come around to realize that ooh, this is the gap we need to fill. Um, I just I'll be curious to see if they they know how to fill it, right? Because they need to go and get specific skills, not to pat ourselves on the back again. But like they got to find the right people who aren't just out there just um parading whatever tool they're they're put in position to hawk, right? You're there just to move one set of tools, so you're gonna be showing up with hammers looking for nails.

SPEAKER_01 14:06

I mean, honestly, both both tool sets are so capable. I'm sure they'll find good uses for the toolkits and the models. But it is interesting how they're using anthropic staff to do the heavy lifting, because what I think is really needed in this market is someone to be picking up on the um uh on the deployments, you know, and the the consulting piece is definitely key, but it it has to be based on a solid understanding of what works in deployment. So um you'd really want those teams integrated as much as possible.

SPEAKER_00 14:41

So there you have it. So that's the news that's dominating headlines right now is uh at least uh five point something billion dollars uh dedicated to open AI and enthropic creating AI consultancies to do what frankly David and I have been doing for 15 years and knew was necessary, and they're coming around. So I guess that's that's encouraging. It's a question of um do they have the right uh ability in remit? A whole other um element

Why Enterprise AI Deployments Stall

SPEAKER_00 15:10

of that discussion then comes down to the uh the audience, right? The the substrate in which they're gonna be doing this, which is portfolio companies of private equity firms. And so um that's a whole other um set of dynamics to consider. Uh ultimately we've we've helped and advised portfolio companies of of private equity firms in the past. Um, it'll be interesting to see whether those those portfolio companies are are being asked to to pay for this service or if this investment is just gonna subsidize all this work. So hey, free consulting. And or um, and if if the PEs have the authority, are they majority owners? Are they gonna be the ones who get to just sort of declare you're going to be doing this, that is the only choice, or or is it what we've seen more in sort of the minority ownership structures where it has to be through sort of influence and um and yet it's still up to the operational leadership of the portfolio company. And that's always friction and uh and uh constraint that slows down any project. And I guess to extend the discussion even more, I mean, David, something that you and I talk a lot about, we'll continue to talk a lot about, which is uh just because AI is there does not mean you're ready for it. And no, absolutely, absolutely.

SPEAKER_01 16:27

I mean, I I do I do applaud both companies for being bold and going after the business processes for automation, because that is where we're gonna find it. And quite frankly, that find the true value of AI. And quite frankly, that's the area where uh organizations are struggling the most at the moment. Um, so it's you know, it's borne out in the 95% of POCs not making it past um, you know, the proof stage and into deployment. Um, so you know, I do think this area does need a bit of a shake up. Now, what'll be interesting, and I I do think it comes down to um the a few things, I think the AI ready point is absolutely you know spot on, and that's where consulting really can help, because I think a lot of companies will have a rough articulation of their business processes, um, but it will be out of date or it will not be complete. And time and time again, I mean, when when Ian and I, first time round, when we were doing this as part of um Symfony Ventures, we found that a lot of clients uh would tell us um their processes and we would map them out. And then when we went into deployment, uh we would find that there were new business rules, and so there'll constantly be changes. This just because all of those edge cases are really hard to code for. Now that's fine if you're in uh more of a um predetermined automation, perhaps there's a better word for it, but robotic process automation and the like, um, API driven automation. Deterministic, um deterministic. Yeah, because um in that in that setup, you code for the for the paths that you want to automate, and then you can throw everything else out an exception. So what you typically started with is high automation um on the happy paths, but actually uh quite a large number of cases going through to the exception path. Yeah. Now, when you're now approaching it in AI, what we're gonna find is just without a good understanding of what those exceptions are, you're not gonna be able to code for them. Um, you'll not be able to prompt and say, you know, uh, if it's this case, throw it that way. So uh it's more of a um an opt-out rather than an opt-in model of automation. I think that's an area where a lot of companies are gonna struggle to that sort of lack of depth of understanding of those business rules is is the challenge. And don't get me wrong, organizations obviously understand their business roles, but that knowledge tends to be within the heads of the operators still, and very rarely in the documentation that is available to these forward deployed uh operators and engineers.

SPEAKER_00 18:56

Yeah. Yeah, I mean, it's it's that, I mean, this is this is worth exploring. She wrote a book about this or something, but it's like the the the the book end dilemma. And the bookend dilemma is on one side, automation is really, really good at the deterministic, the the tasks that you do all the time that are mundane, routine, definable. And yet the fact is in that environment, almost never do you have those processes documented and and fully like written down, right? The number, I mean, that I bet I bet everything on the fact that almost every one of these port codes for this for this case don't have good documentation, don't know how they do what they do or why they do it. The number of times people have said to us, right, like I can walk you through it, uh, but we don't have it written down. So I'll have to just talk you through how we do whatever billing or or you know prior auths or or rev cycle, whatever it is. I'll have to walk you through it. So that's that one end of the bookend, which is right, we kind of we sort of know what we do, but we can't, we it wasn't documented or it was a while ago, but it's not accurate anymore. The other end, which which wasn't really open to us in the past, but is now that these AI tools are just that much more spectacular, that much more powerful, is the art of the possible, right? What can you do with all of this data? What can you do to create all of that that added impact and benefit that we've all been excited to do? And usually it's associated with more of the non-deterministic, right? Analytics and pattern recognition and more proactive stuff. And yet, when you get to that end of the bookends, to your point, I mean, most of these organizations haven't had the time to even think creatively about what they would do if they had that superpower. And certainly some of these consultants who were rocking up won't know. And so it's this weird bookend reality where you got to solve this one over here, and yet there's a lot of heavy lifting left to do to even get started there. This magical aspiration for what AI can do for you all is predicated on a solid foundation over here that you don't have. So yeah.

SPEAKER_01 20:56

Um so and I think that's gonna be that's gonna be the really interesting part as well for the likes of OpenAI and um anthropic. Are they gonna are they gonna find partners? Are they gonna be acquiring into the deterministic uh automation space um buying capability in, or are they only going to try and solve the um you know the the stuff that AI is brilliant at? Because and it's you know, there's always, and and Ian and I could could talk for hours about this, we've seen this throughout our careers, there's always a like, oh that that generation is dead to me now because this new shiny generation has come along. Um but the reality is is that if this was solving 80% of the problem really well, then what you're looking for is how do you um incrementally add the new generation in order to deliver the end-to-end process, not how do you scrap it and start again. Absolutely. Um and as much as you know, we've you know, the world's been successful, uh, industry has been successful in in pursuing automation, still only a fraction of that deterministic automation scope has been addressed. So I imagine these. New consulting companies are going to find a lot of opportunities for, and this is, I mean, God, this is a tale as old as time, right? The amount of times that you'd go in trying to um Disney reference. I've got the Disney song in my head now. Thank you so much. It's just going around in circles. This is the uh the impact of kids and families. Um, but the you know, the amount of times that a new tool has come out and people are going in and seeing how can they solve you know these these age all problems with this new tool, then you find like seriously, that's how you're delivering it. And actually, it's it's something quite basic. And I think that's what these two consulting firms are gonna come across. They're gonna go in looking for problems that I can solve and uncover a whole trove of opportunities for other tech stacks. How do they address that? That's gonna be really interesting to watch.

SPEAKER_00 22:47

Beauty and the beast. Um, and and I mean, I had this just exactly. That's right. Beauty and the beast. Well, you get to determine who is who. Well, welcome to the um beauty, beast, beauty. Um the I mean, I had a discussion yesterday with somebody who's in again, tail as old as time to to use your Disney reference. Thank you. Their team had the the executives in their team had just recently said we're no longer doing gen AI, we're now doing a Gentech. And it really is just it's it just keeps going where it was. And we're we're not doing RPA anymore. We're doing, we saw this, we're doing AI now. We saw massive initiatives shut down in the RPA realm because they were moving on to AI, which you know, in that case, that just failed famously, and then they went back to RPA because they realized it was actually doing valuable, measurable things for them. Uh, and then we moved on to to to chat GPT got everyone excited, so we moved on to the LLMs. And so chat GPT got everyone excited, so we moved on to LLMs, and so everyone decided that they were going to solve things with that and just generative. And now everyone's I mean, it just the the number of times, and you were this was a funny memory that we I won't I won't name names, but you were you and I were in a in a presentation years ago where someone was calling what we were doing at the time quaint and antiquated, and it really it got your goat, and uh and and the point is that each one of these tools uh is compounding in value. They they you should you shouldn't throw out one tool in your toolkit because a new one just showed up. Keep it now, you've got more tools in your toolkit.

SPEAKER_01 24:26

And so not that I've yeah, not that I've been scarred of it, but it was old fashioned and quaint. So um was the uh was the exact line. But I just thought that was I mean, to me, that kind of epitomizes the issue, right? I mean, there's every vendor has got the new bright, shiny thing, and everything else is antiquated and uh old fashioned and quaint. So um, but the reality is just not like that. I mean, imagine if you were in any organization and every every two years you threw out all of the the implementations you did the last five, you'd be um, you know, you'd never get anywhere. You'd you'd be stagnant. So I do do see an opportunity, and I really hope it's taken where this is used as a tool in a much wider market.

SPEAKER_00 25:04

Well, it just creates the resistance to change that we uh we always point at as the reason change happens so slowly, is you just you you invite it when you show up and you tell everyone that their their thing is old and they should do this new thing, and naturally they're gonna get their backup. This is okay. I mean, isn't this why the Roman Empire spread so capably? It's because they would adopt other people's traditions. They didn't show up and say you were wrong, they said, hey, you yes, and and you didn't, and so you were happy to adopt whatever they brought. So I just went Roman Empire historian on you. How cool is that and improv as well. I mean, it's just the uh the improv uh style of uh transforming your business. Oh, I mean, this this is such a renaissance show. Uh, but yeah, so that's that's it. So to wrap that topic up, fascinating, a just absolute crap ton of money has been thrown at creating um brand new AI consultancies out of thin air that will absolutely require roll-ups of existing firms. It will inherit all of the traditional um approaches, good, bad, or or otherwise, that they would bring, because you can't brainwash people and start them fresh. It's all predicated on um the movement of one uh bill of goods, a a particular type of software suite, capable as they are, um, into an environment that's generally challenging to drive change into. So it will be very interesting um to keep track of what happens. Let's move on.

The Bookend Dilemma In Automation

SPEAKER_00 26:31

So we've got we got some other topics, uh, David, and we can we can uh we can pick and choose. Um so one of them is in coding. AI is being used in code generation, um, partly just because the people who are coding AI are coders who code, and so coding is the thing that they know. And so they're they're using that often as the example of where um you can apply this, right? I mean, so separated from like back office transactional processing, the grungy world that we've lived in, but like writing code faster seems like a magical and super powerful use of this stuff. Um, but it would appear that it's getting worse at doing that thing.

SPEAKER_01 27:14

Yeah, I mean, this is fascinating. I mean, I I would I'd say that and there's a challenge here as to whether or not is there a business case in AI written code. And I stand by the the boring and sensible um uh perspective that I've had for a while on this, which is I think it's a great accelerator. I think you can get a lot more out of a senior developer that knows how to use AI in order to generate a lot of that base code or to you know run specific agents like uh, you know, for information security reviews and audit and the like uh for testing out of components. I think that's where the sweet spot lies. I think um where you get into a lot of trouble with AI coding is where you've got people who are not developers trying to outsource the whole task. So that's where databases are deleted, and um, you know, you end up with a load of code spaghetti, which um is harder to fix than it is to rewrite. So, you know, I I I think it's a little bit of the um, you know, the old hype cycle and you know, over and peak of overinflated expectations. I think there was a point where they thought nobody needs developers anymore, we'll just have these $20 subscriptions and uh we'll get all the code we could possibly need. I don't see that happening, at least not in the foreseeable. But can you get stuff out of these LLMs in generating code that you need uh in order to assist your developers? Absolutely. And we've seen that work to great effect. I mean, we're getting 2x, 3x, the amount of productivity out of a developer who's um embracing these technologies. And, you know, I mean, there's there are there are downside, there are bit bugs to fix, and some things do slip the net. So you've got to be a lot tighter when it comes to testing, just because you know it's they're hard at a spot. This isn't a syntax issue, this is just a different bit of code that's gone in to solve a problem in a fundamentally different way than you want to. Um, but you know, the the benefits are real, so it's all about how do you embrace the technology as an enabler and accelerator rather than as a replacement.

SPEAKER_00 29:15

I I mean I I was in a saw a panel speaker last night say that sort of trope about how a company's not going to hire developers anymore, they just don't need them. And it just it those those sorts of hyperbole are headline grabbing and they just they they they sound either exciting or frightening, but in either case they they get our attention. Uh but it just it's a little it's just frustrating because um because I mean that that's not a viable model in which you have literally zero human oversight or input. And you know, I'm I'm uh I'm guilty of this and and you and you suffer from it, which is you just sort of you you cognitively offload and stuff gets pumped out of of one of these really magnificent tools and you treat it as as great. You just assume that it is. I've handed you full reports where I say, look at this, and then you read like the first two sentences and ask me one question, and it all comes crumbling down, which is annoying. Um, but I I accept it that I cognitively offload sometimes to um to a claude or a chat GPT or another. Um, and it does uh on the face of it look great. And so on the face of it, you could write lots of code um that underneath might not be as structured as you need. Maybe and you know there is the argument that AI is as bad today as it ever will be and it'll just get better. And so maybe in the future, um, we really don't need humans at all anymore for any of it for the creation or the use of this code. That's you know, what's the point then? But uh um but uh we take a dark turn. Yes, exactly. That's right. But it doesn't seem as you said, it doesn't seem tenable. And right now, um organizations are realizing just because you can pump out ten times as much code, you still need humans to to take a look at it and then ultimately use it. And so so uh is is the role of developer dead? No, it will evolve a bit, but as you said, peak of of of sort of expectations, we've hopefully we're coming off of that and we're gonna start applying some reason um to it.

SPEAKER_01 31:26

I think it is a maturity thing, I think it is an expectation piece. I think there's huge benefits to be had there. I mean it's um I I've long referred to you know, like ChatGPT or uh or Claude or any of these tools really, but predominantly it was Chat GPT in the early days when uh the level of hallucinations was higher than it's like talking to someone in the pub. They are they have got an opinion on everything and uh they are confident it's the right one, but um but it usually has some gaps in the logic. So uh that's where I think we are at the moment still with a lot of this stuff, especially when you're you're like producing sort of you know, getting it to do the cognitive offloading, it'll do that 20% so well, so quickly, with so little prompting and and effort that you uh you're like, wow, I got 80. So the 20% of the effort is in the 80% of the output. So you get something 80% right and it looks great. Um, but then getting it that last 20 is almost taking you as long as if you've got the right foundations in place from the outset. Amazing. Um, and I think that's that's the same thing across you know, any use of these technologies at the moment.

SPEAKER_00 32:32

Right. So you heard it here here first, folks. The the G in GPT stands for Guinness. And and that is why geezer, maybe or geezer, geezer on Guinness, and that's why it hallucinates and seems so so confident. Okay, moving on to the next one. This is a this is a concept that's sort of it's confused me a bit, and I I I understand we need to rebrand things, um, but there's there's this evolution from um software as a service, and then we decided to use all the same words and then flip them around a bit um to service as software. Our friends at at HFS Research um have embraced this and and put

AI Coding Boosts Seniors Not Replacements

SPEAKER_00 33:10

the the concept out. Sequoia and other venture capital firms are doing this as well. And effectively the idea is that you know you sell services like software to enterprises so as to um to change the the way that you deliver and change the way that they buy. David, what's your opinion on it? Because um I think it's it's really interesting.

SPEAKER_01 33:32

I mean, it's a little bit of a mind uh bender first appearances, but the the one that I related to um uh is to give a shout out to Rippling, who um who we use for some of our HR and uh payroll needs. Um now brought to you by Rippling. Yeah, if you'd like to sponsor, we'll be happy to do that. That's right.

SPEAKER_00 33:49

That's right to take your tools. So uh any of the exactly Guinness wants to sponsor this. Let's just actually we should reflect on this more. We need to this is our new goal.

SPEAKER_01 34:00

We need to rethink this whole episode. Yes, exactly. Um Maserati is the uh the example I know. Um okay, so back to uh back to the example of software as a service or service as a software. Service as a software, right? No, software as a service. You see, this is just this is the problem, um, at least for my adult brain. So the the problem that we've got with um or the opportunity, let me start again. So the opportunity I think exists in the market with um service as a software, especially when you see it in turn up in um applications like payroll processing. This is an area where I think most people would assume and expect that it's software heavily driven. But I think the reality is there's a lot of ghosts in the machine. There's a lot of humans in that loop that are making this work um the way that it does. And um, whilst it appears fully automated, it's a mix. That's the perfect example in my mind, as of services or software. Always have to double check. Um, and I think it's it's got a lot of uh opportunity, and it'll it's it's basically a commercial model in my mind rather than the delivery model or a fundamental shift in how services are packaged. It just comes down to pricing. Um, we've been doing it for a while now, so we price our services at Magentic uh based on a subscription model and an amount, and the mix of AI and people to deliver it changes over time. And I'd say it's it's been received with mixed uh mixed views. Some people get it and are really happy with that SaaS sort of way of viewing the world and will go in for that that monthly subscription. Others are um are much keener on having you know an understanding of the cost base and how it makes up and to um procure services in a very different way than they procure tech. And then you get into the whole IP space as well, where if people want to own IP, then you need to look at different models for that.

SPEAKER_02 35:54

Yeah.

SPEAKER_01 35:55

Um, so I think it's a it's a there's a lot to one pick here, but I do think it is a useful way of thinking about some services where actually you don't mind how much is delivered by automation and how much to deliver manually, which I would argue should be a case for a lot of a lot of customers out there.

SPEAKER_02 36:13

Yeah.

SPEAKER_01 36:13

You know, I mean that's more of how how efficient um a company is rather than um how uh cost effective and cost competitive they are.

SPEAKER_00 36:23

That's an interesting point. And and ultimately, not to I mean to be respectful of all buyers, but ultimately, as you said, and the point is you should you you you buy uh an outcome, you buy an impact, you buy a value. Um, and all too often you're you're you know, people want to see uh a shiny thing, they want to see a demo. Like show me your tool, and we're just we're conditioned to that. And that's so if if services software is a

Service As Software And Outcome Pricing

SPEAKER_00 36:48

way to try to move past that and say, look, why do you care? It could be hamsters and wheels, it could be people plus AI, better together, which is is our philosophy. Um and that's all that should matter. Um, and so so get your your head out of the out of the out of the gears and out of trying to find witches, because then you end up in these these weird tech bro discussions about like uh I don't know, torque and horsepower and things that honestly don't matter. Um, which is one of the things that drives me a little crazy about just the the the GPT world too, where every 15 minutes there's a new model that that clocks at different speeds and does different things and and is just is is wholly irrelevant if it doesn't actually create an impact and value, which as we said earlier, it's it's a lot of it just isn't isn't doing that yet. Just because it's faster, it doesn't solve your problem. That's not what you're going for.

SPEAKER_01 37:43

So um I think the other thing I'd sorry, I just one last thing to add on it, and I think it really applies into the BPO space. And actually, we've had conversations this last week with working with um you know a client and a partner in a BPO space about how you know AI could be deployed. And it's the same thing that we saw for years when we sold our company to um a BPO front office um service firm and we took the the the you know the leadership positions in the C-suite. The challenge was is how do you embed innovation and change and technology if you're stuck on an input pricing model? So um a lot of the time it would be a cost per hour or a cost per FTE. When you've got your, you know, and the industry went that way um pretty early on and stayed in that that sort of level of maturity just because there's a little bit of an aspect of concern over trust, but it's a bit of also being able to compare apples with apples. So if one vendor is charging me X per hour and another one is charging me Y, then I can do that immediate comparison. I can see what their quality measures are and it allows me to commoditize the market and control the uh the cost base. But the unintended consequence of all of this is that it stifles innovation. So what you then have is you have uh companies that not only are not incentivized to innovate, but are actually punished for innovation. Um, so that if you are on an FTE-based pricing model and you and that's the majority of the revenue receive, you know, are you going to disrupt that FTE pricing? Are you gonna bring down your FTE um count and therefore your revenue by 40% for um you know a few tens of thousands of consulting or SI effort? Um, you should, the good ones do, but inevitably people with targets um mean that they they only sort of the the squeaky wheel gets the grease. You only apply that to the uh the clients that are really demanding the innovation. And a lot of the industry is left in an inefficient model. So hopefully this shift to uh services or software will help sort of you know transcend those issues that we've seen in BPO for decades now, where people have been pushing the outcome-based pricing, the gain share models, but just not really getting that traction.

SPEAKER_00 39:55

Yeah, that very well said. It all comes down to the first principles and the economics of the commercial models and incentives. If we don't get those right, it's it's irrelevant what capabilities uh come next, right? The technology we've we've seen evolve and get better for decades and was never the reason, ever the reason that these transformation efforts would um would stumble, would fail, um, would restart again and stumble and fail yet again. And that's just gonna keep happening. And so that's that's why the the first discussion around these AI consultancies is fascinating, because if it doesn't address the the root cause of um transformation um uh failure and and struggle, then then it's just second verse, third, fourth, fifth verse, same as the first, and

BPO Incentives That Punish Innovation

SPEAKER_00 40:45

it'll be fascinating to watch. Um, which brings us to our last topic, which you know is is is not new to to us, certainly, um hopefully not new to anybody, but there was uh recently a uh uh Boston Consulting Group uh report that that said that um that there's just a misalignment between CEOs and boards of of companies and the the rank and file. Uh, we've seen this for decades, David, but that and and I referred to it a moment ago with the person who said that their CEO said we're not doing generative AI anymore, we're now doing um agentics, which is they're just so hype influenced, they're so driven by whatever the newest trend is, whether they pick that up from their favorite analyst firm or their favorite international leadership conference, or wherever they get this thing, where the the whichever, whichever uh you know, the club they're hanging out at. Um there's some value to big picture thinking and challenging their organizations to adopt the new and to think differently and to address, you know, adopt a new mindset. Um, but so much of what they do is just um bring in the most recent hype cycle with little or no understanding themselves of what is involved, and certainly then an unfunded mandate that says go do it without knowledge of what it is, and in many cases without the budget or the resources or the the plan or the expertise to actually get her done. So, I mean, is this surprising? But there you have it. So that's that's what's happening.

SPEAKER_01 42:23

And I I think it's interesting. So, I mean, the the calling out, the naming and shaming of the CEO and boards, I think's interesting because I I wonder if it's one step beyond that. I mean, CEOs and boards are there in order to maximize uh shareholder value. And can you imagine being a firm that's out there not talking about how you're gonna be utilizing agentics or not shifting towards a um service as a software model? So, you know, I think, and you know, we've seen this in our past as it's not a unique perspective, but like these publicly listed companies that are driven by a uh quarterly strategy, because you've got to make your your your payback and your awards happen in that quarter, you know, they're they are they are at the driving seat of this, and the boards and the CEOs are really just just trying to you know vocalize what the market wants to see so that they can maximize the value. Um, I mean, it's why I I like privately owned companies um and think that they can be better disruptors. It's uh because you you can do what's right rather than what's perceived to be sensible.

SPEAKER_00 43:26

Um, that's a great point of view. I mean, just in and tend to be the the small and medium businesses, this is their time to shine, uh to your point. Although the large enterprises that do have the larger budgets will spend a lot of money, as they always have, as we benefited from, um, on on on the on the hype based projects. But if you can stay focused and committed to a transformation that extends beyond a quarter, then then you could win here. And and it it requires back to the readiness thing. It requires some heavy lifting. It requires understanding what it is that you could do to fully disrupt and just

Boards Chase Agentic Hype Without Readiness

SPEAKER_00 44:07

crush your competitors to be a completely redesigned organization, to rethink how you do everything from the ground up. And then there's the process redesign and the data readiness and everything else that goes with it. But if you do that, then you're you're in a pole position to really capitalize and leapfrog anybody who's who's frankly who's unpicking the last um you know the last agentic program where they where they deployed um too many um without the right constraints or guardrails or governance, and they're now just picking up the pieces, um, which will be fascinating. Um again, I uh uh we we find this very interesting and valuable and powerful. Um, but there will be pendulum swing back when everyone realizes I I was told I don't need a you know, whatever, a sales department. I was told that I could replace every one of my SDRs with agentic SDRs, and uh and now I've realized not only have I gotten rid of the the industry knowledge that sat in the heads of my SDRs who understood how to talk to my customers, understood our strategy, and we're also struggling to find where all of the the emails sit, right? It's it automatically wrote things for us. We don't know what we're following up on anymore, and it's kind of chaos. Maybe it won't be that bad, but um, I think there'll be uh a reckoning and then an unpicking and then a uh uh a building back up in the next 18 months.

SPEAKER_01 45:37

Um I think so. I think I honestly I I these these topics do all head in the same direction, right? I mean the uh um the pub. Exactly, exactly. You're gonna go chat chat to that GPT in the pub. Guinness. Um but brought to you by but you know, the the the focus on getting into enterprises and doing larger projects, the that we see with anthropic and uh the PE focus. Yeah, um the shift away from AI will do it all, and actually you need you know some professional um insight into making sure that you're you're delivering the uh the the right combination um and you're using AI in order to supercharge your your people rather than replace them, that we've seen with the uh the futurism article. Yeah. So Coya's view about moving away from talking about software and selling features, but selling services that are foundated that are foundational to that is is a purchasing of outcomes. Um and you know, the the BCG saying that CEOs and boards need to move away from hype and focus on what's really delivering. It's it's all leaning to the same thing, right? I mean, look at what outcomes you want to deliver, you want to achieve, you want to get for your customers, for your stakeholders, and uh focus in on what the best way of achieving that is. Everyone should be considering agentics, ai, all of these these capabilities, but it doesn't mean that the latest generation is the only thing that's going to deliver that value or the best thing to deliver that value. You know, look at a combination of of different technologies and delivery models to really give you

Final Takeaways And Subscribe Request

SPEAKER_01 47:20

what you need.

SPEAKER_00 47:20

Love it. That is that is a great place to end. Thank you for for joining me. And uh, we will do this again, maybe once a week, as we discussed. And uh, and yeah, thank you guys for tuning in. If you're listening, subscribe, like, upvote, whatever you do um to one of these things. Uh, and please join uh David and me again on uh on the Magentic show, soon to be the world's very best podcast on uh the future of work, life, and culture and how AI and agentics and all these other tools are going to influence it and hopefully impact it for the better. So uh David, thanks a million, and uh talk to you again soon. Be a pleasure. Thanks, Ian. Take care.

unknown 48:00

Bye.