Will AI Supercharge Our Output or Sink Our Standards?

26:05How do you place ads on Facebook and Twitter and so on, and how do you optimize campaigns? Well, now we have enough data, to your point. So it's highly likely that an AI model can make at least as good predictions and assumptions as people can do. So where does that leave us? Where do our human traits fit in? And I would say a lot of times there, it's then about the human connection. It's the relationship with the brand, the relationship with the person on the other side. Understanding what does our customer, our consumer actually want and need. and maybe going through these different options that AI might give you.

26:45That experiential knowledge that, like, I know what it feels like to buy a new gadget and be excited about it. Yes. That's a thing that's really going to be hard to capture at the data level, I would think.

26:54Corey Noles:Or I know the feeling that makes me, you know, prefer Mac, you know, for example, over PC. I know how Mac makes me feel, and I know how Mac makes other people feel, so that's what I'm trying to capture that. And how it feels to watch an LLM do something that I do really well, better than me on occasion. But at what point if we're relying on AI, for example, in Meta, like Meta's AI is now making the ads for us, at what point does that degrade brand trust, to your point about brand? How do you find that line? I think it will erode trust when it goes sideways. The second there is a fully automated process that goes wrong one time.

27:40Even if it's one out of a hundred, one out of a million, it's the one time that counts. So how do you mitigate for that one time that it might not go right? So I think a lot of times we need to have humans in the loop. What was it, last year, two years ago, Apple released a short commercial, I think it was about the iPad, that created a big controversy.

28:05Corey Noles:Oh, the Squish? Yeah. Yeah. All I remember was the beautiful guitar that got smushed. So, right. And I'm sure there were people that were reviewing this. But again, there too, we need to pay attention to how is this being received by different people, different parts of the population, different parts of our consumers, different regions and so on. Nothing there changes. But something you mentioned about what does it feel. So I teach management information systems to undergrads at a local university. And in one of the modules, we talk about augmented reality. And by the way, one company that I think does that pretty, pretty well is IKEA.

28:49So everybody has probably been to an IKEA store or has an idea of what IKEA as a furniture store is. And I show my students this example in the IKEA app. look, here's a chair and you can switch it from gray to red to yellow and you can turn and you make it bigger and smaller. Now, would you buy this chair just based on the AR experience? And most of the students say, no, I wouldn't. I say, well, why wouldn't you? It's tech. It's cool. You can see it in your space. You can check out right there. You can have it delivered to your door. Like, yeah, I want to sit in the chair. I want to see how comfortable it is.

29:22Is it hard? Is it soft? What does it feel like? So to me, these experiences, what does it feel like AI cannot replace, right? Those are the decisions we want to make because at the end of the day, AI is not going to sit in the chair. I'm going to sit there. So it needs to feel right for me. Yes, technology can help if you're out to like red or yellow or gray or some other color. But, you know, that's also, by the way, where I think the online, offline retail still has a leg up if you can do both really, really well. I think so too. What it can do is get me to go to the store and sit in the chair.

29:57Corey Noles:But there is a value to seeing it, but it's one of many touch points. It's not the sole touch point, I think, is the kicker there. Yeah. And by the way, I think that's why car dealerships still offer test drives. Yeah. Yeah. So, again, physical, in-person experience, I think that's at least for now where we see a lot of the value. Again, AI can help us make informed decisions, give us more context. But at some point, it's the human-to-human or human-to-physical object interaction that still wins. So you teach about agentic AI and systems that act for us. Does adding autonomy boost the upside of AI or does it multiply the risk?

30:49Both. Both. I thought that might be coming. Checks out. Yes. Our answer is yes. So, you know, cheaper or prices is only one dimension. And yes, in business and towards our stakeholders, we oftentimes look at how can we reduce cost? And we also look at how can we at least maintain the same level of quality or even improve that? How can we maintain or improve customer satisfaction? How can we sell more? How can we do that faster? So price is only one. And a lot of times I like to think about this as the ion triangle of project management. Time, cost, quality. It's very, very hard to optimize all three.

31:38I would say it's impossible. So yes, price is certainly one part. But how do you ensure that the output that is generated is relevant, is of high quality, is something that's usable, is something that you would confidently give to your end user or to your end customer. I think there we need to spend a lot of time and diligence, and I'm actually seeing large enterprises go in that direction and think about the governance. How do we make sure that agents, for example, only have access to the systems and data that they need to have access to? So we don't expose data that is not intended for this person maybe in a call center or this customer to see.

32:19So quality, security, super, super important. It multiplies the risks if we're not doing that right. Otherwise, then it's just bad decisions faster. Which is not good. I wonder if some type of data segregation or on-demand access that isn't constant could play a role in that in the future because it's a really interesting line you have to balance there between what we want it to have and what it has in order to be able to access this information because usually that access extends further. So I've recently had some really, really insightful conversations with security experts, cybersecurity experts, and I recommend anybody to check out the OWASP top 10 for large language models.

33:14I think they're working something for agentic AI as well. OWASP is an independent organization. And, you know, seeing what some of those recommendations are that security professionals make and give you, it's very, very insightful because you also learn what are the things that could go wrong. So how can we mitigate for those from the get-go as we develop systems, as we architect new applications? how do you make sure that, for example, principles like zero trust are embedded from the very beginning? We don't trust any piece of the architecture more than we need to or at all. So same with limiting access to data on a need-to-know basis.

33:54What do you really need as an agent to do that task? And some of those core principles that the domain has developed over the last 20, 30 years of the internet cloud and so on, And now with AI and agentic AI, to me, that's a very important starting point. Because at the end of the day, it's not just about price and getting a good result to the customer. It's also making sure from IT, from a governance, from a security point of view, that we're actually safe and sound and secure in the way that we run this. Even a legal perspective, if you're in, say, healthcare or education, maybe, you know, where certain data is protected.

34:35Absolutely. So to me, it's super encouraging to see that there's so much activity in that space and people are figuring out what other things that we need to know and what guidance can we give to others.

34:48Corey Noles:right i guess like on that same subject so the one thing that satya nadella talks a lot about is he'll say um you know the other thing we need to figure out is the the i forget the exact term he uses but essentially governance of it so like right now if you or i were to create an autonomous system that goes out in the world the ai has no liability i have the liability you have the liability so figuring out like basically like who who is liable for what an agent does is a big part of this. And I think something Microsoft, as a huge company, thinks about a lot. Yeah, I think that's super important as we delegate more agency to software.

35:26Today, if you delegate agency, you might delegate it to your attorney or to your tax advisor to prepare this for you and file it on your behalf. So it's pretty clear, on one hand, what those guardrails are, it's for this specific consultation or it's for this specific tax return or tax question, and also the power of attorney, if you will, that you grant to that other person, that agency that you give that other person is very well defined too. Yes, they may file your tax return, but they cannot access your HIPAA-protected health care information or they cannot buy a car on your behalf. So in many respects, those principles, I think, need to apply here as well.

36:13But then, to me, any system has an owner. So who is that owner of this system, of these agents, of these swarms of agents at some point, if you will? And to me, that's where the liability comes in. Yes, you're delegating agency to software now, but in my view, you're still responsible for those results because you are setting this piece of software free in the world and are unleashing it. I'd say it's probably the same principle that applies if you own a trucking company and you own a truck and you hire a driver and send them out into the world in that and a problem happens. That's your problem for not taking good care of your truck, for not doing essential maintenance for not you know keeping it licensed and inspected and insured and making sure which which probably just came up with four new industries that'll find a way to circle in here i'm sure uh your ai insurance yeah that's funny um i guess like so like this this the when

37:19Corey Noles:prices drop on ai it increases the scale with which we can deploy it so that's where sort of like obviously liability scales with with when you're sending out a ton of agents on your behalf but shifting it a little bit more back to, I guess, like the human here. So what human skills in a world where you can scale jump in value? Like where should companies invest in upskilling as opposed to, you know, offloading? Like if we're saying that's the right way to go about things. So just the other day, I had a conversation with an AI leader at a nonprofit and I asked them the same question. And the answer was super insightful because he said, look, yes, we can use AI to help us figure out who are the optimal donors, who is likely going to make a gift and to whom and what amount.

38:11But at the end of the day, it must not feel transactional, right? The second it feels transactional, you're not as valued as a person. You don't feel as valued as a person. And you're very likely not going to make another donation to the same foundation or through the same organization again. And to me, that was very insightful because in business as well, we have the same situation. There's been this old or longstanding saying, sales is people doing business with people. And I think that, again, holds true. Yes, there are ways where we can augment or maybe partially automate tasks like business development, initial outreach, cold outreach, sending personalized messages, and so on.

38:57but for especially large transactions complex transactions complex sales cycles we want and we need to have a person there to build that trust with the customer here's what we can do for you we're also here to listen what are the pain points you have where are things not going so well if you need somebody to escalate that to I'm not giving you a 1-800 number I'm giving you my cell number and I will help you fix this or get this fixed for you so the personal relationship to me is something that we must not automate and then we must not automate away. I agree. So the other part to me is that everyone will eventually become a leader, whether you lead people and you have responsibility there or you lead systems or systems of agents, if you will.

39:46And having gone through several iterations of leadership training in my corporate career with increasing responsibility, it starts with how do you lead a team? How do you manage a team? How do you manage teams and teams of teams? Yeah. And what was constant in all these trainings was how do you also communicate and how do you delegate? So as a leader, it's super important to delegate. You need to convey what is the goal that I want you to achieve? What information is available? What else do you need to acquire? Who should you work with? What does good look like? And what's the time frame in which I would like you to complete this?

40:25Yeah. Now, as individual contributors, some of those things might be intuitive. Some of those things might not be as structured. So we might not think of all these five things. And if we don't, then probably the results are not as good as they could be or not what we expect. We see this in very simple terms now coming back to AI. When you submit a prompt, write a, again, job description for a podcast host. Yeah, sure. Right. It can do that perfectly well. reads like a job description, doesn't say anything about our show here, doesn't say anything about your previous experience, you know, things like that.

41:06So that's where, again, being able to delegate. Now also, as we increase the level of autonomy in these systems, gets super, super important. And communication, being able to articulate it. So interpersonal relationships, communication, and being able to delegate. To me, those are the three key skills we need to teach more so that humans can work with the software in the optimal way. I think the only thing I would add into that list is that I really think the ability to critically think is going to be a little extra important compared to where we have been. And largely because I think that's an element where there will be some struggles and the ability to not just use AI, but to understand how it fits into the puzzle and be able to really problem-solve your way through some of what's around the corner.

42:04That's great stuff, though. I think it's already tough, or it was already tough before generative AI, you could argue. It sure was. It's only exacerbating that, I would say, They were highlighting the need for it even more. And it's tough, right? If you delegate too much to AI of research, synthesis of information, critical thinking, and you just accept the results as gospel, it gets really, really hard to build and maintain that critical thinking muscle in a sense. so you know there have been studies around this topic of over-reliance on AI recently there were some coming out what was it called your brain on chat GPT or something yeah we covered that right so that's a good one to see and what was it a couple a couple days ago I think I read something about this this past cohort or year of students graduating you know how much have students actually learned and retained over the last two years since the emergence of generative AI?

43:16Or are they just getting a degree and they're actually not as knowledgeable as you would want them to be or need them to be in a business? And, you know, certainly what applies to early in career individuals applies all the way up the chain too, as that risk increases. So long story short to your point, right? Critical thinking is really, really important as well. And it's a matter of how do you actually train this now and how can you train for critical thinking and evaluate? Is something good, bad, or just okay? Yeah, that's a good question. And I think it just leads us well into governance too, because, you know, just like in other ways, governance costs money.

44:00What would you say is the minimum viable set of guardrails that any organization needs right now so savings aren't wiped out by scandals from AI use across the way? So I usually like to think about it in three dimensions. One is, what's a basic level, a foundational level of awareness of AI, generative AI, agentic AI, what have you? How does this work? The most important things to me are the technology is great, very powerful and mighty, but it's not infallible. We've seen this time and time again. Also, there's bias in the data that these models are trained with that trickles down to the results, to the outputs that you generate.

44:48So we need to check for those biases as well. And it's not just gender or ethnic biases. It can also be about professions and many other assumptions, right? It's probably as biased as we as a society and even as individual people are with the worldviews and the data that encodes our worldviews. So foundational understanding. Second one to me is the security and access that we already talked about. What do these systems really need to have access to and what are they allowed to do with the information? Is it just processing? Are they wrangling the data and writing it back? Are they making decisions?

45:31And then the third one is then on a technical level, right? We've seen this with prompt engineering in guardrails, for example. You're only allowed to do this. Don't do that. Or if you receive a question to which you don't have the answer, be honest. Say that that's not part of your scope or you cannot answer this or point to a different resource. So foundational, people enable training type things. Two, security. And then number three, technical mitigations as you're using these tools. That's a good call out. Really good call out.

46:07Corey Noles:Yeah, that becomes a bigger deal with like a Gentic AI where you're letting the AI make the decisions for you as part of the process and incorporating the tool use. So when can you use tools? When can you not use tools? What tools can you use when? that's all part of that equation. And you know, I'm still somewhat on the fence there if natural language is really the best way to give instructions. So I would say for a good reason, right? We've had programming languages for decades. If this happens, then do that. And it's very clear and very explicit. If your program runs into an error, you can trace it.

46:48You can say, okay, you know, this variable has this value, hence it doesn't move forward or it throws an exception. Now, if we move to agents and agents give other agents instructions based on natural language. I mean, we've all played whisper down the lane. Yeah. You start with one thing, you end up with something totally different. I think that is a real and not to be underestimated risk. Yeah. As more and more agents converse with each other, check each other's output, give feedback and so on. And coming back to security, right? Even a malintent of one agent instructing another one to do something that they're not supposed to, right?

47:29Or go off the rails or, well, you know, it's perfectly fine. This is an exception. This is just one time, right? So, you know, just thinking about the social engineering of people applied to technology, I think it's something that we just need to have on the radar as well. and just think about as we build these architectures and systems. That's a good call.

47:55Corey Noles:So let's just lightning round through a couple of these last questions. So let's say like falling model costs, right? We've established that DeepSeek is free open source AI. Obviously, it's not free to host it yourself, but you can technically download it and run it at a certain cost, depending on where you host it. a lot of the cloud providers are lowering costs. It's this sort of shift to the bottom. How do we, in terms of price, how do we who are using it stay competitive or become more competitive without also bringing all of our costs down because of the cost of intelligence going down? So intelligence is just one factor of your equation, I would say.

48:40Whatever service or product you provide, there's a whole lot more to it, I would hope, than a wrapper around an LLM. If that's your business model, you might be in trouble with the next upgrade you want to spend or make. We've seen this with PDF wrappers and chat with your PDF type things. And all of a sudden, it's a standard functionality and one of the tools we use on a regular basis. So please don't build an LLM wrapper. You need to have some moat of your own and some secret sauce and expertise. And I think it's that secret sauce and expertise that makes your business unique. And that's what customers come to you and pay you for.

49:16Whether you use an LLM or an LLM by OpenAI or Anthropic or Google or what have you, it doesn't really matter so much. So what is the value that you create? And to me, that opens up a whole new opportunity around what business model might you want to create in addition to what you have or to replace what you have to evolve. And there we're seeing the first companies move into something called outcome-based models. So companies like Intercom that do customer service or I think Sendesk was another one in that space, they are moving to outcome-based models. So instead of saying you pay per user in your customer contact center or you pay per ticket, you pay per successful resolution.

50:02Super awesome, right? All of a sudden you have a direct metric that your customer is looking at anyways. what is our success rate and maybe even what is our customer satisfaction rate. And you only pay per successful resolution. Now, from a vendor point of view, that creates a challenge too, because what is considered a successful resolution and how do you get that agreement with your economic buyer that this was really successful? Because the challenge you might run into is that your buyer, that your customer says, well, this one wasn't quite right, so we're not going to pay for this one. And by the way, there were 10 others like that or 100 others like that in the past month or in the last quarter.

50:41So no, no, no, no, no. We need to take them out. All of a sudden, yes, you have delivered value and you're incentivized to improve your product and your model so you can deliver outcome-based pricing. But again, the human element that comes into play. So outcome-based pricing to me is a big opportunity now that we can delegate more to AI, that there's more automation, that there's more autonomy and overall more intelligence in the process.

51:08Corey Noles:I see outcome-based pricing is kind of like the switch from the deterministic code to probabilistic language. It's sort of the same. It's like deterministic. You pay for 200 API calls or whatever it is. And now it's more in this nebulous range, which is better but harder to track. Yeah. And from a vendor point of view, right, if this idea holds true that there might be fewer people interacting with the system because you can automate more or not everybody needs to have a license or a user for a specific tool, then how do you monetize it? How do you, at a minimum, make the same revenue or even better increase it?

51:50And again, that's where I think something like outcome-based can be a fair model and something that can propel software vendors into the next era. And it basically shifts a lot of the risk back to the vendor instead of the consumer. Absolutely, right? Ideally, that helps reduce the amount of shelfware products that somebody has bought, but they're not using it or they're not using them yet or not using as much as they've paid for. So you only pay for what you use. And to your point, from a vendor perspective, you need to make sure that your products work really, really well and ideally outperform the competition so that your customers stay with you or, first of all, come to you.

52:31Yeah, because now you're shouldering the burden and cost of a dud. Exactly. We've been there before in the industry and it used to be called business process outsourcing. So in a similar model, I envision that we're going to see this as well.

52:52so in your nine-step ai readiness framework you stress clear kpis what two or three productivity metrics in your opinion cut through vanity stats to show some real gains so we're back at productivity um we made it back to productivity it all comes back to product oh So I would say don't just look at productivity. Let me answer it this way. I would say look at what are your macro level KPIs. Where do we want to be three years from now? How do we measure that? Is it more revenue? Is it lower cost? Is it increased customer satisfaction? And then derived from that, what is the KPI tree? What are the supporting KPIs underneath that would help us create more revenue, reduce costs, get higher customer satisfaction?

53:42But the important thing is to tie it to a KPI that you can measure before and after so you can show the effect. Not just happier employees, not just saving a minute on an email or two minutes or something. That's nice. But what I hear from a lot of CEOs is I cannot really measure the productivity gain or it doesn't translate into something else. It's not hard dollars. It's not, again, increased sales, reduced cost. So look for those use cases, those scenarios where you can tie it to at least one of those two. Those seem to be the more promising ones to measure the impact. And again, build momentum and sustain momentum as you go along.

54:23As opposed to the more qualitative, squishy numbers that are... Yeah. Yeah. Look, we've had these aspirations of fewer clicks, happier employees for at least the last 10 years in the machine learning AI, Gen AI domain. And they're good. Yes, we want happier employees. Yes, we want fewer clicks and increased efficiency. But unless you can really measure it, it gets really, really hard to argue why this is the right solution, the right product, the right feature, the right use case to pursue and to move forward with. So hard dollars is what usually wins. Makes sense to me.

55:02Corey Noles:I have one last question before we get to the end here. And if anything else Corey wants to ask, But I guess, do you have any like specific tools that you recommend, if any come to mind? If not, that's totally fine. But when I hear about a lot of this stuff, I always try to think about, OK, so what is the actual like we talked about a little bit about your workflow? What is the actual process or tools that companies can use to do some of this? So it's a very broad question. It depends on where you are in your business, what your function is, what business function you're in to begin with. But I think generally you can look at simple things like note taking, whether it's built into your productivity suite that you use on a regular basis or some alternatives that can help, again, summarize what are the key things that we talked about.

55:54Can you send me the action items or send them out to all the participants so we have something to work on until the next time we come back? If you're in sales, there are lots of products augmenting go-to-market. So it could be the contact enrichment, lead enrichment with the help of AI, doing prospecting and outreach with AI agents. There are a dozen startups in that space as well. Some of the earlier ones were around marketing and copy, blog posts, websites, white papers, reports, these kind of things. it really depends where you are. If you are a little tech savvy or have an affinity for technology, workflow tools like Zapier, like N8N, make.com, to me, they're really, really great because it's visual, it's low code, no code.

56:46You can drag and drop items. You can connect them. You can connect your LLM. You can build a workflow and still have some AI there, but you don't have to be a trained statistician or a mathematician. You don't have to be a software engineer, but you can already experiment with it and feel it. And you can see what is possible when you bring AI to that workflow. And again, maybe it helps you cut down your workflow from four hours to 31 minutes and 19 seconds as well. I would hope so, right? But the cool thing that this enables you to do then as well is think about what am I doing? What am I doing repetitively?

57:25Do I really need all these steps? and what can I cut out and of that what remains, what is really core, what can I automate and how can I automate that? And whether that is your workflow going from a podcast to newsletter to social posts or something in your finance department, HR department, procurement department, wherever you are, the same basic principle applies as well. Yeah, tackle the thing you like the least. What's the one thing that just drains your soul a little bit every time you have to waste three hours doing it or it feels wasteful you know exactly so i i have a fun question i want to ask real quick um suppose it's 2040 some you know some teenager or someone wanders across this episode what do you hope they'll say that we got right and what do you think they're most likely you to laugh about the way we talk about AI today?

58:25So, I would say to me, it's similar probably to how we think about electricity these days. It's there, it works, but it's enabled innovation vastly and far beyond what people envisioned 100 years ago. What could be possible, whether it's powering factories or just having light in your house? look we have computers that were enabled by electricity that are powered by electricity we have so much more in innovation and i think at a minimum we'll see that too so we're underestimating today what will be possible and what we'll see over the next 15-20 years similar to electricity i also think everyone should have access to ai i agree and hopefully that is the case over the next 15, 20 years, that it will premier society even more and will be accessible by pretty much everyone.

59:23But then I would say also, similar to electricity, from a very early age, we learn what not to do with electricity, right? What not to put into your power socket. No forks. So I would hope that over the next 15, 20 years, we're doing a good job at educating our youth as well on what do you do with electricity and what do you not do with electricity. Very, very early, right? If you learn this when you're 30, you've probably been shocked a couple of times. So teach it early and make sure it sticks. um what what did we get wrong um or what are we getting wrong now uh that uh looking back in in 2040 people will laugh at such a hard question um to me that that is underestimating right maybe people in in 2040 will will laugh about the fact that we use this just to write emails what do you mean people are writing emails right what's it all yeah that's such a great call out so so hopefully on a more social and beneficial side, we will use the technology to solve some of the biggest challenges that we see in the world today.

1:00:41On a more personal side, maybe we will change the way that entertainment works, hyper-personalized experiences, Netflix in your AR, VR glasses, split-second changes of scripts and scenes and whatever. those things are already there I don't think it'll take 15 years to develop, probably 15 years for it to become a huge mass market and time to get good yeah, exactly and beyond 3 second and 5 second snippets but the fact that we're already there I think is so exciting so those are some of the things that I would see again, we're underestimating what it can really do and what it will do. But again, we need to teach the basics of the do's and don'ts so we can prepare for that future that, you know, helps us all prosper.

1:01:35I always wonder how long it will be before someone asks the question, did you really let humans read your x-rays? Like, that's a subject that keeps coming up and I keep seeing in studies. Like, you know, I think that'll be an interesting topic going forward in medicine, in finance, in, I guess, in just about everything, probably. And what do you mean you had meetings that could have been an email or that could have been an avatar conversation? You got together for an hour, six people, and talked about that? Really? What a waste of your time. Exactly. Five minutes and our avatars hashed it out.

1:02:21Now, you know, is that the future we want to create and we want to live in? That's a different story. But, you know, looking at how things are working in recruiting at the moment, right? Resume ranking to help recruiters sift through hundreds of resumes is great. On the other side, if you're an applicant, if you're filtered out by an AI system without any feedback, without an opportunity to get transparency or appeal or even stand out, it's really, really hard. It is. Right? And then using AI to tweak your resume to fit the job description so your chances increase to stand out by another AI that filters you out.

1:03:04Corey Noles:It forces everything back into a closed network system where essentially you can only trust people who you personally know and vouch for. And it makes it actually more of a closed society. Right. So quite the opposite from creating a more equitable and fairer process and fairer opportunities. On one hand, it's relationship based. Or again, you need to invite candidates to come to the office to see if they even exist or if they're an avatar and to test their skills. And what do you test for? Right. So take the AI away, draw on a on a whiteboard. How would you do that? Right. Walk me through this situation.

1:03:41How have you done that type things, behavioral interviews? Yeah. Again, coming back to the human element, I think the human element is going to be a much, much bigger factor and a much more important factor than it already is today.

1:03:55Corey Noles:hopefully in an even better way go ahead what what if it got to the point where an agent who observes you all the time is the one applying to jobs for you and vouching for you and everyone has like the same kind of system so they can all trust this system's review of you and you know the agent says hey here's my here's my candidate it's sort of like your actual agent like in hollywood right like vouching for you and saying this is this guy's skills this is this person's skill they're really good at this so you should consider them for this job i've seen them i've been watching them over the past 10 years so i know how good they are yeah it might also be a sequel to 1984 call it 2025 yeah 2025 oh my gosh what a neat call there's there's a good book in there somewhere there's there's definitely a good movie yes yes well andreas would you mind uh tell Tell us a little bit about your show.

1:04:54I want to talk about your show. I want to talk about your book because I've really enjoyed what we've hit on in both. Awesome. Sure. So, look, about three years ago, I started a live stream and podcast called What's the Buzz? AI in Business. So far, I've interviewed more than 100 AI leaders on how they have brought AI into their organization and turned Technology Hive into business outcomes. And to me, those are some of the most exciting and most insightful conversations because they're raw, they're unfiltered. There were people doing this work on a daily basis. And that was one of the big reasons for me to start the show, to hear from others in the industry, how are they doing it?

1:05:31How are they succeeding? Where have they failed? What would they do differently? And then also, quite frankly, if I got to do it again, how would I do it differently? So What's the Buzz is a biweekly show that you'll find on LinkedIn, on YouTube, as a live stream, on any podcast platform. If you prefer audio. And I've summarized the first 60 plus interviews in the AI leadership handbook. I encourage you to check that out as well so you can learn how you can turn technology hype. We'll make sure we have a link below as well. Yes, that would be awesome. Yeah, you find me on LinkedIn. I post daily about how you can bring AI into your business.

1:06:11I've got several courses on LinkedIn learning that I already like about agentic AI, about mitigating your risks as you go through AI projects. And I'm just super excited about the opportunity getting to help many, many businesses and business leaders figure out what do we do with AI? What should our strategy be? What are the most important and promising things we should prioritize? And most of all, how can we bring our people along on this journey and also make sure that they're enabled and upskilled to thrive in that economy? I love that. And this is an area where I have, you know, Grant and I both work with AI here at our parent company, Technology Advice.

1:06:47and a number of the projects we get involved in, it's very common to come across a situation where it's like there is no information on the problem we're trying to solve. Like we're going to have to sit down and do this the old way. Like, you know, we've got to put our heads together and think critically and, you know, really kind of problem solve. And I'm enjoying that your podcast, your book, Others like you as well have kind of slowly started to surface here, whereas people are gaining experience and working through these hard problems that that they're sharing what they're learning because it's needed.

1:07:29You know, we're very fortunate to have a number of people who are very well versed and not afraid to get their hands dirty in this kind of stuff. But that's not the case at every business. You know, some places don't have that. Yeah. So, you know, if you're looking for a second opinion for some guidance, advice or even to inspire your team, please feel free to reach out to me on LinkedIn or go to intelligence-briefing.com. We'll have the link in the show notes as well. So you can connect with me and would be happy to have a conversation and help you where I can. awesome well as is kind of our little nerdy tradition here we have a kind of a round robin call question at the end that's usually a lot of fun so five years out do you expect we'll celebrate a productivity boom or regret a flood of low quality output that we have dumped onto the world.

1:08:26What should we be looking out for? So we might not even need to wait five years. Yeah. That's my answer, right? Again, look at your feed, look at the newsletters that you get, look at some of the AI-generated comments and summaries about financial statements and earnings reports and things, and you'll spot AI in many, many different places already. Most of the time, I would say it's good if it's from reputable sources. a lot of times it's pretty bland and generic. So we're seeing this AI slop. And again, by the way, quick plug, I'm working on two courses with LinkedIn Learning to help leaders give guidance to their team.

1:09:05How should we use AI? Yes, I want you to use it, but please use it responsibly and check the sources and check the information. And likewise, if you're an individual contributor, how should you use AI responsibly in business? So we don't create more AI slop. You know, already today, the internet looks nothing like it looked three years ago before ChatGPT. We're starting to see the first content generators there. And some of us were wondering, well, isn't that a downward spiral if AI creates content and learns on that content and creates more content and learns on that content? So we're seeing that creep into many, many areas of media already.

1:09:46So my recommendation would be look for authentic voices. Look for people who have the experience, who've done it before, who are sharing the experience and who you can reach out to. Real people with real experience because they can also read between the lines a lot better than AI can currently do that. Yeah. And nothing beats that. I agree. But I'm curious. What about you? You want to go grand? You want to go first or you want me to?

1:10:10Corey Noles:No. No, no. I'll go last. Okay. uh what i would say is that uh there's i think just like with everything else we do you know i always lean toward content because i've worked in publishing for so many years but in any field i mean let's go back to the 1970s when mass production really really kind of ramped up it was like like okay we figured out how to mass produce just about everything and and the quality of stuff went down and I'm going to mention guitars because that's a subject I know well and when you look at 70s era Gibsons there are lots of changes that were made specifically to cut costs a little bit here and make changes a little here and yes some of those are awful but not all of them you know so I think we're going to be looking at something a lot like that where yeah there's going to be some bad stuff out there but there's always been some garbage out there it's not an idea that's new uh what i hope we'll also see is a shift away from ai as a thing that needs to be caught or detected in favor of the idea that something is either good or it's bad and and and none of the rest really matters that the rest is just noise either either either this thing is good or it is not good, regardless of how many M dashes and semicolons it might happen to have in its toolbox.

1:11:42I, uh, I really feel like there's going to be a difference there, but yeah, I think we'll definitely look back and be like, Oh, I remember those days on some things like, uh, you know, we're going to look back at a lot of three-fingered early Dolly photos. We're going to look at, uh, you know, uh, early attempts at social media with AI that are going to be really cringy and bad. But truth is, they're probably all going to be cringy and bad. So we may not be able to notice the difference as time goes. But I think that's all of my take. What about you, Grant?

1:12:15Corey Noles:Oh, gosh. Well, so one data point I want to share. So Gartner released a report recently, you might have seen it, that was like 40 % of all agent tests that are happening right now will probably be canceled by 2027. And I think the thinking goes that because like right now it's very early prototype, very experimental stuff to a lot of the points that you brought up, Andreas, like there's reasons where like, you know, when there's one agent talking to another agent, a lot gets lost in translation and the failure rate spikes pretty quickly. So I think it's probably going to be more like a 10-year horizon for productivity to boom in the agent era.

1:12:57Corey Noles:I think we've already kind of seen a lot of the gains of the generative AI initial wave. But I think whether it's a boom or a flood of low quality output depends on the economic picture, like how many people have access to it, how many people are producing low quality at scale. and if companies are so productive that they're now doing layoffs and how that impacts the economy, whatever else is going on in the economy, that could impact how much you're producing too. Yeah, it might be that in 10 years we don't look back on weak blog posts and lousy social media. It might be that we're looking back on something entirely different.

1:13:42Who knows? And about the Gartner factoid, right? Six, seven years ago, Gartner said about 80 to 85 % of AI projects don't deliver the value they initially intend to create. So 40 % of failing is great. That means 60 % are successful, a lot better than 15 or 20 % of AI projects. That's a really good counterfactual. Yeah, it is. But we also see a lot of times this traditional hype cycle, right? So now we have vendors who have been adding AI, agentic AI into their products. It's new for them. Yes, they are marketing it. Yes, there is an opportunity. But it's up to the customer to introduce that and implement that in their business and figure out how does work change if all of a sudden I have this agent, this AI thing that I can delegate to that can do certain things on my behalf.

1:14:32A lot of times we see a lot of optimism around this technology, what it will be able to do. It doesn't matter if it's agentic or generative or machine learning or what have you. So it takes human experimentation and experience to see what actually works and what makes sense in our business. So the promises, the expectations we have, yes, they're huge. They're hugely inflated. But I think that's also good, right? Because we are venturing into this and we are looking to explore and to experiment and to figure out what actually makes sense. And that's a deeply human process that will take time and where we will also sort out what works and what doesn't.

1:15:11Are you on the 40 % or are you on the 60 % side? So to me, business as usual. Yeah. And, you know, I always joke about this, but in baseball, you know, if you get a hit three and a half times out of 10, you know, you go to the Hall of Fame. You make millions of dollars a year for failing 65 % of the time. And I think that's an interesting call out. Well, everyone, if you want to dive deeper, please check out Andreas' show, What's the Buzz? Pick up his book, The AI Leadership Handbook. Links are in today's notes and everywhere you get your podcasts. Please take a moment to like, subscribe, comment.

1:15:49We'd love to get engaged in the discussion with you and talk a little more about all of this and anything else that might be going on. So Andreas, Grant, thanks so much. It's been a lot of fun today and thanks for helping us separate Bargain Bin from Bust in the world of AI. Awesome. Thank you for having me. Yeah. So to all of our listeners, stay curious. We'll see you next week on The Neuron. Farewell for now, humans.

1:16:30We'll be right back.