The magentIQ Show Ep. 4 | Token Maxxing and Org Hacking: Designing the Agentic Workforce That Actually Works

Ian Barkin and David Brain return for Episode 4 of The magentIQ Show, picking up where the headlines leave off. AI is everywhere at work right now, but the loudest signal is not better results. It is token maxing and layoffs justified by an AI story that often runs ahead of reality. Ian and David dig into what happens when leaders measure AI adoption by volume of usage instead of business value, and why that choice quietly rewires incentives across the entire organisation.

The conversation moves through the latest wave of org hacking making the rounds: Coinbase's one-person product teams, Jack Dorsey's ambition to cut his company down to two layers, and the broader fantasy that agentic AI can absorb the messy human parts of building products and running teams. Ian and David name the constraints those models ignore. The pace of a single operator. The bias of a single perspective. The management work that does not disappear just because you can generate a status report. They also surface a risk more leaders need to say out loud, which is the use of AI to manage performance when the system will confidently fill in blanks, hallucinate context, and send signals that ripple through motivation and retention.

From there the lens widens to enterprise AI implementation and the rising tide of AI consultancies, forward deployed engineers, and services arms trying to turn licences into outcomes. Ian and David tackle commoditisation, lock-in, the puzzle problem of orchestrating a thousand-piece agentic stack, and why buying into the inner circle of a model provider does not solve the hard part of deployment.

In this episode, Ian and David unpack:

• Why AI-driven layoffs are often a budget reshuffle dressed as transformation, and what is really being freed up
• One-person product teams as org hacking, and the speed and bias limits the pitch leaves out
• Flattening the org chart versus the real work of coaching, context, and care
• The risks of letting AI manage people when the model is built to sound confident, not to be correct
• Token maxing as a vanity metric that floods teams with low-trust output
• The graduate backlash to a decade of "AI will take your job" messaging, and what it signals about the next workforce
• Why enterprise AI implementation still needs governance, security, and services to land
• The Sunday tinkering and Monday reality framework for building real AI fluency
• A practical path for small and medium businesses: learn personally, prototype safely, then harden for production

The takeaway is constructive. The agentic workforce is real, the opportunity is genuine, and the operators who win will be the ones who design for value rather than volume, who keep the human craft of management intact, and who treat citizen creativity as an input to disciplined deployment rather than a substitute for it.

Listen now and tell us where you stand on the question shaping the agentic era: are you optimising for tokens, or for outcomes?

Find us in your favourite 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

Grumpy Intro To The AI Moment

SPEAKER_02 0:00

Yeah. I mean I I guess so in in the spirit of trying to be productive and and uh and and uh and that's what we're aiming for? Oh right. Yeah, no. Sorry. I that's why I'm stumbling because like I just wanted to whinge more. A college graduation speaker told the senior class that AI is the next industrial revolution. And they booed her. As David put it, she was talking to the horses. This week's episode is what happens when companies start measuring AI by volume instead of value. Token maxing, one-person product teams, Jack Dorsey wanting two layers between him and the work. The Sunday tinkering version of AI meeting the Monday reality of actually running a business. It's me and David, unvarnished, grumpy in the best way, with about 40 years of trench-earned insights and skepticism between us. Join us as we try and figure out what's going on in the new agentic world of work. And now, on with the magentic show. Welcome back to another discussion on the magentic show where David and I try to figure out just what the heck is going on in the world. Hi, David, how are you doing? Hey, I'm very well. It's really nice there. I like looking at your background. I'm gonna just admire that tree. Look at that.

SPEAKER_01 1:30

We have a few nice, a few nice months to make the most out of. Is that tree always that color or is this like a fall thing? No, no, this is all oh yeah, nice um bamboo. Not bamboo. What am I talking about? It's uh it's an Acer, a Japanese maple.

SPEAKER_02 1:44

Okay, uh Japanese maple. There we go. All right, well, I like it. I want one. I don't know that it would grow in Florida, unfortunately. Probably not. All we have is palm trees.

AI Layoffs And Budget Reallocation

SPEAKER_02 1:52

Um okay, so um we've got a sort of meandering show uh notes, and uh we'll we'll both sort of we'll play off one another as far as what is as we want to talk about. Um I guess top of the there's a few major stories. One is we continue to hear about uh layoffs, uh layoffs in in in IT companies, layoff and financial services companies, almost all justified as sort of AI driven. And in many cases, it's because we don't need all these people anymore. AI is going to do the work for us. Uh, but I think the prevailing wisdom is two things. One, perhaps there was an overhire during COVID, um, which if you if you Google it or Chat GPT it or anything, it does say that there's some truth to it. So it's it's not just a line. Um, but also it feels like it's probably more freeing up budget to then go figure out AI because there's no way in hell that these companies have figured their stuff out to be getting rid of hundreds or thousands of people because they've AI fied, agentified the roles. So I don't know. What do you think? Are people have they AI'd it already? Is that why people have laid off so many?

SPEAKER_01 3:02

Nothing else we're seeing in the market would suggest that every enterprise has cracked this.

SPEAKER_03 3:07

Yeah.

SPEAKER_01 3:07

So um, but what I can imagine is everybody would like to declare that they are an AI first company. It's great for stock prices. So, you know, what I can imagine is that there's a lot of firms that are relying on other forms of automation in order to give them the real benefits, yeah. Um, and they are becoming more efficient, but it's just not with you know the latest version of Claude or uh uh you know, or Spud at the uh base of it. So um yeah, I wouldn't be surprised if it is some form of automation now. Yeah. I was gonna say I was speaking to a friend over here who helps with um uh assessing uh candidates actually for uh I think it's for MIT. Um so uh he'll remain nameless because he already gets too many people contacting him. Um and uh yeah, so he's been doing that, and he also helps with um assessing uh people for uh graduate places in the firm that he works. And what he's noticing is a massive reduction in the amount of the people getting graduate roles. So I believe that the data that came out this last week showed that that's starting to increase again in the US, but I think globally we've really seen an impact on those graduates.

SPEAKER_02 4:15

Yeah, yeah, we'll get back to graduates in a second, because that's that's a fascinating and and I mean it. There's I don't know. I I I there was a panel I listened to last week where where someone said that they are really glad that they aren't uh a 20-year-old today, um, just because they're at peace with where they are and and because man, they wouldn't want to be trying to figure out a career start right now. And another person who said they wish they were a 20-year-old right now just because you know, because of the the options and opportunities um afforded to you because all this stuff could help you start major companies. So who knows? Literally, yeah. I I I don't know. I feel like I'm in the first camp. Um is it's way too confusing. I'm glad I'm I'm old and crusty. Um the the one thing that occurs is and having lived out in California um for a while no longer, but there's always like this culture of like of hacking, like biohacking. Like how can you how can you be healthier? How can you like do the sort of those those um

One-Person Product Teams Sound Risky

SPEAKER_02 5:20

like starvation stages and then different pills and uh so there's always biohacking? It seems like there's sort of like org hacking going on right now, right? Because um one of the layoff uh the stories was from Coinbase, and Coinbase started suggested that what they're gonna try to do is effectively um uh use AI to create AI augmented teams where they go down to as as small as a one-person product team, where um just because AI is all the other people for you, all you need is one person to to develop a product. That feels like org hacking. And uh I don't I don't know, my again, my old crusty self feels like every I mean the entire logic of like you want different opinions, you want like lots of people because it makes better products. And if you reduce it to one person and AI that's just generally sycophantic and tells you that your idea is great, then um it's gonna be pretty interesting to see what what these product teams come up with.

SPEAKER_01 6:22

Well, I and I you know what I I can buy in that you could have a one person um you know resource that does takes a feature from start to finish with um with the value of using different AIs and do QA and to do requirements, do documentation, da da da to assist in the development. Um but you still only can move at the pace at which that one person is able to progress things, and you're always gonna end up with the biases of that one person. So um they may not be able to spot some of the things that they should that that's not that's not going right in that product. So I mean, even with our use of of AI agents across the development and testing and all the rest of it, we you know we still wouldn't propose that any customer takes like a one-person product team. I just think it's I think what that and and and I think when you start hearing that stuff, you really wonder how much innovation that that that firm is going to be going through. Maybe in Coinbase, they don't need to progress their their application at all that much. But if you're in a sprints to build something out, you need to move quicker, right?

SPEAKER_02 7:25

I mean, and we we're we happen to be friends and investors in a company started by a former um the chief product doctor of Coinbase. So maybe well, I'll ask him next time we chat. Um he uh yeah, he's uh he and I are gonna see each other in two weeks at the mentioned MIT, but I've got my my business school reunion, 20-year reunion coming up. Um, and it's interesting, most of the the speakers, or at least the speakers I registered to go see, are all talking about uh effectively the the the human plus AI element in the workforce. So it seems like like this is the this is top billing. We're all trying to figure out just sort of what organizations need to look like. I think business schools, if if they do in fact teach people how to run businesses, which I maybe they do, um, this is critical. I mean, the the the the hybrid of your resources available to you, not just cash and and physical materials, but humans and then this virtual uh labor element. You gotta you gotta have some frameworks and models to figure out how to how to

Flattening Orgs And Losing Care

SPEAKER_02 8:28

manage it. Um this is another one I think you'd like this, because I this is me doom scrolling on TikTok, but uh there was a um Scott Galloway, Professor Scott Galloway from New York, um, who's got very strong opinions and this is a pretty smart guy. Um he was commenting on a different podcast discussion um with uh the the the former uh TikTok uh founder who uh Jack Dorsey. And Jack was again sort of org hacking. Jack was talking about sort of levels of of like layers within a company, and how he thinks his group company right now has like five layers between him and the bottom, and he wants to get it to three or four, and then ultimately he just wants everyone reporting to him. He wants one basically two layers, like just the CEO and then everybody else, because you can somehow magically identify um as a grizzled operations manager and and and C-suite manager and and everything else. Does that sound awesome to you?

SPEAKER_01 9:36

Uh it sounds like hell, to be honest. Um, I don't know. I I that's what I think by Quip. I mean, I I you know I love working with the people and the teams, but um you know, just HR performance, career growth, making sure, and you know, I I think what I've I believe if you want the best performing team, you've got to take some responsibility for developing them. So um with that, you you've got to if you're spreading yourself over, you know, any more than 15-20 people, you're not doing that justice. So um, yeah, if you've got I don't know what size team he wants to end up with, but if he's considering a world where 200 people are reporting into him, I mean imagine a job where you get none of your manager's time. Um yeah, doesn't sound very peer.

SPEAKER_02 10:24

Literally, even if if everything else is effectively multi-agentic and it's doing all the things, as you said, to have zero human interaction because and andor to have your have your your emotional and personal and career development needs met by yet another agent where you're literally just talking to a chat bot and say, like, thank you, David, for expressing your your your need for I don't know, for growth, development, a holiday, a break, some some some just uh some therapy. Like, um what a great question. I mean, again, it's it sounds like hell. I think it does sound like hell.

SPEAKER_01 11:03

I mean, I you know, I'm today I've moved my stuff around because of you know people's personal lives. Um, you know, one of our team and uh you know won't know names, but um is going through some personal stuff. Um, and you know, that means that the the rest of the team's got to rally. And you know, my job as a manager is to make sure that those things are covered. Yeah, you can't do that with 200 people or whatever you have reporting into you. So I think it's fine with the idea that you've got a 10-person business or an organization, but yeah, it doesn't scale.

SPEAKER_02 11:30

And why would you want to be the same? But it is interesting. And sorry, why would you want to be the bottom layer where you're just like a cog that's completely ignored and doesn't have have the level of of intimacy or care or culture? Anyway.

SPEAKER_01 11:43

Well, I tell you what, one of the interesting things, and I gotta be careful as to how we go into this one, is um seeing the use of AI in managing as well. Um that is that is really interesting when you start to come across reports which um hallucinate on the data and come to very different conclusions around the performance of individuals.

SPEAKER_03 12:12

Yeah.

SPEAKER_01 12:12

Um, and uh just you know, I mean, I s I use hallucinations, and some of you might be thinking that, yeah, okay, well, there's the the traditional sort of hallucinations that you get where it just says black is white, whatever. But I think also part of it is just, you know, I've I've I've often thought that um the LLMs are, and I think I've mentioned this in the last podcast, but that they're the uh the the confident person in the pub that's got an answer to everything. So um, you know, in that sort of situation, you can imagine a uh where the AI is asked to comment on someone's performance and pull it together, even though they don't have the full data, they will reach a conclusion, right? They will come up with the um uh the details as to uh, you know, even though it's got part of the information, it will, it will fill in the blanks. Yeah. And that's that's the risk when you start dealing with with you know people's careers and their lives.

SPEAKER_02 13:02

I I completely agree. I mean, again, do you get frustrated with with me when I I hand you AI slap and I it's just a bunch of stuff I think is is brilliant, and uh and then you scrutinize it. And and so A, the volume of output that comes is tremendous, and you can create more more of everything. You can you can get off a call, um, which we're we're doing because we're actually we're pulling together some some training programs um for various different audiences, some private equity firms we're working with, and some other um small medium businesses. But you get off a brainstorm call, you you take that transcript, turn it into an outline, and it's just infinite. It's just so long. And so you just look at it and you're like, this is this is great. This is really good stuff. Um, you don't have enough time in the year to review it, let alone to validate it and andor act on it. So this is gonna happen. I mean, again, let's get back to the Dorsey, like one layer deep organization. If if every one of those people has a report that's coming out of coming out of Slack and GitHub and ClickUp and whatever project management tool you use, it'll inundate you with so much shit that you will not be able to compute it um or process it. And then you'll be left to just defer to it, and as you said, and and a bunch of it'll be wrong. You'll be making bad decisions that will just antagonize that that team of 4,000 who report directly to you. Yes. Um, yeah, exactly.

SPEAKER_01 14:33

And that's the other thing, Ray. I mean, the the the cost of um mismanagement is not just the individual that is affected, but it it has a ripple effect across the organization. And so it starts impacting the uh you know the uh motivations um and the retention of the whole workforce. Yeah.

SPEAKER_02 14:52

Yeah, it'll be it'll be interesting. And and and I think again, this is the you you uh you see it in some of the this is the whole like Luddite analogy that we had 10 years ago of people breaking looms because they were they they saw their their their jobs being threatened. Um, but there really is gonna be, as there always is, it uh a resistance to change and a pushback on on how some of this is happening, but also just the way it feels, right?

Token Maxing Versus Real Value

SPEAKER_02 15:19

There's just this top-down mandate. Um, a lot of it's around token maxing now. Like, are you using as much of this stuff as humanly possible? That's I mean, that's gonna hopefully that isn't in the headlines for too much more longer, because I don't want to say token maxing very often. Um, but token maxing is literally, I mean, like it's like using as many lines in Excel as you humanly can. Like, are you using all the Excel? Like max it. Um, and and often it's an unfunded mandate that comes with incomplete training and and a lack of clarity. You could you could you could blast through tokens real fast. I mean, you you see that in some of the programs that we're developing, is is there's certainly a way to consume as many as possible, but there's also more efficient ways to design uh agentic workforces.

SPEAKER_01 16:05

Yeah, I mean I I think the the token maxing is a direct correlation to like the the same thing that we're seeing uh being tackled from various different directions, and that is we got an amazing capability, but we don't know what to do with it. Yeah, maybe if we just force people to use it, they'll come up with some good ideas. Yep. Um, and you know, I think there's there's a sense to to that up to a point. But um I think the challenge comes when you start managing people on the basis of how well they are using the tokens, and the measure for that is volume. You know, so I mean, if it's a case of, you know, I'd rather have someone in my team that's using it 10%, but for that 10% they're using, they're leaving a huge amount of value than someone that can, you know, never write me an email without it going through uh, you know, the M-dash generating hell that um that makes everybody sound like them them uh each other.

SPEAKER_02 17:05

So value you say, what's the point? Why are we doing this? Yes, you're suggesting that. Critical it returns to value, yeah. Right, exactly. Um, yeah, and and the other interesting element, and this is just this is going back through our history, but ultimately the top-down mandate again. Um decade after decade, wave after wave, it's been the CEOs who've declared, and you know, not to pick on CEOs. They're lovely. We love CEOs as co-CEOs ourselves. We're we're delightful. Um, but um you don't have an you don't have a lot of time. Um, and so you're you're just trying to to move an organization forward, you're trying to have vision, you've been told that this next wave is the one that you really need to embrace to stay relevant and not to be uh you know beaten by your peers and competitors, and so you push it down. Um, but it brings up uh it brings up middle management, which always seems to be the the whipping boy where everyone wants to get rid of it because they think it's expensive and unnecessary. So back to the Dorsey comments, he wants to get rid of middle management and just go directly to, but unfortunately, CEOs are distant from the the day-to-day, right? They they they they lack the context because they can't spend the time to have the context of what's going on in the trenches. Middle management sort of does. You get rid of them, Lord knows what these people are doing, they're just being incentivized to use as many tokens as possible. And as a result, yeah, like how do you get value out of just a team that doesn't have the nurturing and the human support of uh you know, collaborators, like one-person product teams and and or or managers? I mean, so much of the institutional knowledge sits in middle right, whatever you call them, but people who've been around a bit. So um, so this is gonna be an interesting. I I I think it'll end up it'll end up depleting organizations of valuable context, insight, and experience that will be very hard to to rehire for um if if and when they realize that they've just they've stripped too much, they've gone down to the bone and they need to come back, and the agents aren't gonna have all that context in them just magically. So anyway, that's my point of view on that. Um where do

Graduates Boo The AI Talking Point

SPEAKER_02 19:25

we go from here? Okay, so um so one of the the hilarious stories, truly love it, um, is uh just and this is gonna get more and more coverage. This will be on news stories. This will this will hit Twitter and LinkedIn and Instagram and everything else. Um, but there was a graduation speaker at University of Central Florida, um, who who's standing in front of an entire class of graduates and um and does what everyone's doing now, which is talks about how AI is bringing on the next industrial revolution. Which, you know, that's easy fodder, probably written by a uh uh an LLM anyway for her. And what is so delightful is the entire student body, like all of these graduating seniors, all almost in unison, boo her. Like, like, like voraciously boo this woman, which shocks her. If you can't like go find the video, Google, um Google Florida graduation speaker, AI booed, and this will come up. But so she's shocked, doesn't like looks off the stage to like someone to try to figure out like, am I in trouble here? Asks the student body if she can continue. She's like, Can I keep going? And then proceeds to basically the next line of her of her story, which was sort of to go back to go forward, which was in in my day, effectively, we didn't have AI. At which point, the entire student body cheers at the comment of we didn't have AI. And I can't blame them because these poor kids are just being inundated by news of effectively whatever you wanted to be when you grew up, whatever you've been studying to be, isn't going to be. Right? Whatever, I mean, I don't know what the programs are at the school, but like if it's engineering or computer science or legal or consulting, or it doesn't even matter. Like every one of them is being inundated by these just overbearing messages of all those jobs are gone. I don't blame them, that they're just booing the stuff at this point.

SPEAKER_01 21:31

It it does seem like a a bad time to uh to bring that up, right? At a at the party, the celebration of of three, four years worth of effort to uh to get a degree. And uh I think the industrial revolution uh analogy is fascinating because she's not talking to the uh the managers in that um yeah in that example. She's talking to the horses. Yeah, you know, so uh that's uh I can't imagine you'd have got a much better reaction in the last industrial revolution either.

SPEAKER_02 22:00

Really good point. She is literally talking to the horses and and andor the people who would be moved into like factory towns who would just be treated as cogs, right? So so uh I think we need there's a gonna be a reckoning soon about um how to balance the messaging around this, yeah.

SPEAKER_01 22:21

Yeah, I mean because there was like well, I think there's the messaging, and then there's the reality as well, right? I mean, as I said earlier, there are you know, graduate recruitment, we've seen it in the UK. Um, I think I saw something more positive coming out of the US this last week, but um, you know, the graduate recruitment's been way down, and it's hardly a surprise because you know, we've had decades now of offshoring and um now you know decades of automation uh through its various forms. Um I think it's all now being put being wrapped into an AI sort of you know brand because that's somehow more more acceptable um as a firm to be doing it. But uh yeah, I do think the the workplace is affected. Um and it is it is you know it's an opportunity for disruption, but it also um, you know, that disruption comes at the cost of those being disrupted. Yeah.

SPEAKER_02 23:13

I I I agreed with a speaker at a panel I was at last week who who did emphasize that this will be a rebirth. I think we talked about this last time, rebirth of of sort of the liberal arts, as far as don't don't try to go be necessarily a computer programmer, just go go do English literature and uh do something you enjoy, but also you know, that's pretty helpful for prompting, like know how to communicate better, know how to Yeah, yeah, yeah, yeah. Yeah, yeah.

SPEAKER_01 23:40

I I have moments because um I'm better put my son through an expensive uh education. Well, five more years of an expensive education, and I keep sort of reflecting on it that actually for that amount of uh of cost, he could get a really decent, you know, van and uh and either uh all he needs, all the tools possible to be an amazing electrician or a plumber, you know, or one of those types of skills.

SPEAKER_02 24:04

Honestly. Well, and and and those those on site, on the ground. I mean, right, the heating, ventilation, air conditioning, those are that's gonna be that's gonna be where the uh some folks like uh uh Mike Rowe, the guy who has the uh show over here in the States, the Dirty Jobs, where he's always celebrating sort of the blue collar. Yeah, right. Those those you can't identify. Right. There's I mean you can you can you can automate more of the Internet of Things element of it. You can have better sensors on these things so you get better information, but it still then will just tell you proactively that someone's gotta come out to my house before it becomes a a sweat box as my my AC fails in the summer in Florida. Um somebody's gotta do a truck roll. So so yeah, maybe, maybe that's maybe those are the jobs of the

No Single Tool Solves Enterprise AI

SPEAKER_02 24:55

future. Um so we have long, um, just because we've been doing this for a long time, we we have long believed, long said that the way to survive in in this space, um, from a from an actual viable solution and deployment perspective, is to understand that this is there is no one tool that's gonna solve all your problems. Um, and just dumping a tool in a hole, which we've long seen is the aspiration of the vendor, but not the right way to do things. Um it's something that we that's I mean, we've embraced, it's we've started startups, we've funded other startups, and we have a uh company now that that believes in this just to our soul. That messaging seems to be coming out now. Um, so we told you so is my synopsis of that intro to this topic. Um but right, I mean, you we were talking about this just earlier. The the the presence, the existence of these new AI consultancies um speaks volumes that are um implicit and perhaps a little um unsaid. So so pick up on what you were saying to me before, because I I do agree this is it's a fascinating thing to watch.

SPEAKER_01 26:15

Yeah, I mean, I it it it does strike me that I mean, you know, obviously we're in in this this massive growth area. So they use the word desperation as perhaps the wrong term, but there is a uh a real concern as to how the AI companies are going to get into the enterprise use cases, be it SMBs and you know, or um or the large uh large organizations. And um, you know, uh you just start seeing uh the different approaches play through. So first it is the you know, the deployment company, as we spoke about last time, and you know, these these JVs trying to get into the um the PE world and they'll be successful. I shouldn't say try. Uh next, we're seeing that each of the AI firms investing in their own uh consulting services, their own services firm for consulting and implementation, these, you know, the the uh the massive amounts of forward deployed engineers, I'm sure we'll start seeing turning up as part of a um an enterprise purchase of these tools. Yeah. Um I think it's I think it's right that the the world needs help figuring this stuff and implementing it. Um and the other approach at the moment that we also see some people in the market use is just just throw tokens at it, throw licenses, get everybody to use co-work without direction. You know, I mean, so so those are the two parts that seem to really favor the deployment and um of and the selling of licenses.

SPEAKER_03 27:44

Yeah.

SPEAKER_01 27:45

But you know, is it selling value? And is that where we're gonna be seeing the uh the the best case studies in the future, you know, the most disruption? Yeah. And my my theory on that is that I think we'll we'll start to see it more through the service arms of these companies. That will definitely help get into the space. But it's unless they want to become a full service sort of transformation um outfit that's tackling every different aspect of of um digital transformation, business transformation, yeah, they're gonna still struggle. The other thing is, is it does strike me that if you look at a lot of these tools, um, it's very hard to for them not to become commoditized. So, I mean, you've got those foundational models, the three or four sort of massive ones in the market. Um, you know, it's very hard to approach their levels of maturity, but among them, you can pretty much, you know, track each month there's going to be another one that's considered the best, the best at coding, the best at prompt engineering response, um, the best at cowork. And, you know, these tools are being heavily sort of invested in. But I really do wonder if any organization wants to take a skill set which in some ways is is more general and commoditized and easier to swap out, yeah. But then bake it into one stack through, you know, using Claude's consulting and implementation arm, you know, so that they're tied into Claude, even though you know there may be a better solution out in a month's time or three months' time.

SPEAKER_03 29:13

Yeah.

SPEAKER_02 29:13

Yeah. I mean that the play is clearly to to move units. Um and you said the it seems like enterprises um more and more were always uh starting to worry about lock-in and stickiness, right? They're they're they're they're hip to the to to what's being done. Um and and and yet they do, I mean, the big, big enterprises, the ones with with uh enough uh resource, don't want to miss a trend and a fad. And so they all want to be close to the sun and and be part of this party. I I had a a chat with um with the a friend of mine who who's closer to the sun in the Silicon Valley um ecosystem uh a few days ago who said that to to get into the sort of the circle, the cohort of cool um within one of the big LLMs, one of the one of the big brands, they there's a their price to buy in, right? There's there's a certain minimum that you have to invest to to be part of the inner circle. And it's you know, but committing to buying their software. And it's and it's not and it's not small. It's like hundreds of millions of hundreds of millions of dollars committed to buy by the software. The problem is you literally you buy hundreds and millions of dollars of it, you don't know what the heck to do with it, and you don't know how to place it anywhere. So ultimately, what you're gonna do is you've bought in you're you're hanging out with the smartest people who are saying sort of esoteric things about the future of of you know cognitive and neural synaptic and and it's exciting, but you can't you can't get the stuff in your organization, um, which is which is only a bubble waiting to happen, right?

SPEAKER_01 30:57

Because so what do you turn to? You turn to co-work and token maxing.

SPEAKER_02 31:02

Yeah,

Consulting Arms And The Unicorn Myth

SPEAKER_02 31:03

yeah. Token. Well, the and the other interesting I I've long thought about this just because my career has been in consulting for for so long, but it's just why don't you do it for yourself? And usually it's because because there's there's a lot of complexity to a lot of things that you don't you don't have the ability or the resources to be the specialist in. And there is if if we were to call consulting sort of a rainbow spectrum, there is a rainbow spectrum of consulting, right? There are consultants who do like management consulting and strategy, they do opportunity assessments, there's ones that do learning and development, there are ones that do change management, there are ones that do then implementation, there are ones that do continuous improvement, right? There are specialties and there are specialists in each of those for a reason. Um, and yet the narrative right now, especially around these forward-deployed engineers, is all you need, David, all you really need to be successful at this, is someone who's an absolute expert in your domain and an absolute expert at the same time in AI and the design and deployment of a process and the deployment of the AI and the management of it and the redesign of how you do things. So basically, we are looking for a flip in unicorn, whether it be in a team or a person. And you know, time has shown that that's that's not that's not a viable expectation. So we're we're hoping we're hoping all of a sudden somehow, I don't know what sort of training program makes you like a grizzled 30-year-old expert in finance and accounting, you know, period and closing, and also process re-engineering, and also the codification of that into AI.

SPEAKER_01 32:52

It's not just, I mean, as if that wasn't enough. It's not just that. You always take more ridiculous amounts of time to dedicate to these projects. Yeah. And this is where this is where the LLMs get you, right? I mean, you you get started and you can prototype uh vibe code something that's like that looks looks close in next to no time. But if you want to build out that functionality set and get it to a point where it's actually at that MVP stage, you can, you know, you you you end up fighting with the tools um after spending hours on them in order to to um to get close to an outcome. And at that point, you're thinking, you know, I mean, as the person who's the subject matter expert who's building this out, you there's there's got to be a better way, right? There's got to be a better use of my time than trying to fix the uh the issues that's coming out of these tools. Meanwhile, I'm running out of tokens and um you know I'm I'm losing bits of functionality every time I I fix one other part.

SPEAKER_02 33:54

And my to-do list of my day job is just piling up because I need to get it, and and so the the the positive side of the you know, the glass half full side is uh again, it allows you to to to prototype and to brainstorm and to ideate or whatever consulting word we want to use um in ways that you never could before. So that's awesome. Um, but back to back to specialization, um, you can't be expected to be a specialist in all things. And and again, to your point, if if you vibe code something up, that is exactly what your organization needs. You got a small, medium business, you understand your industry incredibly well, and you know what the SaaS providers have not given you in the past, and you're able to craft it even to 99%, that still then leaves the not only the deployment, but the safe deployment and then the ongoing support of this this platform that you've developed, right? Because if you're a demand expert, you're probably not also an expert in back-end systems and um cyber security hardening and compliance and risk and and and and and uh nor should you be. It's unfair to expect you to be all those things. Um, and so there'll there'll be there'll be um service specializations that'll pop up, right? We're having discussions already of like, can you turn my vibe coded thing into something that's safe and and deployable over time? Yeah, yeah, absolutely. Maybe um, but what we'll bring in is we'll bring in all the rigor then to ask you, okay, I see what you're trying to do. Um let's let's go back to first principles and figure out if this is the best way to achieve that.

SPEAKER_01 35:37

Um, or you know, maybe it won't be by picking up the project you started in Claude and carry on vibing. It'll be uh you know, looking at what you want to achieve and ideally building it into a tool which enables you to swap things in and out in the future.

SPEAKER_02 35:51

Yeah,

The 5,000-Piece Software Puzzle

SPEAKER_02 35:52

yeah. I mean that's that's an interesting one. I you know, I try to figure out in the old days, in the old days it was just uh uh I got no good visual for this in my head, but it's it's kind of like puzzles. And you know, when you've got a kid and they're young and you you open up a puzzle of little ducks and and um puppies, and the puzzle has like 12 pieces, right? And you know, it's it's the big pieces and it's easy, and you're looking at it, and it makes the picture. Um, and then when you move forward and there's infinite, um, so let's pivot to the the use of this analogy, infinite software vendors, there's LLMs coming out of the woodwork, there's your existing providers that all have agents now. You can you can vibe code everything else in between. So the SAS pocalypse means you're getting rid of some of that with like discrete things. You end up with a puzzle that's like 5,000 pieces. And um, and and as you said, and some of the pieces can just swap out, and it's it's much harder to orchestrate and control and govern that um than than your than your puppy and duck puzzle, which has fewer. And the puppy and duck story is like ERP, CRM, some some like mainframe systems, like in the older days where I got my career started, was I was implementing ERPs. And that was that was a big thing. You spent way too much money and time on it, but it was just all thing. Like there was you could customize it, but it was hard to go off the rails. I mean, it was it was Oracle, SAP, PeopleSoft. Um, now it's uh a million little things in place of it, which could be really interesting, but I don't know that organizations are sophisticated enough to coordinate a thousand-piece puzzle.

SPEAKER_01 37:36

Well yeah, and and if you were to do it, you wouldn't want each person trying to to build out their own bits. That's right. You wouldn't want each each member of your team to try and navigate that thousand pieces and work out which are the right ones and which ones have infosec risk and you know what IP is owned and retained and where the data stays and all of that. So, you know, then you get into more of a point of needing an enterprise toolkit. Um but you know, you need and the capability in order to make it make it make sense, and that's where the reality of the AI enterprise implementation is meeting the the sort of the pitch and the hype of a SaaS model, which just wants to shift tokens, right? Just wants to shift sign up like subscriptions. Yeah.

SPEAKER_02 38:26

Yeah. I mean, I I guess so. In in the spirit of trying to be productive and and uh and and uh and that's what we're aiming for?

SPEAKER_01 38:38

Oh right.

SPEAKER_02 38:39

Yeah, no, sorry. I that's why I'm stumbling because like I just wanted to whinge more. Um but but but productive and prescriptive as far as you're let's say let's okay, I'll I'll I'll I'll do I'll define

Practical First Steps For SMB Leaders

SPEAKER_02 38:52

you. So you're uh an executive operator owner of a medium-sized business. You're inundated by this news constantly. Um you're you're freaking out a little bit because because you're told that if you don't adopt this, then your your closest competitor is going to eat your lunch, or there's gonna be a a native AI native company that's gonna come up out of the just out out of the blue and and just clock you. Um what do you what do you do first? Do you do you do you get everyone trained in cloud code? Do you bury your head and say what do you do?

SPEAKER_01 39:28

I think so so my my advice to this person um is of don't panic, don't sort of believe everything that that guy that is you you know that you meet socially or at the school or whatever keeps telling you about how they've um already achieved like automation of 20 roles in this tool, because the next week you meet them, it will be another tool that they've just done exactly the same thing, right? So so don't get too caught up in the hype and the uh and um the over excitement and exaggeration of people that are using technologies. Get to know them yourself. Um, you know, just do it for personal things. Like, for example, I mean, you know, a great place to start, and it's one that I'm helping several friends through, is just as they prep as their children prepare for exams, is like how do you use AI in order to support um tutoring, revision, you know, creating apps and games and worksheets, you know, for around your your child's education. Um, or, you know, what can you do in order to help your home finances or, you know, whatever it might be, or help that book that you've always wanted to write? You know, just use it around something. It doesn't need to be around the capabilities that you need to automate in a business, but it's something that you've got a passion about and you've got the the time to put into uh to really experiment and get to know the capabilities of the various platforms.

SPEAKER_00 40:52

That's cool.

SPEAKER_01 40:53

Then what I think you can do is applying it into an organization context in terms of how would you use that on a project, prototype, you know, like fiddle, try different new ways, you know, play with different theories, just prototype. When you get something that feels like, yeah, this is about 40% of what I need it to be, then look at how do you how do you put it into um into a proper project, into something that's enterprise ready and something that you've got control over, yeah, so that it's it's gonna stand the test of time. I mean, we don't fall into the trap of um, and we saw this uh in the RPA waves, I don't fall into the citizen developer trap where you have organizations full of inspired people going off and building automations and then never finishing them or leaving or having them run without the support. That means that you you you've got something uncontrollable that you need to pull back from in a couple of years. Yeah.

SPEAKER_02 41:52

Yeah, no, the great guidance. There's uh actually we'll we'll have him as a guest in the show in the future, but uh one of our friends, uh Phil First from uh HFS Research just had his event in New York this week, and one of his keynote speech comments was was the difference between Sunday and

Sunday Tinkering Meets Monday Reality

SPEAKER_02 42:08

Monday. I liked the way he described it, which is exactly what you just said is helping your friends in their lives. So the stuff you do on Sunday, um, and the dissonance between then what you do on Monday, which is it's just not happening in your your operations as much. But Sundays where you you tinker and learn and build a vocabulary and an awareness of the stuff, and then Mondays where you got to figure out how it does fit into um the the reasonable guardrails and and structure of of your actual organization. Um and and you're right, uh the it's tough because because they're more so than ever before. And you know, you know how I feel about citizen development and and individual creativity, it's it is so powerful and it's absolutely an asset and a resource you want to tap into, but but in a way that's that's reasonable, responsible, and safe and scalable. Um so to your point, um a bunch of a bunch of an initial prototyped ideas has a significant value to it, um, but only if there's a plan to to learn from from what came and then to to iterate and then put into production, safe production, um, what it is that that stemmed from it.

SPEAKER_01 43:22

Um well we saw this with um, you know, back in the back in the days of RPA and Blue Prism user conferences and the like, we had, I'm trying to remember, it was was it npower in the UK utility firm that was talking about how they would use RPA as a quick way to prototype uh in order to then move it into the base product. Um, and I think it's that those you know, everything you're saying about um uh you know, user led um citizen developer fit in exactly the same here, right? So you use the creativity, use the inspiration, prototype, test ideas, but then once You've got something, get it off their plate, you know. Still involve them in the project. They'll have to test, they'll have to guide and advise. Yeah. But like don't slow them down. Don't turn, don't turn your, you know, your best structural engineer or your best accountant or your best lawyer into a really bad developer. Right. Yeah. Get get someone else there to help and pick it up. But um play to your strengths.

SPEAKER_02 44:22

Which is which is, and then back to back to the Coinbase one person product teams. I mean, the who who knows better what is needed than the people who need it. And so to some degree, it needs to be a collaborative effort because you want your best engineer or lawyer or whomever to be contributing. And this is a much more powerful way than than the tools we've ever had traditionally. Um and we know this. I mean, we it's funny, we were we were looking back at um frameworks that we used 12 years ago around a sort of a scan-focus act of helping organizations look at their organization, understand where opportunities might exist, and then you know, coming up with the criteria to like let's pick how we're gonna focus on which projects would add the most value. V-word comes back, and and then which ones are we gonna move ahead with. Same framework that serves us well to this day. We used it last week with with uh with a discussion. Um, but that that focusing and acting element has a has a more collaborative prototyping um potential available to it, but it doesn't change the dynamics of organizations or complex, data is not necessarily um clean or structured or yeah or or accessible or um and processes are absolutely not documented to to really um to to build on. Um but how exciting that you can accelerate that that exploration together almost in real time, um, to then then put the right the right roadmap and the right time to to building to building it right. Um and in that case, right? Faster ROIs than ever, right? Just because we're not just whiteboarding some stuff that maybe eventually could be turned into a wireframe, you've you can move much faster in that regard.

SPEAKER_00 46:21

Yeah, no, definitely.

SPEAKER_01 46:22

Yeah, the other thing I I think of, and it's uh oh my god, I think every week we're gonna end up with some sort of Disney quote. Uh this one, courtesy or ratatouille.

SPEAKER_02 46:30

Like every Booty and the Beast was the last one. That's right. That's right. What's what song are you going to do?

SPEAKER_01 46:38

But like uh what everybody can cook or whatever, right? But everyone everyone can cook, but should doesn't mean like everyone should cook. I think I'm probably misdirecting that film somewhat. But I mean, you know, the the thought of uh the you know, the the token maxing going out there, every like get everybody sort of thinking about how they can use it. You probably don't need or want that. I mean, and also you've got to think about the messaging that it's sending out to the people that are doing the work that's probably best suited for AI. So, do you want to give someone who's doing say accounts payable clerk uh all day? Do you want to give them, you know, an a target they've got to use tokens in their work? What does that mean for them if they're successful in automating that that work?

SPEAKER_00 47:22

Yeah.

SPEAKER_01 47:23

Um so you I think I think you've really got to focus it in on where are the people that uh in the organization that um will truly benefit from it. Then actually, I think a little bit of not token maxing, but give them the resources, give them the time, you know, give them some creativity. Like I don't care if this is to do with your son's homework or your you know diet plan or your recipes for your meal planning or your you know your accounts payable. Just learn it, understand it, you know, test it, push it to a bit of its limits. Yep, then that gives you the informed view of where where this stuff can work. And that's really what you need out of those people is just that in informed understanding um to help design the new ways of working.

SPEAKER_02 48:06

I like it. Gusteau on token maxing. Yeah, exactly. I love that movie. Love that movie, so good. Um, but you're I mean, yeah, no, awesome, uh, uh awesome analogy. It's it's hilarious. We do think in Disney movies. Um but but you know, it it's it brings up too, and and we've had discussions about this for decades, um, around just incentives, right? What is, you know, what are the incentives you put in place and what is it you hope to get out? And there's always unintended consequences that come when you put an incentive in place. And in this case, using as much of a resource as possible is an interesting, you know, first order incentive that you know what percolates out of it um over time. It it's it feels like uh it's it's an incentive in flux that it will be revised and and improved over time. Right now it's it's a little simplistic and and isn't gonna get in in most cases what you're hoping to get out of it. So and now I have Ratatouille's theme song in my head, which is awesome. And now I want to go to Paris. And now I want Ratatouille. Um so um well we'll we'll I think maybe we wrap with the most recent announcement that just came out yesterday, which is um again, because it's not a day if there's not a new announcement or drop or anything else like that. And and obviously this this podcast will be produced and come out in a few days later. So came out recently, is what I'll

Claude For SMBs And The Last 20%

SPEAKER_02 49:39

say. Um Anthropic dropped clawed for small and medium business. Yeah. And that's fascinating just because not only is it is it a I mean, I'm not sure that I mean any small and medium business could have used anthropic before, but um the distinction here seems to be um pre-packaged um agents that do common things that small and medium businesses do, like uh like period and closing. Um and that's that's gonna be fascinating because they they do say that it hooks up to all the systems that you use. So it uh mentions like QuickBooks and HubSpot and Canva and other things like that, which by the way, it it I think it connected to anyway. So there's it's it's it's not new it hooked up to those things. Um but the message now is that every small and medium business can just use Claude to identify all of those um those functions. Um could be, I mean, it's more of the narrative, um, but it still leans on the small and medium businesses to have all of the readiness elements in place, which they just they don't have. Yeah.

SPEAKER_01 50:55

And it's it's a great pitch. It's a um, I think it's something that will help a lot of companies think through how they could deploy it. You know, what does this this sort of triage one look like? I think to a certain extent, and this is you know, my natural sort of skepticism coming through a little bit, so uh it's a bit unfair. So the the caveat is out there.

SPEAKER_02 51:15

But I think David is is grumpy and old man, yeah. And unfortunately, it comes from just decades of experience. So go ahead, let's hear it.

SPEAKER_01 51:26

But I think I think it will be the similar sort of story in that you'll get to play a little bit with it, see what it is before you sign up your your uh team plan, you'll sign up the team plan, you'll then go to implement it fully, and again, you'll probably get to 70, 80 percent uh of an implementation, and then it will sit on the desk for three weeks, six weeks. You're not blaming the tool, you're blaming you for not having the time to finish it off. You're um and so you don't cancel the subscription, you keep the subscription going for another, say, three, six months, um, at which point you still have perhaps you've implemented some basic sort of reports and dashboards that you may or may not use. But are you actually achieving the full vision? Um I doubt it. And the reason for that is because every, and you know, Ian, we've seen this so many times. You can on a surface and in a product um mindset, there is no reason why you would have anything different in month close, period end close, uh, in one service firm or one manufacturing firm to another, yeah, until you get into the detail. And then you start seeing all of the complexities, all of the nuances, all the different reporting and statutory and blah, blah, blah. And then that's where the detail is. And trying to get that into a place, I mean, it was is, you know, we saw this back in Cap Gemini days, like when we were working there 18 years ago, you know, with the global process model. Um, and uh, you know, global process model was great for selling, it wasn't necessarily something that was used a lot in delivery because um, you know, nuance, you know, you you every client wants to know that they they are buying into best practice until you get them, tell them that they need to change, change the way they work, and then all of the nuances come through. That's so I imagine that's what we'll see in the SMB talking caveat. I haven't seen it, I haven't played with it. So um this is just natural sort of skepticism on my behalf. So I apologize for that. But you know, history tells me it'll do a really good job of getting you to 70, 80 percent, and then it will get tricky.

SPEAKER_02 53:28

You know what's funny? So I was I was on the solutions sales team, and we we had uh one slide in the deck, and I gave this presentation a hundred times, if not more. And we would get to that point where we talk about our global process model and we talk about how we've incorporated the best practices of finance and accounting from our other clients into a global process model. There was, I can't remember a single time when the the prospect we were pitching to asked a question or gave a crap. And we would move past that slide and never come back to it ever again. It was it was it was never something that they perked up and really wanted to double click on and understand. Like literally, and I remember I remember specifically feeling the sort of like the the just the the just the cacophonous silence after that slide. And I was usually the person presenting that slide. And so this idea that we have figured it out for you and we've got the way to do it, and these are big enterprises, so maybe it's different for a small or medium business, but they didn't care. Um and mostly because they were all convinced that they were special and different, and it's because they were really they all had as much as I hated to admit it at the time, and I now accept it uh now, um, they were all different. Yeah, and so so maybe these solution, yeah.

SPEAKER_01 54:55

Maybe these package solutions solve that, right? Maybe it that's where AI can give the nuance and the uh the differentiation that allows you to do things separately. Yeah, but uh again, it's a it comes back to those themes. Like, I mean, is it gonna be broad enough, broadly capable enough to candle, handle, to candle, to handle the uh the the variety of business rules and the unique nature of each business? And secondly, is it something that you know SMBs have the capability to deploy themselves, you know, or and the time as well to make it work?

SPEAKER_02 55:24

Well, and that's what's gonna be interesting because ultimately they'll they'll see, right? This, this you've you've you've gone for a particular version, you've decided to lock in, they want to be sticky, they want to scale, they'll see that you've gotten 80% of the way there and that it's sitting, right? Their their customer success management team, which will will all of a sudden have to grow exponentially. The services investment, right? Will will follow up with you and say, like, hey, David, I see that you started as something and that you didn't finish. What can we do for you? Um, and the challenge is, especially in the smaller side of the small and medium business, um, these are not organizations with massive budgets for consulting. And so, so these these companies will have to decide on the bang for the buck as far as is putting a forward-deployed engineer or whatever we call them, um, almost on site to help you get it that last 20% of the way, which we know is, I mean, from a Pareto perspective, that 20% is is is is 80% of the heavy lifting as it relates to customization and um and then just the the hardening of it so that it's safe and that it runs reliably. So um fascinating.

Closing Thoughts And Listener Invite

SPEAKER_02 56:33

So anyway, I I guess for for anyone who's still listening, we we certainly um are are true to our word that this is unvarnished and this is honest perspectives on the agentic era, and we don't sugarcoat and we bring a whole bunch of our our grizzled trench-earned experience to the table. And uh, and yeah, thanks for thanks for listening this this far. Um David, pleasure always. I'm gonna go whistle the Ratatouille theme song for the next two hours, and uh and we'll probably just jump on another call in a second.

SPEAKER_00 57:08

So uh sounds right. I'll see you there. Yeah, thanks everyone. Great show. Thank you. Bye-bye.

SPEAKER_02 57:15

I hope you enjoyed that episode as much as we did. Thanks a million for tuning in. If you liked that, please subscribe and follow along. We have so many great episodes planned. We can't wait to share them with you. And if you're feeling particularly generous or inspired, please share this show with your networks on LinkedIn or wherever else you share things, and post about your favorite moments. And finally, if you feel like you have a strong point of view, get in touch. We'd love to have you on the show. Thanks for listening to the Magentic Show.