Automating Chaos, Building Custom Tools, and AI Readiness w/ Michael Greenberg (Part 1) - Ep. 22

SPEAKER_00 0:00

Today's guest is Michael Greenberg, founder of Third Brain Digital Operations and Automation, and the author of the Digital Operations playbook, AI Readiness for SMBs, and someone who has spent the last four years in the trenches helping small and mid-sized businesses actually get ready for AI. His take? You can't automate chaos. You have to fix the foundation first. Michael also hosts his own podcast, Five Q's, with Gent of Tech, where he talks with innovators and business leaders about AI and the future of work. Michael, thanks for being on, man.

SPEAKER_01 0:35

Thanks for having me.

SPEAKER_00 0:37

So when it comes to AI, uh one of my favorite topics, but which camp do you fall in? I I kind of see two pretty distinct camps of people I talk to. There's like the the big scarcity camp and the people who are uh kind of sitting on the sidelines waiting for something to be packaged and handed to them, and then there's the founders and uh that are going out and uh playing and and creating stuff. Uh I think I know which camp you fall in, but what's your what's your overall overarching view?

SPEAKER_01 1:10

Yeah, I think it's my job to be in the second camp. Um, I'm probably I'm not as like AI hype as many people that I talk with and know. And I'm I'm certainly not in the oh, everything is going to change in three years and all the jobs are going to be erased um camp either. But I also believe it is very useful now, and we experiment with it, we build tools with it. It has fundamentally changed the way we do much of our work at Third Brain, and it's allowed us to change and improve our business model as a result. And so I think I'm a very big fan, and I think it's a very powerful tool, but it is certainly not replacing us tomorrow.

SPEAKER_00 1:58

Yeah, no, I and I think that's where you if you get onto X, and I know you have a pretty big following on X, so you you see also what goes on in there, and there's a lot of uh probably in both of our feeds, there's a lot of people uh that do overhype and put a lot of clickbait out there, uh, which is sadly I think is driving uh a lot of the scarcity for some of these founders who are like almost it feels like think that they've missed the boat or they're too late to start, which um is you know is is something that I have strong opinions on. Uh, you know, like it's it's what I think what once you get going and once you start playing, it really your growth and ability to understand and use and build with AI really does uh multiply pretty quickly. Uh it's just getting over that initial hurdle where I mean it is overwhelming too on Twitter and on X, like there's new things shipped every day. Uh, and it's kind of hard to keep up with it all. Um but uh um but I I agree with you with it being at a place where it is useful, uh, where you know, like I I think if we zoom out and look 12 months ago, um how useful uh it really was. And at least in my business is uh it's you know drastically different. But what what have you seen as the biggest uh changes over the last 12 months or even just this year, uh, with how useful it is?

SPEAKER_01 3:37

And so uh is it alright if I date the recording?

SPEAKER_00 3:44

Yes.

SPEAKER_01 3:45

Okay. Uh so we're we're recording this the the week after Fable 5 was uh pulled off the market. Um and it will be back. Uh I think we all know that. Uh there's no way it doesn't come back. It was not a drastically better model uh than some of the models that have come out this year and last year, but I think the big thing, at least that I'm seeing, is the quality of models inside of harnesses. So those things that we use to create agents, the models that they are building now and training now are drastically better at operating inside of those harnesses. And so that means that if we have uh and like we have an internal tool that we use that does quite a bit of the consulting work for us and that processes all the documents and organizes operations and makes recommendations. And the ability for that tool to operate without a human giving it another prompt from six months ago to now is three or four times better. Um and that that is starting to happen everywhere.

SPEAKER_00 5:16

Yeah, I mean, and I think you know, especially when you're getting into it as well. Uh, and I like I see a lot of those uh like we're using certain workflows and tools that we're building internally for uh the same, you know, similar type of use cases on a consulting uh side of things, but it's the yeah, like it's like uh I think it's the difference between maybe six, twelve months ago where it was individual prompts and uh where it really wasn't that streamlined in the way that you would use it, especially from a consulting basis, to now being able to shift to having it work on loops uh without without the human input, is is like um I mean it's it's changed the way that we do uh reporting for clients and how data is processed and like conglomerating different sources of data all together that with AI removed from the equation are all important pieces to the big picture for you know how we're serving clients and you know for how anyone serves different clients? Or um, but it like the ability to bring that all together and not have to prompt it each time you want it to run uh is something that definitely has been uh an advancement that uh you know, at least that um I'm enjoying right now. Um but what other kind of things um like what's a specific example of maybe six months ago versus now that uh has uh has really changed uh and affected the like the output, especially from a consulting kind of point of view?

SPEAKER_01 7:06

I mean, honestly, I don't think we've had as much change there. Um and this is mostly just because like we can do more faster now, but ultimately the quality of our work I wouldn't say has fundamentally improved or changed. Um the ability for us to price certain kinds of fixed scope projects, like if we're replacing an existing tool that maybe they're paying on some exorbitant annual contract for and replacing it with an internal version, that is now cheaper for us, and we can scope those more effectively. But fundamentally, most of our work has not really shifted, and that has more to do with the fact that the work we do is foundational. It's the step before AI in many cases, or and we were already building a lot of automations, we were already building a lot of these pieces, and so it it's for us at least, it's much it's more we can do more faster for some of our clients rather than we do anything that differently, um at least in the past like six to twelve months.

SPEAKER_00 8:20

Gosh, yeah. So the work that you guys do, um I mean I know you you talk about uh having a good foundation first. Um, and I think you probably see the same kind of thing that uh that I see and a lot of uh people see in um client service type work is um the you're trying to automate ad foundation just uh you know is just putting on autopilot uh things that don't work or in the first you know in the first place. It doesn't fix the the core problem. So um I so I definitely like I uh am you know fully uh aligned with your approach with uh like fixing the foundation and um I think uh the best the best way that I had ever have ever seen it uh really portrayed was uh in when I was reading um the Walter Isaacson um biography for Elon Musk and it w Elon was talking about uh when he was coming into I think it was uh one of the uh Tesla factories um and he's his team were trying to automate and you know add tech to um the assembly line and he came in and he's like no first you have to fix the how things work in the assembly line before automating anything um yeah and that's you know that really was something that clicked for me and and is how we're building out systems uh on you know finance and accounting side of things for clients as well but operations too I think I feel like operations and AI are kind of like two words that more and more go hand in hand uh with different companies now and um and so like what's uh what are some what are some areas that founders operationally are maybe not looking at at you know their company's operations through the right lens or where's where are some of the biggest operational flaws that ai can help but that need a good foundation first uh yeah good question uh so first off we we have an internal phrase uh that I often talk about with clients you can't use a chainsaw if you can't swing an axe um and so you know we're very much have the process down by hand before you automate it uh don't bring a power tool in before you have a system that can support it uh and sort of the the corollaries are garbage in garbage out which I think is a classic one and you can't automate chaos uh so if there is no order there's nothing to automate um I guess sort of the the broad message is that AI is not going to solve that problem for you it might be able to point out where the problem exists if you have the right setup um but and it might be able to recommend like oh here are the best practices on how this is done inside of an enterprise or inside of a larger org but if you're an SMB if you're in Main Street if you're running like a few million dollar a year business or smaller or a little bigger you have too many exceptions to have that strict fully automated process you have to be able to bump out uh to a human still and I I think we run into that all the time where people are trying to automate everything you know oh let's fully automate our outreach process or let's fully automate AP and realistically it's oh well none of this is actually standardized and when we get to AR then it by it turns out like oh well this client has totally different terms than all of our other clients and yeah we hard coded our standard terms in so what are we going to do here?

SPEAKER_01 12:41

Well we're gonna create a single exception for that client no that's just bad design um and so uh yeah I I think fundamentally you kind of have to have it working with people and the great thing is right now AI is not that much cheaper and in many cases it's not cheaper than just using some uh global talent to support and so at this point you can still very much rely on lower cost global talent who knows how to do these things and who can learn and help you design the processes and then when it comes time to hire the second person to do the same thing you can deploy automations to take 50 or 60 or 70% of the work off their plate.

SPEAKER_00 13:26

Yeah no it's um no it's funny because we uh well like we serve kind of I think similar types of clients and it's the word that I uh always come back to is uh is is there's just too much like there's too much nuance with these companies like yeah like you you explained it perfectly like for an example real world example that we see every time we have a new client they all you know most of these clients all have receivables and payables and when any of these clients come on board there's never been a client that has done AP this exact same way as the other as another company and there are so many nuances some there's nuances for literally every single vendor or customer and it's like oh no this is this is Chuck I you know I play golf with him he pays me and cash and pro shop credits like you know and it's like okay we well we can't work off that um and we can't create a system that's better when there's you have multiple different rulebooks like like how how would you how would you possibly create an SOP for that you'd have to create one for each vendor or each customer and and so we that's one of the first things we uh we do when we come in uh is we identify those processes that are like okay let's let's try and uh find ways to streamline this as best as we can because like also with the types of businesses we're working with like I'm not gonna make you like I'm not gonna try and force you to change certain nuances that are just like quarter like quarter your DNA and like is like you know like that there's just certain certain nuances that are just always going to exist and it's honestly the a lot of the beef that I always have with different uh in the past different softwares and as in tech stack especially um ones made for the accounting and finance industry that particularly the ones that are branding themselves as AI automation and I don't even have to try these tools to know like when they're when you know when they put these stats on their homepage on their website and you know automate 90% of XYZ and it's like yeah but based off whose process right and it's like that's and there's so many nuances and like I've never had a uh a a software uh for anything that works perfectly across the board for each client um which is where uh one of the biggest things for uh that um I've spent my time on uh this year has been uh you know I was like I realized we've got to this point and in time where you don't have to you don't have to put up with it anymore. You can't you can just go out and build tools uh specific to the nuances you encounter and so that's what we've replaced a lot of different softwares in our tech stack like we we used to pay five six hundred dollars a month for a a reporting tool that I really didn't like because it was just this you know generalized templatized way of looking at a company's financials but then you would have a client that would love to you know I really want to see uh you know this presented this way or can we do this and like they're just that they all lack the uh the customer uh the customizable uh ability to actually give a client exactly what they want and then it's like then it's like I feel like okay now I'm in a hard spot because I expect to give the client what they want because that's what they're paying us for as an outcome. But we're handicapped by our tech stack because it doesn't allow it to us to do it where now we don't even have that uh reporting uh software subscription uh and I just I've spent the last month or so completely rebuilding our own um reporting uh building our own reporting tool and where you can bake in the nuances across clients and again they're they're they're not perfect uh because it there's always uh more nuances uh but it uh they're a lot more valuable than these cookie cutter softwares that are out there yeah and that's I'd say a large part of what we end up doing with clients is identifying where those nuances those edge cases exist that you should build an automation for and where you should not where you should still be using a human. What's your take on communication? See I'm a I my I'm very much in the camp that client communication will be the last thing if ever replaced by AI because it's pretty obvious um in a lot of cases when you're being communicated with an A an LLM messaging you uh or emailing you versus a human. What's your take on that?

SPEAKER_01 19:08

So I think at least in like professional services where we live uh our clients are paying for people um and with third brain I'd say maybe 80% of our engagements are embedding team members directly into client organizations and so for them they are communicating with a third brain team member who is living in their company all day every day uh full time and that means that they never have to talk to an AI. But if your deal if you're like a SaaS company or you're an e-commerce company and you have hundreds or thousands of customers and you might have hundreds of inbound requests every day using some sort of chatbot support to cover the first layer or two makes sense. It just comes down to oh like we have a client now where we are building them a new platform and something like 60% of their client inquiries are related to either password management or billing questions. And in almost every case those can be handled by like a wiki or directing them to the right page on a on a site. And so that's a case where we would deploy AI for that. But the other 20% of cases are what the humans really want to be working on. And if we're able to I mean if we're able to prevent them from having to hire additional customer support staff or even reduce the number of staff entirely then that's the goal and that's the win.

SPEAKER_00 21:00

Yeah uh yeah I mean it's yeah there are uh there are certain uh there are certain things where it's uh uh yeah that I think that's where the frequently asked questions page was born and should be more front and center and more companies should have them including us um but uh uh yeah no like it was funny like I saw I saw once uh recently a um I think it was a LinkedIn post and it was uh guy put a screenshot of um of two two people having a conversation in a LinkedIn DM and it was so blatantly obvious that it was to one chatbot talking talking to another uh and copy and paste and it was like that was the first time I've seen it like on both ends and I was uh I saw that and I was like wow like that's like It's the other side is not going to even be pissed because they're also using a chatbot to copy because but then the flow of the conversation just made no sense. And it was like one person was trying to basically it was basically get on uh like get on the phone and network with this stranger on LinkedIn. And the the opening was given in like the second message where uh you send a call, you know, a call link and it would have been scheduled. Like the yes had already been reached, but then it was just this overcomplicated back and forth between two chatbots analyzing over analyzing uh and responding, and it completely missed the opening uh for each other.

SPEAKER_01 22:46

Yeah, well that's the issue with LLMs in the current system is that they they're trained to keep talking and to keep you on the line, and so even if they reach the yes, unless you have some sort of system that it's gonna catch, oh, this is the natural end of the conversation, I should stop talking here, it's always going to send that next message.

SPEAKER_00 23:10

Yeah, no, yeah, you're right. Um so when you talk about AI readiness, um what does that what does AI readiness actually mean or actually look like, in your opinion?

SPEAKER_01 23:30

Yeah, so I I think there's there's a few the key things that we look for. One is can we point to where the single source of truth is for every type of data inside the organization? So do we know where the definitive tracking of invoices lives? Do we know where we can say, hey, is this a client yet? Uh where are we tracking each piece, each type of record? And do we have a single source of truth for it? Because if we don't, then the record could exist in multiple states and multiple places, and that creates confusion for a system. Number two is having basic processes outlined. And you don't always need them, you know, step by step down to every click mapped out. But we do need to know, like, okay, well, we don't consider a client onboarded until kickoff or after kickoff, after they filled out the quest this questionnaire, or do we consider a client uh onboarded and going once they're signed and paid? That's a big difference inside of a process. And so understanding the basic scope of the processes, understanding what they are, and then building on top of that. So, you know, we've got like that data foundation, and that data allows us to map and build out processes, and then above that, we're able to then define individual roles for humans and potentially for AI or agents as well. And that all then allows us to say, okay, there are really only three things that implementing AI or automation in a business are going to move. It's either going to make the business more consistent in the way it operates, it's going to make the data clearer and give more clarity to the understanding at the leadership level or at the individual level of what people are trying to achieve or of how people are performing, or it's going to increase the capacity of the business. And in most cases, it starts by creating clarity when you reach that foundational level, and that allows you to build more consistency in your operations. And then as you have consistency in the way a particular thing runs, you're able to automate more of it, which then creates capacity. And so generally it runs in that flow. And we know if you don't have clarity, you don't have a foundation that's going to work.

SPEAKER_00 26:06

Yeah, absolutely. And so do you do you think most businesses are AI ready for what's coming or what's already started?

SPEAKER_01 26:18

I know most aren't.

SPEAKER_00 26:31

And it's unfortunate too, because I've um, you know the amount of operations mixed with a you know, like operation AI companies, automation companies that have popped up over the last six months, even is unfathomable. Um and I've um I've been introduced and talked to a few as well, and um you know, like I know you from Tribe, our founder, our uh founder group, um, and I know that um you know the way that you approach operations uh you know and f in these businesses, uh not, you know, like you you didn't this didn't just stem up just because you saw the money opportunity of AI and you know um the amount of businesses that are gonna need this who don't know any better. Um and so and so there's you and then there's um and then a good friend of mine, Brian La Framento, he also has an operation, it's operations first, AI and automations as a component to make that more effective, which I think both of you really do approach this uh in a similar way. Um but the some of these other companies that I've I've talked to, like um it really showed me, and it's like these are companies that have paying clients on pretty large retainers who then just in a conversation with them talking about AI. I mean, I talked to one I think about a month ago, and I and we're just you know just chatting, talking about AI, and I asked I asked him if uh it was maybe a little over a month ago, it was right when uh Claude released uh Routines. And I asked him if he's uh uh if if they've uh used that yet and what because I had not yet it came out like maybe two weeks before uh this conversation I was having. And he he said to me, um, he said, are you talking about that cowork thing? Uh I haven't well we haven't tried that yet. And I was like, cowork came out like what like late summer of last year.

SPEAKER_01 29:09

I don't think it's been quite that long, but yeah, it it's it's not a new tool.

SPEAKER_00 29:13

Um and it it and it's been converse like conversations like that have been the um have been the majority apart from uh you and Brian like talking about this stuff, and I'm like, wow, like this is kind of predatory in a way. Like there's these companies that are framing themselves as these experts in operations and AI, and like I can't like I can't imagine what they're like they're they're coming into these companies and like I like I think there's a l there could be like there's a there's definitely an opportunity for a lot of these business owners who need this help with what you guys do that uh are could be you know really getting taken advantage of from all these operations companies that have just st you know started up overnight. What would you if you were to talk to a business owner who knows nothing about AI and has you know probably the the majority of uh the clients that come uh on with you guys in the beginning who really don't have any real grasp over their operations internally how would you what would how would you advise them if they were talking to some of these companies how to vet them to like cut through the ones that are basically gonna charge them in these retainers and have zero act like real impact?

SPEAKER_01 30:57

Yeah. First off, before does Brian host Wantrepreneur to entrepreneur?

SPEAKER_00 31:02

Yeah.

SPEAKER_01 31:03

I think I've been on his show. Um really I think so, yeah. Uh I uh I'll double check afterwards, but um, I think that there's a good chance uh because I was like that name sounds pretty familiar. I think I might have recorded a podcast with him at some point.

SPEAKER_00 31:21

Well that's that's that friend the friend that uh um has helped me set up uh no trade secrets, and uh we work together on stuff too. Um yeah, um I'm not surprised. He's he's had over 1,500 guests now, so uh that's a small world.

SPEAKER_01 31:38

Yeah, uh podcasting especially is a pretty, you know, the there's only a few of us out there. Uh but I think there's a few things to look for. One is I would not hire anyone who calls themselves an AI agency or an AI automation agency. Two, figure out if they are, and the the breakdown here is fuzzy, but in many cases I have seen just standard dev shops, development agencies, whether web or mobile, that sort of thing, pivot and say, hey, we do AI now too. Don't hire them unless you have like a strong product vision and you just need engineering talent. Um, because they probably don't have the operations consulting. On the other side, you've got people who got laid off from their jobs and operations or who are looking for the next quick buck and they are doing maybe some basic operations consulting, and then they're using no-code tools or low-code tools or open claw, things like that, to set people up. And can that be valuable? Yes. Do I think you can probably hire somebody on upwork for a fifth of the cost or a tenth of the cost and get the same results? Also, yes. Um, so I would recommend in those cases, you just avoid getting locked into any sort of retained contract, and you go find the hourly expert on Upwork that you can hire to guide you through that setup together. Um, but then there's a few other pieces to look at. One, and this is just a general rule for service providers, age of firm. So, how long have they been doing this sort of work for? Um do they have standardized processes and methodologies to do this work? Do they have a team that they've trained to do this work with them? If they're in the AI space, what sort of tools have they built to allow them to do their work? If you're hiring an AI consultant or an AI agency and they have no internal AI tools that they actually use, huge red flag to me. Um and then finally, and this is this is very much on the operations and on the transformation side of things. So not as applicable if you're like a solo founder or you've got a very small team. But uh a lot of our clients might have teams of 20, 40, 50, 100 people on them, and digital transformation, the change management process becomes a much bigger piece of this. There's really only one way you have successful digital transformation or AI transformation in a business. And it follows three parts. First, you get the people on board, then you understand how they do the work, and then you design or build or pick out the tools to fit those needs. And so it has to go people, then process, then tools. If you see them starting with tools, they are not a great fit for most organizations, to put it uh kindly.

SPEAKER_00 35:06

When you say getting the people on board, um I'm curious. Um, do you ever run into the founder? Obviously, has engaged with you to uh because they're on board, but then um, and you know, uh you know, we've seen this uh from some of our uh kind of you know more legacy mom and pop businesses, you know, that are uh you know you know teams on the in these companies that have been there, uh, you know, they're of the older generation, um like a hesitancy to get on board with with uh technological changes. I'm not even talking about AI, uh, but just like even uh you know, like a common one is uh is getting a team um to change like uh getting everyone fully on board with just the process of hey, let's um let's as a step one, let's let's get rid of paper checks in your in your business. Let's let's let's let's let's start there. That's a real easy one. Um you know there's uh there's some really great tools there that you know we can collect vendor ACH information just with their email, but if for some reason that vendor absolutely has to receive a check, you you can pay you can pay them with a check through the platform, but you don't have to write the check, you don't have to sign the check, you don't have to go to the post office. Um makes everyone's life easier, but like that's one that there's there's uh a shocking amount of kickback on uh for some people. Like how how how how do you navigate when a founder's on board, but maybe their team isn't fully on board because of their you know uh resistance to change, or I've we've done this for four years uh and you know and we're still alive, like why why change what's not broken? You know, that way of thinking.

SPEAKER_01 37:13

Yeah. So the majority of our clients are post-transaction legacy businesses. So I run into this all day, every day. Um, and I'd say it's not even every time that the founder's on board. Sometimes we just got the COO or a head of operations or somebody like that, or we get brought in by like uh some other department for work there, and then it sort of starts to spread throughout the org. And there's there's a few tricks we've got that sort of help get people moving, but the biggest one is sitting with them, making sure they're heard, making sure you understand why they do the things they do and the way they do them, and helping show them that there's a better future. Um, you know, if you've got somebody who's 63 years old, they're retiring in another year or two, and they've been doing it that way their whole life, they're not gonna learn. I I can just tell you that. And then we got to go have a conversation with the client and say, look, we can force this thing through, but Gary's not gonna change the way he operates. He's just playing golf and maintaining his book of business till he retires. So either you hire him an assistant who's gonna actually run these things for him, or we come back there in a year or two when he's getting ready to retire or after he's left. And that's sort of the worst case scenario. Um, in most cases, though, they're not at the end of their career, they're just maybe in their 40s uh or 50s, and they haven't had to change, uh, especially if they're on paper. A lot of times the business is in an industry that hasn't gone through that digital transformation at every level to get onto the cloud and things like that. And then you've got sort of two main tracks. One is showing them how it makes their life easier, how they can do less work, how they can get more done, play more golf, and how they can react. Yeah, exactly. They can play more golf because hey, I don't need you to input everything into the CRM. I just need you to call this number after every meeting and tell it how the meeting went so that way it can log the notes into the CRM. Um and so building, you know, if you have to build an extra step to then make it a more natural process for them, that's not the end of the world. If they are still resistant at that point, then we often have to find a champion somewhere else in the organization or the department first, where we can start that person on the tools, and then they start raving about how much better everything is, and they get the little award on their desk, or they get the shout-out in the meeting. Oh, and so-and-so built this thing and it's amazing. Uh and people want that recognition, they they start to get on board through the social pressure instead. And I I mean, obviously, that doesn't work every time. You can't solve every single one of those cases, but those solve the vast majority. Um, and we've seen you know, examples of even engineering teams where, oh, operations is now outcoding the engineering team because they've adopted these practices. And the team member that we embedded with them, they are getting a lot of praise and they are shipping more faster. And they shipped the entire backlog of operations work from the engineering team. They did it just in a month. And then, oh, well, how are these guys doing that? We gotta get on board. Let's learn how to use these tools, and it spreads from there. And so I I generally tell clients you know, that sort of adoption of new technology spreads more like a virus than it does like a commandment. Uh, you've you've got to get it bubbling up inside the org and then watch how people adapt to it.