Smarter Starts with AI

Webinar Transcript

Mohamed Hamad (Host): Hello, hello, hello everybody, and good morning. Welcome back. We are very excited today to be talking to Mitch Schwartz from Ops Machine. And today’s topic is Smarter Starts with AI. And this is quite a big topic and there’s a lot to cover over here, but I’d like to introduce first Mitch.

Mitch is a good friend of mine and he is a process designer and an AI consultant and a CGLCC certified consultant. So I’ll let you give your background and a little bit of an introduction before we get started Mitch.

Mitch Schwartz (Guest): Yeah, thanks Mo. So, I mean, one I trained as a designer originally actually designing products and the whole thing was what problems are we really solving for people rather than just making something that’s cool for us or fun or fancy. 

And as I’ve carried those skills forward into process design, that’s what I’ve really helped teams with is like, how do we take these systems that are maybe seem really complicated, but they’re supposed to do so much, and how do we make them work for us in a way that’s great and not painful and really moves things forwards?

Mohamed Hamad (Host): Yeah, that’s great. And I think that really lends into what you’re doing these days, with AI consulting and helping teams understand how to integrate AI into their processes, into their business, and into how they do everything basically. In saying that, it is such a big topic that everybody’s talking about.

There’s a lot of promise with AI in terms of how to use it, when to use it, how to get the most out of it. And every other week there’s a new feature and a new something. But as we’ve discussed, over the last six months to a year, when discussing AI, it’s really down to how and when to use AI in your organization.

So the when is a big piece here. And maybe you can speak a little bit about how to figure out when to use AI and when not to use it. Because you know there’s things that are, they’re really good at and other things that they’re just, you know, let’s, let’s skip that.

What do you think?

Mitch Schwartz (Guest): Yeah, I think the most popular answer to when is like, oh my God, right now you’re so far behind. What are you doing? So then after that answer, it’s like, so what’s repetitive? What’s draining you?

Right. One thing that I always tell people, and I think a lot of them find this grounding, or at least it looks like they do at the moment, is, you know, AI is a tool. As an automator, it’s just like another thing that we have. Whereas, like, people used to say, hey, can you automate this? And I’d be like, that’s going to be really expensive.

I don’t know. That’s crazy. Now we’re like, yeah, I think so. I think we can do that. I frame it that way because. Okay, great. So how do you go about automating something that doesn’t turn into a mess and that brings everyone along for the journey, doesn’t create all this resistance? You do that through discovery.

You do that through mapping out your processes. So I’m saying that really abstract to frame it, but what that means is what are the things that, I’m doing frequently? High volume is great because you get a lot in return for it if you try to automate 17 steps.

And we often don’t realize, hey, make this LinkedIn carousel. Great. That’s figure out the topics. That’s, write all the different slides for it. Create a CTA, tie it back to the product, or what you’re selling. Then it’s create the images.

Right. So that’s so many things. But hey, can I get AI to read about my company and then suggest topics for me? Yeah, from there can I have it say, hey, here’s what each slide might look like?

Yeah, great. So when we break things down into steps and figure out which individual step is taking the most time, that’s really the best place to start. And then if we just automate that one part, then like, hooray, we could save so much time and we can do a little bit of extra clicking to bring it from here to there and build it out further.

So that’s how I really like to think about it.

Mohamed Hamad (Host): Okay, I mean, what are the things that you would say, AI? I mean, and this is saying that it’s always changing and it’s always moving. So whatever we think is good at right now or terrible at right now is going to change completely in a couple of weeks. But historically there have been a few things that it doesn’t really do too well unless you really customize it in a way or really work your magic with it to get it, get the right fit.

But the first thing people think of is writing blog posts. Writing social media. Writing is usually the big thing that people think about the most. But what would be other scenarios you think that AI would probably do really well to help out teams really improve their internal efficiency.

Mitch Schwartz (Guest): So there’s a few. And so one obvious one is sales. But I think the core in what I’m saying is like research and pulling out insights. And so I was sitting outside of an event that they’d sent me the list to. I got there 15 minutes early and like, wow, I’m so prepared in the head.

And I never looked at the list of attendees and I had no idea who to talk to. So sitting there on the bench downtown, I pulled out Claude on my phone and I had a project that already knew about my company and that made it easy to do this. I said, hey, here’s the list, like, here’s the PDF, who should I talk to?

And it gave me a list and I went, okay, like, what are each one of these companies? Can you research them? And it went bam, bam, bam. Like researched every company, re-ranked them for me. And then there we go, like, here’s the top guy you want to talk to. He runs an accounting firm. There are 300 people and that’s just ripe for AI automation, for these reasons.

And that actually turned out to be a great person to talk to. So, if we’re getting tons of emails from clients, can we pull what they want from it or insights? Crunching a lot of data is a great thing for AI to do. It’s so good at pulling out patterns. That’s one of my favorites.

Mohamed Hamad (Host): Absolutely. We had a previous webinar with Joelle Irvine, and one of the things that she’d like to do was to do deep research, on user behavior and user needs, using the deep research tools, but also diving into Reddit and community posts because it can aggregate so much information, from so many different places and synthesize that I think synthesizing large data sets or going through large swaths of information, back, you know, the documents that are on hand or research papers and all that is, is great.

Mitch Schwartz (Guest): I’ll just jump in ahead of your next question. On top of that is for anyone. I don’t know if you’ve used Notebook LM for this but when I’m doing consulting and it’s been a week since I spoke to the client I’ll take the entire transcript of the last calls, throw it into Notebook LM say, hey, this is, you know, my situation, focus on this stuff.

But, make a, you know, it’ll make a five minute podcast for me to re-soak my brain in the whole project and remind me of all the details and nuance that I’ve absolutely forgotten in that time. And I know people use it for sales calls as well, so it’s really powerful.

Mohamed Hamad (Host): Yeah, absolutely. Notebook LM is a godsend, especially when you’re trying to get overviews of, big things like giant reports that are 30, 40 pages long and you just want a five minute podcast to listen to while you’re doing the dishes.

Mitch Schwartz (Guest): Yeah.

Mohamed Hamad (Host): Now, you know, one of the things is knowing when to use it, and getting creative with the places that you can use it in, is how to talk to it to get the information out of it. You know, there’s the promise that, just with a few sentences or one line and if you say it right, it’s just going to give you magic.

Right. And the idea of prompt engineering, which is a term I’m not a big fan of, how do you conversate with AI to really pull the things that you need from it? And, that’s tougher than most people think.

But it’s also a simpler concept.

Mitch Schwartz (Guest): Yeah, absolutely. I think it’s both. Right. It’s like the simple but not always easy. What I’ll notice is sometimes I’ll have these huge epiphanies and then a week and a half later I’ll just be doing the, you know, the lesser version of it again. And so I’d say it’s a lot like people.

You’ll get quality of response in proportion to how much space and time you have for setting up the conversation. And so there’s a framework that I like to use which is RICO. And so we have the role. And that’s like, oh, you know, you’re, you know, expert, you’re like Ray Dalio and I’m going to ask you about the economy now, or, you’re an expert, such and such, like, I don’t know, senior developer or whatever thing.

And there’s the instructions, there’s the context, and there’s how you want the output. And that’s a quick checklist for am I telling it what it needs? But in the story I told you before, I had gone through the process of making a master prompt, which is a lot of critical information about my company, the products that we sell or the services, our goals, our ethos.

And so it had that already when I started. So other than having these sort of preset up environments that aren’t crazy complex to do is, I like to use the voice transcriber and I yammer at it. I’ll be like, hey, this is my thing. And I’m that and this happened.

And I’m annoyed about that and I don’t want this crap. And the other thing, what I really want is, can you tell me, how would I do this? That’s one version. What I find works the best is, hey, giving it that context and saying, I want to write a blog post but I don’t want it to be some AI sounding crap.

So what are the concepts of good writing right now? What are timeless ideas to write? Something that really grabs people. Then when it does that research, now we have all this context loaded in and so I can say, okay, great, so I want to write about this thing.

How would we bring that into our framework? That’s where it becomes really powerful. Even though it’s sometimes a bit annoying to do that step at the beginning of every, conversation, it’s so worth it. And it saves that AI whack a mole.

I mean, I did a whole presentation and I literally said to it, like, I do not want to do AI whack a mole on this. We are going to go step by step. You are not going to run ahead of me. You are going to just move a few steps at a time. We are not writing the whole presentation.

We’re just doing the outline. We’re going to get that right. Like, yeah, stuff like that. Really reining it in.

Mohamed Hamad (Host): Yeah. I think, to come back to what you were saying before, there’s the role and then there’s the context and being able to speak candidly to it and be expressive of actually what you want. I think there is a legacy of us, as human beings talking to computers through a search box.

When we use the Google things, when we’re keyword heavy, where we would type in, you know, men’s shoes, Montreal, size 9 and expect results. But now we would want to express what we’re looking for. I’m looking for men’s shoes, to play tennis.

And I’m a size 9, but I’ve got wide feet and X, Y, and Z. And really kind of talk to it as if you’re talking to someone, as if you’re walking into the store and talking to the person that says, hi, how can I help you? You know, and give them the breadth of what you’re thinking.

And that usually for me works really much better. Because there is the role, there is the content and AI doesn’t really know who you are. So being able to express a little bit more gives it an understanding of what you’re all about and can answer things a little bit better.

Mitch Schwartz (Guest): You say that AI doesn’t know who you are and you’re right in an isolated new conversation. But with ChatGPT where it can access all the other conversations like just, and I think you’ve tried this, but to the people watching, just try going to ChatGPT and saying based on all our past conversations and then say like write up like a FBI style profile about me or what are my growing edges.

And it kind of knows you but you do have to remind it that and it’s not so different from people because often we work with employees and it feels like micromanaging and there’s ways to do it. But hey, like remember what are the best practices? So yeah, I would say on that point, I don’t know, it kind of does know us or it can figure us out with reasonable accuracy.

Mohamed Hamad (Host): I think for businesses, in general it’s great to have briefing documents. So documents that you would have to give to a new employee or to an intern for instance, just to give them the lay of the land, those types of documents as baseline to be able to train it to understand what you are about what the business is about what the products and services all of that and then have that foundation and build on top of that usually works really well.

And that also adds to the sustained relationship between you and the AI in general.

Mitch Schwartz (Guest): Yeah, absolutely. It’s a great way to think about it is what would an employee need from you to do the job well?

Mohamed Hamad (Host): Exactly. There is the idea of thinking of the AI as a tool, but then there’s also as a team member and how that relationship pans out or how that relationship manifests between a tool and thinking of it as a team member or as a personality is an interesting conversation to be had.

It’s quite the philosophical and meta conversation.

Mitch Schwartz (Guest): Yeah, I was gonna say some people consider their co-workers to be a bunch of tools. But that aside, it’s funny to split up where it is to be treated like a human. Because I find a lot of people get way better at prompting with just that.

Pretend you’re talking to a human. Right. It’s not Google Home, it’s not Siri. It can understand you and then, but no, like it can’t do this end-to-end task yet. And trying to make it do that whole thing is going to be insane. Or at least not insane, but it’s like a 10k effort when all you need is like a $100 effort to get it to the next level.

So interacting with it in that way, yes, but maybe not having the end-to-end expectations of that at first, I think is the sweet way to the sweet spot, way to split that up.

Mohamed Hamad (Host): So what you’re saying is basically take what you’re expecting out of it and break it down into smaller chunks and work and iterate with it over time to get to the end product instead of expecting it to give you everything under the sun with just the single prompt.

Mitch Schwartz (Guest): Yeah, like going back to the example of making a LinkedIn carousel. If I just say, make me a LinkedIn carousel and I go straight into Gamma and do that, it’ll get me something. But if I have my company context and I like to use clothes. So if I’m there and I say, hey, we’re going to write a carousel and it’s based on this document I’ve already given you and here’s the goal of it, then it outlines it and then I iterate that and I take that and paste it into Gamma.

So I’m going to basically work with it step by step. And by doing that, yeah, now it’s a tool in a sense. And that allows me to know the process and from there, okay, sure, people like you, or I can help say here’s the technical way to automate that better.

But when, you know, when I work with teams, it’s almost always that, let’s break down the process. Let me help you figure out all the inflection points and let me help you see that this one thing is actually six things. And then now let’s zoom in on what makes the most sense to actually automate first.

Mohamed Hamad (Host): Yeah, that’s, breaking things down. And there’s a lot of times where you would think that, you know, it seems simple to you, it seems pretty straightforward, but when you really break it down, you’re like, okay, this is a lot more complex, there’s a lot more nuance, but only because the fact that as human beings, we’re used to doing things, in a way that makes the most sense to us, but we also compress concepts in our minds.

Just looking at the chat over here, there’s a few questions coming up. We’re going to answer the questions. Towards the end, if anybody else has any questions, please throw them in the chat and we’ll get to them. And saying that, now getting the thing that you want from AI is one thing, but there’s always situations of it making up things, hallucinations, assumptions, there’s bias.

Mohamed Hamad (Host): And this is where the human in the loop kind of comes in. How do we make sure that it doesn’t veer off the path and really either sticks to the trained information? Or how do we get it to challenge its bias? Or how do we get it to check its own work?

Yeah, I mean, so one of the ways is to ask, right? And then it comes back to something I’ve already said, is if we want better quality, then instead of saying, hey, check yourself for bias, I’d say, hey, what are all the ways that typically AIs have bias and what are the best ways to suss those out? I’ll let it do that research and then I’ll say, okay, so apply that against what you just created for me, or even have a separate conversation that has just that one job with that context preloaded, and then bring the result in.

So that’s okay. The using AI to do it kind of way. In concept. And, so I, you know, that’s the end of my sentence. I thought I had more. That’s the number one technique, let’s say with AI itself.

From there we could talk about some of the bigger frameworks maybe, or, the ways to think about bias and issues in AI.

Mohamed Hamad (Host): Do you get it to benchmark against the framework? Against any of the frameworks that you mentioned? Or is this something where you read the content, it gives you, or read the output and you see the bias and then you say, check it for this kind of bias because you’ve identified something.

Mitch Schwartz (Guest): I think both are important. Right? Why not run both checks independently?

Mohamed Hamad (Host): Fair enough.

Mitch Schwartz (Guest): So once again, we can say, hey, here’s a framework like the AIA, which is the European act, which my best guess is that’ll be sort of like the GDPR of AI, meaning they’re the first ones to regulate it. Other regulations will end up resembling it quite a bit.

And anyways, if you’re following that now and some regulation comes in where you live and changes the rules and makes what you were doing no longer legal. Well, at least you have some reasonable defense to say, okay, I followed the only thing that I could find out there, so absolutely will change.

But it was never nefarious. We were doing our best to follow something. I’m not a lawyer, but my guess is that’s a really strong way to go. And so we can take that framework in which lays out, for example, levels of risk, like what types of decisions are okay for AI to make and what type of counterbalances or measures need to be in place for different types of, let’s say AI interaction.

Right. Is it low risk where it’s like putting filters on photos? Is it medium risk where it’s a chatbot and we say medium risk, but we’ve seen some horror stories of what chatbots tell people, or is it, you know, high risk where it’s making decisions that could affect people’s livelihoods?

You know, where it’s like HR recruitments, things like that. So, yeah, basically understanding that and then having the AI understand that and together as a team operating within those frameworks, to me is the best way to think about doing this.

Mohamed Hamad (Host): Okay, so that kind of comes back to the first conversation topic where we talked about when to use AI. And if it is appropriate for AI to be doing a specific task given certain legalities or sensitivity to the task at hand.

In general, it can help with, let’s say, the lift of certain things. But should the final decision be predicated on what is defined by the AI, or should there be a human in the loop actually making the final review and, and all that?

Okay, so we’re going in for some questions here. And we have a question here from Ivan. He’s asking, with the barrage of new AI platforms launching every week, every other week, what advice do you have, for companies who are keen to get AI into their ops but are not sure where to start or are overwhelmed by the pace of this new technology.

Mitch Schwartz (Guest): Yeah, I love that. So I happen to know it’s Ivan and not Ivan. And thanks for joining us. But with that, unknowable, just based on reading his name thing said, so I think stop worrying about it is like such a dad answer, like stop worrying.

But what I’m saying is, AI is such a huge field now. There’s so much going on that nobody can keep up with it. Like, I can’t. It is anxiety inducing. And so I think the first thing is learning to just let go of that and then go through the process of understanding what do I need.

Right. Like, I worked on a project which was doing some ad generation and we use something called Placid AI and it’s not actually Placid API. It’s not even AI but we use some AI to generate some ads and like. Cool. But now Gamma does it. Okay, great. And there’s like six different slide presentation platforms.

But like I’ve used Gamma and it’s more than good enough. So great. Hooray. I found that I needed slides done faster. I played with a couple of them. I chose that one. It’s easy for one person, so the process is a little bit longer with a team, but it’s always identify what do we really need, figure out what platforms might be the best fits, do a quick trial, small group with one or two and then we can move forward with it.

And if something comes out in the future that’s better, then we weigh is it worth changing our workflow to that or are we pretty good anyways, right? I think chasing shiny objects is going to become increasingly killer. So I would say grounding ourselves really matters.

Mohamed Hamad (Host): Absolutely. On my end, I think that most people get the most out of the AI tools that they have. So whether it’s Gemini because it’s part of their Google Workspace, or Copilot because it’s part of Office, or ChatGPT because they upgraded or Claude like you, it’s really.

All of the features are going to blend in and every single one of them is going to have the same features at some point and it’s just going to differ in their approach. And to borrow an adage from photography, because I’ve done photography for a bunch of years, the best camera is the one that you have in your hand right now.

The tool that you have is the one that’s the best. If you’re going to keep running after the shiny object, you’re going to get stuck in. How do I configure this? You’re going to get stuck in configuration hell. Just get used to the one you have and learn it and figure out its nuances and do the most with it.

There’s going to be ones that are hyper specialized to very specific things like image generation for instance. And you mentioned Gamma for let’s say creating carousels. And there’s going to be others but the general purpose one was I think do the most for everyone.

Mitch Schwartz (Guest): Yeah, absolutely. And make time to play with things. Play with the tools you have. But okay, play with a new tool. Like today I had some time I wanted to try Gamma for carousels for like two months. Okay. I finally did it because I felt like it today. Leave a bit of space and see sometimes you’ll spark something cool.

Sometimes you won’t.

Mohamed Hamad (Host): We’ve got another question here from Jason. Can you recommend a nonpartisan source for the latest AI info and updates?

Mitch Schwartz (Guest): That’s a good one. I’d have to get back to you on that.

Mohamed Hamad (Host): There’s quite a lot of resources out there depending on what you’re actually looking for. I’m sure there’s enough blogs out there discussing the conversation, but, yeah, we might have to get back to you on that one, Jason.

Mitch Schwartz (Guest): What I could say is there’s this guy I was following, Michael Simmons, Blockbuster something. He’s on Substack, and he really focuses on the core skills of using AI. So it’s not so much about the platforms, but it’s the timeless things. And at the end of the day, I like the surfing analogy because as bigger and bigger waves come, if we’ve built the muscle memory of the core skills of balancing, standing, riding the wave as it relates to AI, and we’ll be able to ride the bigger waves and it doesn’t matter as much which surfboard we’re using.

So that’s the approach that I really advocate at the top level.

Mohamed Hamad (Host): Yeah, no, it’s, getting used to and being able to fluently conversate with AI and getting that cadence, really makes a difference after a while. I think, for us internally, it took us a while to get a feel for how to talk to each AI and which one resonates with different ways to structure prompts and conversations with it.

But I think we’ve got the hang of it and we’re getting the results that we want consistently. So, just really, lends to how you conversate with it. We’ve got another question here from Ivan. Ivan, sorry, could you expand on what some of these skills you mentioned would be?

Mitch Schwartz (Guest): Mitch, super briefly, I can comment. I don’t know if I can do it super briefly, but let’s find out. So one of the things is getting a sense of when you’re just in AI whack-a-mole hell, getting a sense of when you need to back out, go further back in the conversation, getting a sense of what types of phrasings work and don’t.

Just really building that cause-effect feeling of what you’re doing and how your conversation is turning out. And so I think with that you just get better at interacting with the AI. You get better at understanding also when it’s probably telling you something that’s not true.

You just really get a better sense for the whole interaction. So I would say it’s really like it comes down to prompting, prompt crafting as I like to say, instead of prompt engineering or, or just interacting with it in, in the same way that you would build the skills of leadership of working with an intern and understanding when they might be about to go off totally on the wrong path or when they need more guidance or if they’ve understood you how to confirm that how to see what they’re thinking about doing.

So I think it’s really—it comes down to a lot of skills that map pretty nicely with human skills. And everything else is like some tech stuff that you probably don’t really need to know. You probably just need to know how to at the core interact with an AI. And that’s what’s going to matter, I think the most.

Mohamed Hamad (Host): Yeah, for sure. I see it being more around communication. Just like how you would need to communicate with an intern. Don’t make any assumptions that they’ve done or known something before. Be a little bit more verbose. Really articulate exactly what you want and it’s, and the outcome.

And it’s one of the reasons I don’t like the prompt engineering because engineering makes it sound like a foundational institution of knowledge that there’s that to work off. Really this is just communication. How do you tell someone or something what you want, how it should be done and why it’s—

Why you need it to be done that way. And that’s the key part there. If we don’t have any more questions, we’re coming in at the end of this. Just wanted to make sure, if anybody wants to connect with Mitch, there’s that little button right there at the bottom.

You can connect with him on LinkedIn and on his own website OpsMachine. If you want to reach out to him or find out a little bit more about what you, what he does, he does have a wonderful program called the Jumpstart.

And if you want to talk a little bit about the Jumpstart before we wrap it up, please go ahead, Mitch.

Mitch Schwartz (Guest): Yeah, absolutely. So, it’s pretty much what I’ve been talking about. But we help companies figure out how can AI help us. So we go through the very grounding, very fundamental parts of like what is your process? Where are you really struggling? Where are you spending energy? How can it help?

So it’s fundamentally business consulting and process design, but by somebody with technical skills to figure out where the AI fits and then help do that whole implementation. We run what amounts to about four hours, and at the end you walk away with a blueprint for how you can move forward with it.

And, essentially we’ll roadmap. Really?

Mohamed Hamad (Host): That’s awesome. We got one last question from Ashleigh, and this is actually a really good question. Is why the most important for AI?

Mitch Schwartz (Guest): I’m not sure I understand the question.

Mohamed Hamad (Host): I’m, guessing it’s. Is the why you’re doing something or you’re requesting something from it important for it in terms of context? Correct me if I’m wrong.

Mitch Schwartz (Guest): That makes sense. Well, I’ll start answering that one anyways. Until we get any correction, is. I don’t know is why. Like, I feel like all the questions are really important.

Mitch Schwartz (Guest): The context matters. So is why more important than, like, what it’s doing? Is why more important than how it needs to accomplish it? I, I think it’s critical. So I think if you look at your five, you know, W questions framework, that could be a great way to go about prompting the AI.

Mitch Schwartz (Guest): One of the reasons that I like to use voice dictation is because it’s getting all of those contexts, including the why. And I absolutely like to share that. I don’t want AI slop because nobody’s going to read it. Yeah, that makes a lot of sense, to add that in and, often it’ll come out when you’re using voice dictation as well.

That’s why I really advocate for using that instead of just typing most of the time.

Mohamed Hamad (Host): That’s a great idea, actually. I find, I talk different than I type. This might be a good way of including the context with the why when you’re actually speaking to it.

Mitch Schwartz (Guest): Quick, for anyone using Windows, this is not really known, but if you hit Windows and H, it will bring up a voice dictation thing and then you can just speak. You can even turn off the profanity filter now so the AI can really get what you’re saying and get that feeling on the, temperature of your communication.

Mohamed Hamad (Host): Is this specifically for Claude or just the Windows thing?

Mitch Schwartz (Guest): No, that’s Windows wide. You just click in any text box, hit, Windows H like I keep looking Windows H, and it’ll bring up a thing to. Yeah, it’s pretty awesome.

Mohamed Hamad (Host): That’s a great tip, except I’m on a Mac. All right. In saying that it was a pleasure, Mitch, it was great having you. Thank you everybody for joining this session. Jason, we’ll get back to you on your questions.

Mohamed Hamad (Host): In the meantime, we’ll see you guys next time!

Mitch Schwartz (Guest): All right! Thanks, everyone.

Mohamed Hamad (Host): Thank you.