How to use agentic AI as your decision-ready intelligence

I mean how do I you know I don't want to automate myself out of a role but I think it's about framing right it's about framing the idea that you don't have to spend le you can spend less time clicking download cleaning data in Excel like a gentic egg can do all of that for you allowing you to think a little bit more critically about the business context of the report or the the nuances of some type of analysis that you're doing for clients. So I look at that as a as an opportunity. Not everybody does though. And so I think taking people down that journey, showing them the business value for me as a leader in this company and and driving change with my team. It's about showing business value, showing case studies, and showing the fact that and encouraging them to like, hey, it's okay to test this and if we fail, that's fine. Let's keep going. Let's keep optimizing. But I'm already in an environment where at the agency level, at the leadership level, our CEO and everyone down like we've we're all in on AI. So, it's easier to do it that way when there's other work streams and other leaders supporting this kind of cultural change. Much harder to do when there's no clear direction in the company and everybody's kind of looking around saying, well, what do we do with this? And, you know, do we need to be worried about our jobs? Things like that. So, I think it really does depend on the type of environment that you work in. — Yeah. And Tom, I'd love to hear your thoughts because I think outcomes is so important in that scenario in the test and the process like we have to evolve as humans and how we work. We can increase our efficiency and production by 10x if you use AI, right? Um, so Tom, I'd love to hear some of your examples from there. Well, I'm gonna follow on to what Michael said and even back it up a little bit because what's really interesting is Michael, congrats to you. To me, what holds teams back and

companies back and many are through this, but many are not yet, is a confident top executive sponsor. So, within your team, you've got a vision. you see it. You're you you're all bought in and you're willing to disrupt your entire operation and change things because you see it. Now flip it to the company side and many companies are at different stages. If they don't have a confident executive sponsor that al first of all is comfortable and confident talking about AI and really knowing it and they're a little worried. We saw a lot of this with CEOs in big companies say a year ago. they're much more competent, stronger, and have a vision of where they're going to go. But that will that's dead in the tracks. So then you have individual team members doing things, you know, smartly, maybe the right way, maybe not a good way, a bunch of activity that can cause chaos because there's no confident leadership and willingness to disrupt. So that can be a big thing that holds back. Most companies are through that, but some are still not all the way there yet. Then another thing what follows is fear. You know, we had the old thing first. Oh, hallucinations. That's still out there, guys. Still happen, but it's not going to stop where we are and it's not even slowing for companies that are moving off. We had pilot failures. The MIT report that came out a while ago and talked about 90%. That was true because teams were jumping too fast and not knowing how to do it. So, where do you go now? is um really think about it as um having that AI and agentic expertise in your organization and sometimes you have to decide whether you build by rent it and develop it and quite frankly it's often a little bit of all three of those areas

and then now how do you do the build stage effectively starts with clean accurate data if you don't have your data right your foundation is off and you can make very bad decisions and basically you can take agentics and u dramatically scale a bad process which some companies have done and I can't say we haven't done it at times — so — that data is so important Tom — exactly — I always it's always so funny to me when someone's like oh your blue silk or your Yeti is not working and I go and I check the the topic and it's just the word Apple and I'm like and they're trying to understand the brand Apple and they're wondering why the AI is confused. I'm like it's not a magic wand. You got to have clean data in there before you can analyze it. But I think what important elements that you both touched on here is you need to have leadership buyin in processes and a framework of how it's going to use based on outcomes or else you know it's kind of everyone running with their their um a chicken with their head cut off from that perspective. So seems like you've learned both learned that and have the commitment on the leadership level. any um advice on you know how to convince your leadership team or leaders to convince their teams that they won't be replaced. — Um I don't know I don't have any examples of the latter but the first is you know I mean this is still so new there's not a ton of case studies yet or use cases I found the media um you know there are some forums and Reddit forums where people are talking about this and so There's really again I think there's but there's a lot of hype right about aentic AI um just as there is with generative AI three years ago I mean it's now it's kind of this continued cycle and I think just learning right learning what it can and cannot do and there's nothing stopping anybody from learning uh I've learned how to use Photoshop years ago and Dreamweaver I used to build websites in like early 2000s from just doing and learning now you have things like YouTube and there's so much content on YouTube and tutorials on how to build, you know, using Zapier and other form and that's more AI automation, but there's other AI agent, you know, videos out there that can help you learn and there's free tools to use. So, I would just say learn it and then start thinking about how you can apply it to your if you're a small business owner or if you're on a small marketing team or any team, testing it, showing business value and then showing your management on, you know, the fact that it could save you this much money on investment or this much time. Um and that's how I would start out if you just again because there's not a lot of like you know published business use cases of companies who are actively sharing what they've been doing in aentic AI and how it's driving profit or how it's cutting costs things like that I'm sure that will come but I think it's still too early on there's not a ton you know for each industry that you can use to convince executives to move forward with this — you know especially from our healthc care and uh nonprofits and uh financial services. There's a lot of concerns with AI and data leaks uh using AI um especially for HIPPA compliancy and uh patient protection. Can you talk to me about maybe what you've experienced with your clients or internally um with using AI for this reason? — Oh my goodness. I'll jump into that a little bit first. I'm going to probably cover it at more of a little bit of a higher level versus the detail, but hopefully it's helpful. I mean, we cut across all of those industries with um very important secure confidential data like banking and financial services and healthcare and so forth. So, that's a big starting point with our um client base and brands that we support around the world. And it should be also

from a marketing and a social media marketing perspective is okay what data has to be completely secure and privatized. What can be anonymized? What should never go out? And what we've done with a lot of companies for example is we use small language models versus large language models within their internal architecture that we set up for them. so that you have that uh enhanced security fence around that very um important confidential um intellectual property and data. Um, and I think that's the best you can do, especially looking at what data source that um, you're looking at, right? Is this private data or public data? I think that's always critical. And so if it's private, you have to go through your company system is if it's public and you're using Hootsy Tooker, we are fully HIPACO compliant. Um, we have all the right um, compliance um, so etc. So, it's really just making sure the tools you use are using the proper protocols for your industry, I think, is a simple process. So, ask those questions, especially in your healthcare and financial service. They're some of our largest customer base. So, um we we're

we know how to work with them from that perspective. — Excellent point. When we first started testing Ali GPT and using it to some degree, it's again one of the agents in our hygienic framework. Now, we've got Yeti. That was the first question I'm going to wise is IP secure or is my information and our learnings training my competitors models? That's an important question to ask. — I think a lot of questions are coming up around applying agent to data processing and how does that actually apply to content strategy overall? Um from that perspective does Michael Tom do you have any thoughts on that one? I can give you I started testing that there was a white paper that was released last week or maybe it was a couple weeks ago from MIT and it was about recursive language models. — Yes. — And um it was it's fascinating and I was playing around with it and I had I was I uploaded this report. You can Google like recursive language models file type PDF and you'll get the actual report. It's like six or seven pages and it's pretty technical and it's long. That's why I uploaded it to Chad CBT and said help me figure this out. And what I got from it is um kind of a framework for telling the AI what it is you want, what is the outcome you want. So rather than a prompt, it um you tell them what problem you want to solve. And so I had uploaded a data set of a CSV file of data and I said this is data and it was for some new business pitch. Um, and it was data, it was media

coverage, it was social media conversations, engagement, all this great. And I uploaded, I didn't prompt it. I told him what I wanted the outcome to be. Um, and 45 minutes later, I had a report um, in PowerPoint presentation through chat GPT. Um, and it was not bad. Like I looked at it and I didn't think it was going to work. It was getting late. I went to sleep. The next morning I checked and I downloaded a PowerPoint file and it was in the right colors. It was it had insight driven headlines. Now, it wasn't perfect. Obviously, it wasn't even close to being client ready. Um, but it processed the data. Um, it wrote the insights and it took 45 minutes to do so. Okay? And it happened while I was sleeping. Um, now the next step for me is to ensure that the data processing was correct was go through the normal what we normally do, right? We export the data. We'll do pivot tables. We'll do VLOOKUPs. We'll and we'll create charts and things like that. What I'm going to do is actually have one of my analysts process it and then compare. So I want to see and make sure that what the analyst comes back with is what is accurate or not, right? And so that's so data processing and and whether you're cleaning data, you're tagging data, like that can be done with the Gentic AI very easily. Um but I'm still at that point where I don't fully trust the output yet. I still want to verify, right? I trust but verify. Um and we'll see. I'm happy to report back when I have uh we compare the two data sets because I'm very curious about it. But on the surface, it is like it's scary good. — It's so cool that it does that. Now, I've done a couple tests there. I even asked it to combine different uh Excels like I had 10 different Excelss and I needed it and it created and merged one and then gave me insights from it. I was like, — obviously validate and check, but like even just the combining of those sheets is a win. Tom, do you have any good stories to share on that part? — Well, I do and you gave me a good thought. Mike, I'm glad you went first because you clearly are more uh you have more depth of understanding of data than I do. But let me tell you a story that um relates similarly to what we're trying to do with our business intelligence architecture. So, we're pulling data from our um social media platform that cuts across all of our 195 channels. A lot of data potentially, right? The thing was, okay, how do you pull that data and get it cleansed, um, get it organized and feed it into our BI architecture, what we happen to have an agent in our ex, which we call our intelligent experience product suite around smart data for our internal teams to use. So my social media analyst said, "Well, great. Let me see if can take the data pull from our social media platform pulling data across all our 195 channels. You take that download which is a little messy and clean it up and organize it and do all that data prep and then feed it in to our um business intelligence architecture um which happens to be Domo where then we can create usable dashboards that show you know uh trends performance monthto month etc split anyway by any segment for all of our audience and it was extremely helpful. So again, one other agent in our aenic operating frame. — Amazing.