Michael Greenberg - Tactical AI

And how it's impacting or will impact the future of work. Michael, welcome to the show. It's really great having you with us today here on the show. Anything to add before we dive in?

Michael Greenberg: Just the one qualification when we're talking about AI today. Anytime we say AI for all those sticklers out there, I'm really just going to be referring to large language models and the related technology in the current boom.

Always a stickler in the crowd.

Tim Butara: Yeah. An important clarification. Definitely.

Michael Greenberg: Yep.

Tim Butara: Okay. So with this taken into account, what exactly do we mean by tactical AI?

Michael Greenberg: Yeah. So I think there's a, there's a lot of talk of like, Oh, AI will replace a job. AI will do this. But what is AI actually doing today? What are we actually seeing people use AI for?

Where is it successfully being deployed? We've all heard the disaster stories of chatbot deployments to customer facing and public facing communication. What are the alternatives that people are using to those things to achieve similar results? And what are the little, little hacks? We call them magic tricks inside of the company, little one or two step things you can do with a chat GPT or a cloud, something like that.

That's just going to give you. Output that saves you hours.

Tim Butara: So maybe if we get a little bit more hands on, you also mentioned, you know, I, and I hope that you'll share a little bit more about some of your internal examples but what are some really important business areas or some particular use cases where tactical AI is especially valuable or useful?

Michael Greenberg: I think we are seeing the biggest gains in three areas right now. Sales. Marketing and operations. And the reason those three come up is because they have the most, like, directly repeatable tasks. And they have a lot of repeatable tasks that have just a little bit of fuzzy logic that made it impossible to automate with like a Zapier hard set of rules as we do a lot of our work right now, but an AI can say, is this email a complaint or is it an order and routed accordingly off the shelf?

You're not going to be able to figure that out from just a subject line or a single word in the body of the text in most cases.

Tim Butara: So in most industries and in most use cases, you know, we, we talk about AI being particularly useful in cases of, you know, repeatable tiresome tasks, I guess, and I guess that this is what's really tactical about it.

Michael Greenberg: Yeah, it's two, a single thing. So we use AI for email routing all the time. We use AI for meeting summarization and taking like a lot of agency clients have a long onboarding meeting with their own clients. And so new. New client comes on to the agency. Agency has like a two hour call, gets a ton of information in onboarding, and then has to create a giant client brief that describes everything.

So that way the whole team knows we can look at how they make audit, how they make client briefs, and then have a little fine tuned GPT4O mini. It'll cost like three or four bucks to fine tune it. And we can have it set up to be the perfect client brief writer based on the hundred best client briefs they've ever done.

That is at this point, low cost to implement, and we can get a long way without doing that sort of fine tuning in order to deploy it, you can't use just a standard call or note taker app. Because you're going to need to ask 30 or 40 questions across the transcript. And that can only be done with automation and then additional summarization and pulling in context from other data sources that you really need a little bit more of a custom set up on the backend to support a similar same architecture that we use internally would be a lot of our coaching clients or like community based masterminds.

They want to be able to record all of their meetings and then anonymize them and put them into a chat bot. So they have a knowledge base for their members. Super easy to do if you string together the 30 different steps of, Oh, are we recording every meeting? Are we transcribing every meeting? Are they stored in a way that it is actually possible for us to do that?

Do this, do we have the legal approval from everyone involved to use these recordings? There's 30 different pieces on the business side and the technical side that all need to line up in order to actually deploy something like that in a business. And I think most businesses, most organizations are missing that base layer of digital infrastructure that ties all of their data together to be able to deploy more complex AI, which is why we get stuck on, Oh, well, one of my favorite packs, we can take a transcript from a client interview and we can drop it into Claude and Claude's going to make us a workflow.

That's a one step prompt. That might save us two hours. Compared to that onboarding automation, onboarding, client brief creation that can replace an entire job.

Tim Butara: Wow. Okay. I'm sure that we'll talk a little bit more about this particular aspect a little bit later, but first I'm interested, you know, you just highlighted that a good, robust digital infrastructure is a prerequisite for doing all of this, right?

And it makes sense, right? Because, you know, how are you going to implement or incorporate advanced technologies into your digital strategy? If you haven't got the baseline cover, that sounds like a recipe for disaster for me. And I think that we just recently actually had an episode about digital transformation being a prerequisite for AI transformation.

And that makes sense, right? Based on also what you just said. So based on all this, do you have some kind of Playbook or something very hands on actionable to help people, you know, implement tactical AI. Maybe people that are listening right now aren't sure how exactly they should go about it.

Michael Greenberg: Yeah, so first you have to understand where you are in the process and to understand sort of how much can we implement, you have to look at your organization and that could be your department, that could be your team, that could be your whole company, just depending on what you're doing.

Where you sit, you have to ask, do all of our tools talk to each other? Is it easy for us to pass data in between them? And do we know which one of them is the primary source for that data? So do we know which one is the real source of truth? And that's on your data. That's on your software side. On the other side, you have to ask for your business processes.

Do we know who is supposed to be doing what, when, and do we know how they're supposed to do it? And do we have that documented? Because AI is the best intern you'll ever have. It's really dumb. It doesn't know how to do anything, but if you tell it exactly what to do, step by step, it can do so much. And so you need both all of your data in place.

So any sort of AI you want to deploy can actually reach everything. And you need to know the process that you carry out. And only at that point can somebody who understands both sides, then sit down and say, Oh, We can automate this step. AI can summarize this step. We can transfer the data here to here, and then we can have a little agent that checks every day and checks the news and sees if there's anything that we should do and report here.

And that is what we call sort of a level three organization, which means they have the infrastructure and process documentation to be able to support building automated and human in the loop AI solutions. In order to get to like a fully AI, nobody has to touch anything. You're going to, you're going to want to run with a human managing the process or a human, like actively incorporated in the process for at least a hundred reps, I think, if not a thousand.

Tim Butara: So wait, the goal of the human in the loop is to not have a human in the loop at some point.

Michael Greenberg: Exactly.

Tim Butara: But, but it, is that true for, for, you know, in general, or is it just true, you know, for particular use cases? Because I'm sure that there are some industry, some use cases where it will still be essential to have a human in the loop throughout the whole process.

Michael Greenberg: Oh yeah. And I think there's many cases where I can't think of a case where I would not want to have a human at least supervising, if not in the direct approval loop. But. It might have to be that, Oh, if you're doing like a content production workflow, very common for us. It might be that a human has to approve the outline of the content before the AI goes and drafts the first draft.

And so the human might act as the editor and the AI might act as like a junior writer, but then you're still going to need another human at the end of that process to clean it up and to make it not look like an AI wrote it in most cases. Or we have a client where we've deployed little AI that monitors camera feeds for them, and then flags potential violations or issues that the cameras pick up.

And then a human team is going to look at those flags and make the call.

Tim Butara: Yeah, because it's not just, you know, the importance and the relevance of talking about the human in the loop is not just with regards to innovation and advancement in AI, but also the last topic that we were conveniently saving for, for the end of the discussion.

And that is how AI is impacting work and the future of work. And, and I mean, obviously we need to talk about that, but obviously. If we don't want AI to, so to speak, take over our job, then a human in the loop will also be essential for, you know, the sustainability of this entire ecosystem from the human perspective.

Michael Greenberg: The way I envision, at least the future of work, the way I'm seeing roles evolve inside of our organization is one person is able to do a little bit more in the same amount of time. And then occasionally we will make some sort of breakthrough or finish a larger, more complex project that will allow us to replace an entire activity.

And so that ultimately means that everyone has a little more capacity at work. They have a little more confidence in their ability to get their work done and all of the work becomes a little more clear. It becomes more tracked as a result of more of it being automated. I think that's a good thing.

Tim Butara: Yeah.

And also you know, you see, you said that it might replace if we, if we hone it enough, it might replace particular activities, but it also means that, that there will be new activities based on this change that will then have to be taken care of probably by humans. So, you know, we've heard this often, but while it's true that yes, it will, AI will heavily impact the job market.

That means both, you know, the loss of some jobs and some roles, but also the creation of new ones, right?

Michael Greenberg: Absolutely. I mean, inside of like the podcast production space, it's one I know well, I've got a, I've produced a number of podcasts myself over the years, and it was actually my first taste of AI when the cost of transcription went from a dollar a minute down to 10 cents a minute.

This was probably Eight or 10 years ago now, now transcription is like practically free. Well, it is free. I'm watching a transcript show up right now on zoom live. That would have been very expensive years ago. That is a job, just transcription alone that used to employ tons of people. Most of those people, podcast realm, moved from podcast transcript to podcast show notes that bought them another five years.

Because then podcast show notes, AI, AI got better, got better at summarizing, better at voice to capture tone for different shows. Now, AI show notes are 90 percent of all show notes out there, maybe 95%, but the difference between an amazing set of show notes from a human and like your average set of AI show notes is still a very large gap.

And so I think I see this with designers too, right now, all of the B and B minus and lower like quality. People in the market who are getting by because somebody needed a human to repeatedly do this task who needed somebody to follow brand guidelines and make a thousand different variations of their logo on banners.

That person's out of the job, the person who just summarized podcasts all day and couldn't write great show notes and wasn't passionate about the topic. That person's out of the job, but the people who are truly great and who stand out above, like the average person, or right now, I think maybe a little below the average person, they are secure because they have real novel, creative work.

And so I think that's sort of what we're seeing is there's a lot of non creative creative roles in the world and those are getting wiped out right now.

Tim Butara: Yeah. I mean the, the creative work involves a lot of processes that aren't intrinsically creative. And this is where, where things like AI and, and the right the right use of AI can really be the biggest, can really have the biggest impact, I guess.

Absolutely.

Michael Greenberg: You know, all the open AI guys right now, they're saying, Oh, five years from now, we're going to have AGI. And then five years after that, we've got artificial superintelligence. I don't know what that means for us. I don't know if that's true. I do know that right now we are not even scraping the edge of what we can do with the existing technology, just by coding the thought patterns.

Of a specific roles. So one that I know actually the PR company that helped me get on your show uses because we built it is a little marketing analyst. And what it does is it reviews the. Responses and the outreach campaigns that they're sending to get guests placed places. And then it gives an automated report of these campaigns are going well.

This variation is winning super simple, super useful. Used to be somebody's, you know, first task every day.

Tim Butara: Yeah, that that's very cool. And as we said before, that doesn't mean that the person is necessarily not involved in this process anymore. It's just that their role, if they're driven enough, if they're committed enough, they're creative enough, their role will just likely change, but it won't go away.

Michael Greenberg: Yeah. The one at the biggest risk right now, there might be a little buyer alarm to the audience. It's project managers.

Tim Butara: Yeah,

Michael Greenberg: because if a come as companies move to more global remote work, they require a higher level of digital communication and the soft skills of project management become more important as AI removes.

Any of the like more hard capacity planning related stuff, all of that gets wiped away when all of your data is accurately set up in like a click up or a notion. They have their AI tools that layer right on top of it and you know, you're not writing summaries, you're not capturing standups or updates if you have a decent automated setup going.

But somebody does have to follow up with people. Somebody does have to like discuss scopes and negotiate timelines and do things like that. And that is a much more advanced task where I think there's a lot of people who have just sort of shuffled paper and they're getting squeezed out as all of this data becomes more public and organizable.

Tim Butara: Yep. Yeah. As we said, that definitely one of the roles and one of the types of roles, that's like most impacted for people who, who, you know, won't change with that. Absolutely. Michael, thank you so much for joining us today. Thank you for this great discussion and kind of multifaceted discussion going from, you know, the more, the more not super technical, but technical aspects of implementing AI in technical ways to the larger, broader connotations of, you know, the impact on the job market and the future of work.

Thank you so much. As I said, if the listeners would like to maybe learn more about you, connect with you, learn more from you, what's the best way to reach you and to, to get in touch with you.

Michael Greenberg: Yeah. So you can find me at Gentoftech, G E N T O F T E C H on all social media channels, and then you can find my company at the number three R D brain.co.

Tim Butara: Awesome. We'll include both in the show notes and thanks again, Michael. And to our listeners, that's all for this episode. Have a great day everyone, and stay safe.

Outro:
Thanks for tuning in. If you'd like to check out our other episodes, you can find all of them at agiledrop. com slash podcast, as well as on all the most popular podcasting platforms. Make sure to subscribe so you don't miss any new episodes, and don't forget to share the podcast with your friends and colleagues.