Why the Agent Zoo Creates a Performance Problem

Hello and welcome. I'm Rob Wilkinson at CX Today and today we're taking a closer look at AI agent performance and the rise of the agent zoo. If you're challenged with disconnected AI agents entering your customer operations, or if you're just exploring AI as an opportunity to improve frontline performance without losing control of quality and accountability and stay with us because today I'm joined by Gal Rimon, CEO and founder at Centrical, a company who are helping organizations close this gap between insight and action across people and AI agents through its AI orchestration system. Welcome Gal. Thank you so much for joining me today. Thank you Rob. It's great to be here. So Gal, let's just sort of set a little bit of the scene for our audience today. What does the AI agent zoo mean in a CX environment? Today you bring so many AI solutions and providers and there are many builders. So it would be both the different vendors themselves you know like ServiceNow and NICE and um Agentforce Salesforce uh but also things that you will be build by yourself and each one has a role. One could help the agent uh the human agent, the other one can make phone calls and the other one will answer simple questions uh by the customer. And each one will try to optimize themselves. And once you optimize yourself, right, like in the basketball team, if you will optimize yourself, you will we will have maybe great dunks and uh amazing plays, but the team will lose at the end. And therefore, this is the zoo. The zoo is actually the um unmanaged and unorchestrated environment where human and AI agent are working for their goals instead of for the organizational goal or the customer goal. That makes perfect sense. It's such a great analogy. Uh and you know, you're right. We we would never um we'd never leave the human agents um if they've just joined alone to get on with things without a support and and to get them kind of give them context and getting them on board. So why would we do it with with with the AI agents? It does seem a little bit crazy when you take a step back I suppose. Um why why is this problem emerging now? Because we you know CS leaders now face so much pressure to implement AI. Is that driving like too quick adoption or are you seeing other things that are causing this? First, I think that the quick adoption is is great. It's something that we should encourage and we see so many POCs out there but uh I I think the numbers show that 80 or 90% are uh not going live and and and in my even personal perspective I see a lot of um successful POCs that don't go live because it's all about the environment where you you live at right so the AI agent could do great job, but it has a beginning and an end to the process where the process is usually bigger, you know, it's it's the customer experience or anything like that. And [snorts] today the the the the few big questions are accountability and ownership like who owns the process and and then you can say that the you know SDR AI agent owns the whole process. It's part of the process. And second thing is that everyone is putting so much focus on observability, but observability is only half the job. It's like okay, you understand where the problems are, you you basically don't solve the problem. and and and as a result you might have different uh um you know like uh gatekeepers or uh uh owners of the zebras of the lions and but but where is the orchestration both the human part the the leadership part and also from technology perspective that you could uh address bigger challenges and then all the AI agents and human will work together as one team for your customer. I think um recent research from from actionary which we've talked to um points to this kind of fragmented performance systems um and uh disconnected outcomes because that goes hand in hand. Um where are you seeing from your vantage point um organizations uh are most likely to kind of lose control of of AI agent performances? It starts with uh I call it the programs, right? You have a strategy. From the strategy, you need to um craft a program, a plan and then to deliver this plan to the different actors like AI agents, human actors and teams and and so on and and then to have the part of observability to make sure that uh everything is run as you expect and if not what are the interventions uh that you need to to have. And last but definitely not least, once you finish the program, how do you learn from what you did in order for next time to do it better, right? You launch a new mobile phone or a credit card product and uh next time you would like to do it better, right? And I think this is the the main gap that today it still runs in silos and losing this business goal at the top and and connecting everything to the business goal. Absolutely right. Um so I think we should ask the question why why should CX leaders actually treat this as a performance and execution issue rather than simply a technology issue. I think the in in one word it would be orchestration but but everyone is using the term orchestration because it's complicated and and in order to um make complication simple let's start with strategy from strategy you need to bring outcome the gap between strategy to outcomes are usually different behaviors and and in order to reach to different behavior behaviors by the field by the actors right the AI agents and the human agents uh you need to declare very clear clearly what's your business goal whether it's a product launch uh you would like to improve uh CSAT or you would like to increase price or handle better customer satisfaction and and so once you have this declaration of your goals with Centrical which is an orchestration a workforce orchestration system. You're able to basically create a plan very quickly based on all of your past experience, your codex of the business, all of the skills or all of the playbooks, everything that you did before to put this plan and interact with this plan, improve it. Once you have a draft of a plan, you are able to send it to all the different actors that they will act accordingly. They have their own optimization algorithms, but the same instructions should go like communication plan. Learn how to um present a product, a roleplay simulation that will certify you to present a product well as you expect with dozens of different scenarios and uh customers and so on and so on. Right? Once things are starting to be to run in the field, you run the show. You don't just um check where the show is. You have the observability part that realize where the gaps are. Once you identify the gaps, you fix it immediately. You change the plan in real time. And once you end this process, you learn from it and update back the codex the playbooks. I did it well. Let's learn from that. I you know I I didn't meet my goals on this and that. Let's learn from that and do it better in the next time. So you have this compounding feedback loop that will make everyone a better version of themselves both in the actors part like the AI agents and human agents the supervisors team leaders store managers shift managers and also in the leadership part next time I'll know how to launch this product better and this is a true orchestration layer that enables you to run the business and eventually your business is optimized much faster and that gives us a very real and and for probably for the first time a really true continuous improvement approach um which always strive to achieve but never seems to have been able to. Exactly. And uh last thing that is relevant but I think it's definitely not the it's last but not uh list the technology is definitely there now. Um you're absolutely right that that execution level is where we're struggling to to deliver. Um and uh yeah we we we should learn from the past. we've made so many mistakes in our in this industry around you know deploying technology and not getting it right that the you know there's really great opportunity to to do it to do it right this time. Thank you so much uh G. This has been super clear and given a really good understanding of why this uh agent zoo which I love uh is is become actually is becoming a serious you know CX leadership issue right now. Um, for me the main takeaway from this is that AI agents aren't uh just isolated experiments anymore. You know, they're becoming part of frontline operations. That means performance, ownership, governance, it all needs to be addressed early. Um, and um, and we're going to talk more on this. Um, so watch out for the next video if you're watching this one because there's going to be another part of this discussion where where we're going to look at how CX leaders can actually start to own the AI agent performance to build a stronger operating model uh around both people and AI agents. That's going to be then the next installment. So look out for that. Um, and of course you can discover more articles and videos just like this one at cxtoday.com. Uh, but for now that's all we've got time for. So, thanks again to Gal at Centrical for sharing everything. I'm Rob Wilkinson. Thanks very much for watching.