The Customer Success Playbook
Welcome to “The Customer Success Playbook,” a fresh podcast initiative spearheaded by Kevin Metzger and Roman Trebon. Immerse yourself with us in the dynamic realm of customer success, where we unravel the latest insights, inspirations, and wisdom from recognized leaders in the Customer Success domain.
Our journey began with a simple yet profound belief: that meaningful conversations can significantly impact our professional trajectory. With this ethos, we’ve embarked on a mission to bring to you the voices of seasoned and revered professionals in the field. Our episodes have seen the likes of Sue Nabeth Moore, Greg Daines, Jeff Heclker, James Scott, David Ellin, and David Jackson, who have generously shared their expertise on a variety of pertinent topics.
We’ve delved into the intricacies of Profit and Loss Statements in Customer Success with Dave Jacksson, explored the potential of Customer Success Platforms with Dave Ellin, and unravelled the role of AI in Customer Success with all guests. With Sue, we navigated the waters of Organizational Alignment, while Greg brought to light strategies for Reducing Churn. Not to be missed is James insightful discourse on the Current Trends in Customer Success and Jeff’s thoughts on Service Delivery in CS.
Each episode is crafted with the intention to ignite curiosity and foster a culture of continuous learning and improvement among customer success professionals. Our discussions transcend the conventional, probing into the proactive approach, and the evolving landscape of customer success.
Whether you’re a seasoned veteran or a newcomer to the industry, our goal is to propel your customer success prowess to greater heights. The rich tapestry of topics we cover ensures there’s something for everyone, from the fundamentals to the advanced strategies that shape the modern customer success playbook.
Our upcoming episodes promise a wealth of knowledge with topics like CS Math, Training, AI, Getting hired in CS, and CS Tool reviews, ensuring our listeners stay ahead of the curve in this fast-evolving field. The roadmap ahead is laden with engaging dialogues with yet more industry mavens, aimed at equipping you with the acumen to excel in your customer success journey.
At “The Customer Success Playbook,” our zeal for aiding others and disseminating our expertise to the community fuels our endeavor. Embark on this enlightening voyage with us, and escalate your customer success game to unparalleled levels.
Join us on this quest for knowledge, engage with a community of like-minded professionals, and elevate your customer success game to the next level. Your journey towards mastering customer success begins here, at “The Customer Success Playbook.” Keep On Playing!!
The Customer Success Playbook
CSP S3 E6 - Kevin Metzger - AI agents usage in 2025
In this forward-looking episode, Kevin Metzger delves deep into the transformative potential of AI agents in service delivery. The discussion unveils how AI agents are revolutionizing business processes by breaking down complex tasks into manageable components. Kevin explains the architecture of AI agents, their practical applications in service delivery, and the significant developments in AI technology, including ChatGPT's O3 model and its implications for artificial general intelligence.
Detailed Analysis
The episode explores several crucial aspects of AI implementation in service delivery:
Understanding AI Agents
Kevin provides a comprehensive breakdown of AI agents, describing them as specialized LLMs (Large Language Models) equipped with specific objectives and tool-integration capabilities. These agents can interface with various platforms through APIs, enabling them to perform targeted tasks within a larger process framework.
Practical Applications in Service Delivery
The discussion outlines a practical workflow where AI agents can transform meeting management:
- Automated note-taking during meetings
- Extraction and assignment of action items
- Integration with task management tools
- Automated notification systems
- Progress monitoring and tracking
Technical Infrastructure
Kevin highlights key technological developments:
- Salesforce's AgentForce platform for AI integration
- Crew AI platform for agent implementation
- The importance of API connectivity
- Multi-agent systems for complex task management
Addressing AI Limitations
The episode tackles critical considerations in AI implementation:
- Managing LLM hallucinations
- Implementation of verification systems
- The role of multiple agents in ensuring accuracy
- The importance of process definition before AI implementation
Future Outlook
The discussion emphasizes the evolving nature of AI technology and its increasing accessibility through improved APIs and integration capabilities. The conversation suggests that 2025 will be a pivotal year for AI adoption in customer success and service delivery.
Now you can interact with us directly by leaving a voice message at https://www.speakpipe.com/CustomerSuccessPlaybook
Please Like, Comment, Share and Subscribe.
You can also find the CS Playbook Podcast:
YouTube - @CustomerSuccessPlaybookPodcast
Twitter - @CS_Playbook
You can find Kevin at:
Metzgerbusiness.com - Kevin's person web site
Kevin Metzger on Linked In.
You can find Roman at:
Roman Trebon on Linked In.
Welcome back to the customer success playbook podcast. I'm Roman Trebon, and this is our Friday episode where we discuss the impact of AI on this week's topic. We've been talking to Kevin this week about driving team success. And defining your service delivery as a product. If you haven't checked out those episodes, give them a listen. They came out every week. Now we come out with an episode Monday, Wednesday, and Friday. So if you missed it, go back, definitely worth a listen today, though. We'll look at how artificial intelligence can help drive service delivery and what to expect this year with agents. It's great to have you back, Kev, for our, uh, third part of, of your show this
Kevin Metzger:week. Thanks, Roman. I appreciate it. AI is, is obviously one of my favorite topics. Um, so excited to have the show this week.
Roman Trebon:And not, not that I would have asked you to choose your favorite episode of the week, but if I would have guessed, I would, I would have leaned on this episode. So I'm excited to hear, uh, Kev, what you have on this. So talk to us about how AI. Can help with service delivery and you know, what can we expect from, you know, artificial agents as we move into, into 2025,
Kevin Metzger:it's interesting. I'm thinking about how to talk about this. I initially started my thought process on. Talking about things like SLA is where you're already recording a lot of this stuff into systems, and there's a lot of AI functionality that can be applied on the stuff that is measured, right? And it's really starting to drive some predictive analytics through machine learning and things like that. So I was really thinking about service delivery and agents, which is really the hot topic for 2025. If you're following AI, I'm sure you've heard about it. There's a couple of interesting things that I'll try and get into as we talk through this. Agents is going to be a way that you can really get the opportunity to start using AI in a way to get things done for you. And let's talk about what an agent is for a second. An agent basically is Using a large language model that you've given an objective to and giving it the opportunity to use tools. What do I mean by tools? I mean, it can use your web browser. It can use your, it can use APIs in products that you use. So, Salesforce, for instance, has built a product called AgentForce, where they're allowing AIs to integrate directly into their product through their AgentForce product, right? There are actually a number of platforms out there that are allowing you to build independent agents. So there's a platform out there called Crew AI. That's crewai. com is a platform for implementing agents. So basically helps you build agents in a environment where you're grouping them together. And let's talk a little bit more about the structure of agents. Again, Because you want to kind of minimize what each agent does and then get agents to work together. You've got an agent performing similar tasks. So how, how are we going to use this in service delivery? Let's, let's talk there. And this is just one, one kind of discussion to think about. I go into a meeting and after my, and I'm using AI to take notes in the meeting. That's just. Recording the meeting and taking the notes and then you generate, you generate some notes out of the meeting, but those notes have action items. Well, now I can start feeding those notes into a agent group that then looks at the notes and one agent will take the actually extract the action items and decide to see who needs to be. Notified that they have action items and when they're due. And then you can use another agent to work with that, to take those action items and insert them into whatever tool you're using to manage your action items. And the third agent will take them and stick them into emails to notify or texts or however you want to notice or slack to do notifications to your people. And then you can have an agent that's monitoring to see whether those tasks are getting completed on time. Agents are going to be able to do individual tasks really well. And when you're grouping together, you can get actually, if you understand your process and you've defined it, defined your process, well, like we're talking about in service delivery to finding out that entire process, then you can start. Building agents to complete pieces of that process one step at a time, there's a couple of things to understand about agents that as you start implementing them, it's not going to be as easy as saying, Oh, I just want an agent to go take my notes and put them in right? Because it's not one. It's not one step. It's multiple steps. It's going to be multiple agents, but the other thing is, okay, well, LLMs do hallucination, LLMs get things wrong a lot of the time, right? So how do I correct for that? One of the things that some of these newer models are doing. So if you were following the AI news in December, you may have heard that chat GPT just. Made known, they didn't really release it publicly, but they've got the Oh three model out now, right? So they had the old one model. Now they have this sort of three model and there's a lot of discussion in the news around Oh three, potentially being artificial general intelligence. Whether it is or whether it isn't beyond it's a thinking AI, what does that mean? And how is it working and how they put it together so that it's thinking, well, What effectively they're doing is they've got. Really multiple agents. So they, the question comes in and they run that question through the LLM at the base level, numerous times and get several answers. And then they have another agent on top that's actually evaluating those answers for correctness, truthfulness, best answer, that LLM is feeding best answers back out, or I'm sorry, that agent's feeding best answers back out. So you can design agents the same way. And, and you'll have to to get the best answer because now that's how you put some parameters around, Hey, this agent's giving me this answer, but it may have hallucinated that answer. But I've got another agent here looking at it to see whether that answer was hallucinated. Really, like I said, this can start getting applied to any process. Where you have a tool or an external factor that you're using that has, if it has APIs or other types of interface is that you can interface with it, then an agent can interface with that inner, with that interface in that manner, give it an objective, and then it can produce a result that ships it out to whatever the next destination is or other agent is. And then you can get them to work together. Okay. I guess I just figured for this episode, it'd be fun to talk about agents a little bit, how they work so you can get a deeper understanding when you're talking with people about agent or about AI and about agents, you can understand a little bit about what they're doing and how they're building it. I hope this was helpful.
Roman Trebon:Yeah, I think it's great. I think helping, you know, just the more you learn, people learn about AI, right? Like what are agents? Those are terms people hear. What does that actually mean? So it sounds like the technology is getting better. Right. Like they just improving the models are getting better over time. I'm sure that will continue. And the ease of our ability to consume them. It sounds like it's just going to expedite, right. The API feeds to what we can do with them, et cetera. So it sounds like 2025 it's Kev, I'm going to, go out on a limb and say, this will not be the last time. We talk about AI and how it's working, especially around the agents and, and how we can help with customer success and service delivery. So Kev, I hope you enjoyed your series of episodes. I know I did. It was awesome insights. Great to have a chance to actually interview you this time, right? So we're back with more episodes. If you found these mini episode formats helpful, please be sure to subscribe to the podcast and share it with your team, your colleagues, your friends. Kevin and I are super appreciative of your support and we'll be back next week with more customer success tips and strategies. Kev, I know you're as excited as I am for all the guests we're lining up in 2025. It's going to be great. I'm excited for our biggest and best year ever. As always to our audience, thanks for tuning in. And as always, keep on playing.
