(en anglais)
Dans cet épisode du balado De l’IA au RCI, Victor Foulk de CGI anime une discussion percutante avec Niklas Bläsing, directeur-conseil expert de CGI en Allemagne, et Sebastian Glock, directeur du marketing des produits de Cognigy, sur plusieurs cas d’utilisation importants de l’intelligence artificielle (IA) agentique au sein de divers secteurs d’activité.
Le trio s’intéresse à la façon dont les agents intelligents offrent un soutien en temps réel tenant compte du contexte sur plusieurs canaux, réinventant à la fois le service client et les activités opérationnelles. Ils analysent les mesures prises par les secteurs du commerce de détail, des gouvernements et de la chaîne logistique pour être à l’avant-garde, et la façon dont les organisations peuvent surmonter les défis liés à l’intégration, à la gouvernance et à la préparation de la main-d’œuvre.
Au cours de la discussion, ils se penchent sur l’évolution de l’expérience client, qui s’éloigne des interfaces contraignantes et robotiques pour permettre des interactions intuitives et humanisées améliorant la satisfaction client et réduisant les tensions. Des appels sortants proactifs aux réponses hyperpersonnalisées, l’IA agentique permet d’offrir des services efficaces à grande échelle.
L’automatisation existante dépendait souvent de flux de travaux déterministes et de scripts préétablis, frustrant la clientèle et limitant les équipes de service. L’IA agentique fait un grand pas vers l’avant en favorisant l’écoute en temps réel, la compréhension contextuelle et l’action intelligente. Alimentés par de grands modèles de langage et l’IA conversationnelle, ces agents proposent des interactions humanisées et fluides qui surpassent l’habituel message d’accueil préenregistré.
Victor Foulk souligne qu’il s’agit d’un moment transformateur dans l’évolution de l’automatisation des processus, permettant aux systèmes de s’adapter et de répondre en temps réel. Contrairement aux agents conversationnels traditionnels, l’IA agentique a la capacité d’interpréter l’intention et le contexte avec une sophistication comparable à celle d’un agent d’expérience d’un centre d’appels. Sebastian Glock la définit comme le début d’une nouvelle ère, où les machines ne se contenteront plus d’exécuter des tâches, mais qu’elles collaboreront activement avec les humains pour offrir des expériences de service plus intuitives.
En transférant le fardeau de la clientèle au système, l’IA agentique favorise des interactions numériques instinctives et satisfaisantes, améliorant ainsi l’autonomie des équipes de service et la satisfaction des clients.
L’IA agentique est particulièrement efficace dans les secteurs proposant beaucoup de services, montrant une grande complexité et au sein desquels les attentes des clients évoluent rapidement. Par exemple, les détaillants et les entreprises de commerce électronique utilisent l’IA agentique pour personnaliser les expériences, simplifier les retours et gérer la logistique avec une intervention minimale de l’humain. Dans le secteur public, les gouvernements déploient l’IA pour améliorer les portails des services aux citoyens, transformant ainsi la bureaucratie numérique en une expérience plus accessible et réactive.
L’un des premiers signes flagrants de la demande d’IA agentique se manifeste dans l’expérience client, tant dans les secteurs gouvernementaux et commerciaux. Victor Foulk souligne que les effets potentiels y sont immédiatement observables.
Le concept de « clientèle » évolue également au sein des services publics. « Les citoyens sont des clients de la ville », illustre Niklas Bläsing, mettant en lumière la façon dont l’intelligence artificielle peut rehausser les standards de satisfaction et de réactivité. Outre les services aux citoyens, des cas d’utilisation s’observent également en gestion des actifs et en logistique, où les agents intelligents commencent à rationaliser les opérations et à favoriser une prise de décision plus proactive.
Qui plus est, l’IA agentique ouvre de nouveaux horizons, comme des services proactifs d’appels sortants permettant d’offrir un soutien en temps réel en fonction du comportement de l’utilisateur, des données de l’appareil ou des repères contextuels, ce qui était auparavant impossible à grande échelle.
Le passage de l’IA probabiliste à l’IA déterministe accroît la complexité. Contrairement aux agents conversationnels traditionnels, les systèmes agentiques répondent de façons différentes en fonction du contexte et des données. Cette variabilité oblige un changement de mentalité et l’actualisation des modèles de gouvernance, des cadres d’intégration et des balises éthiques.
Selon Victor Foulk, malgré l’enthousiasme suscité par l’IA agentique, bon nombre d’organisations rencontrent toujours des obstacles, dont le poids de la dette technique et la complexité des systèmes existants. Ces problèmes fondamentaux persistent. Parallèlement, il existe une pression croissante d’agir rapidement et de mettre en place de véritables solutions. Toutefois, Sebastian Glock précise que le besoin de bien faire les choses – d’innover de manière responsable et de respecter les balises établies, y compris dans un monde appelé à être progressivement façonné par les technologies d’IA générative et probabiliste – modère ce sentiment d’urgence.
L’état de préparation du système, notamment l’accès aux interfaces de programmation d’applications (API), la sécurisation des flux de données et l’infrastructure prête pour l’IA, joue un rôle clé pour exploiter le potentiel de ces fonctionnalités. De plus, une adoption réussie repose sur une gestion réfléchie du changement afin de soutenir les équipes dans l’évolution de leurs flux de travaux et de leurs responsabilités.
Les problèmes d’affaires clairs contribuent dès le départ aux plus grandes réussites d’intégration d’intelligence artificielle. Il ne s’agit pas d’implanter une technologie pour la forme.
Victor Foulk explique que l’adoption de l’IA agentique s’appuie sur une source d’irritation précise pour les affaires; l’objectif est d’abord de résoudre de véritables problèmes, puis de bâtir à partir de là. Une fois la pertinence démontrée, remarque Niklas Bläsing, la question de l’expérimentation passe rapidement au second plan pour laisser place au rendement du capital investi à l’échelle de l’entreprise et en faire une initiative d’affaires.
Fondamentalement, l’AI agentique ne sert pas à remplacer les gens, mais à les soutenir. En exécutant des tâches répétitives et laborieuses, l’intelligence artificielle permet aux gens de talent d’axer leurs efforts sur l’empathie, la réflexion stratégique et la prise de décision complexe. Victor Foulk insiste sur le fait que l’IA agentique continuera sur la lancée de l’IA générative, qui transforme déjà chaque composante de la main-d’œuvre.
Il en résulte un nouveau genre de collaboration humain-IA asynchrone qui, selon Sebastian Glock, est sur le point de réinventer la manière dont les services sont offerts et perçus. Ce modèle collaboratif accroît les résultats des services tout en stimulant la satisfaction, la rétention et la résilience de la main-d’œuvre.
D’autre part, cette approche stratégique permet aux organisations de s’adapter, de renforcer la confiance en interne et de passer graduellement des processus dirigés par les humains à ceux soutenus par l’IA.
Sebastian Glock est le directeur du marketing des produits de Cognigy, une entreprise novatrice offrant des solutions de service à la clientèle fondées sur l’intelligence artificielle (IA). Il favorise le rapprochement entre la technologie de pointe et les véritables besoins des entreprises.
Possédant une solide expérience en transformation numérique, Sebastian a offert ses conseils à des entreprises de pointe en Europe et en Amérique du Nord sur de nombreux projets numériques, ce qui leur ont permis de tirer parti de la technologie pour améliorer les interactions entre la clientèle et le personnel.
Habile orateur, Sebastian présente régulièrement sa perspective sur l’intégration de l’IA dans les activités opérationnelles à l’occasion de conférences et d’événements internationaux. Il allie avancements technologiques et applications pratiques pour les entreprises afin d’encourager l’innovation et la transformation de l’environnement numérique.
Découvrez comment l’IA aide les entreprises et les organismes gouvernementaux à générer des résultats tangibles et fiables.
Visitez aussi notre page d’accueil sur l’IA pour obtenir des perspectives, des ressources et des nouvelles sur les stratégies fondées sur l’IA.
Lire la transcription de la balado par chapitre (en anglais):
Victor (00:00)
Welcome everyone. I'm Victor Foulk, Vice President of Emerging Technologies for CGI Federal based out of Washington, DC. I lead our federal research and development programs, including artificial intelligence and GenAI applications, and I am thrilled to host today's discussion.
Today, we're going to talk about the real-world applications delivering tangible business value. We're going to cover what industries are gaining the most ROI and the most valuable use cases. What challenges organizations should expect when scaling agentic AI.
I'm excited to welcome Niklas Bläsing and Sebastian Glock.
Sebastian (00:31)
Victor, thank you for having us. My name is Sebastian Glock, and I work for Cognigy, a global leader in AI for customer service. So, some of the largest enterprises with the largest contact center operations trust us with their customer operations. And what we do is we build smart conversational AI agents that fulfill service requests, that talk like humans, that know how to act, and that know when to escalate to a human, and that basically act like super smart service reps in the contact center.
Niklas (01:02)
Hello, Victor. Hello, Sebastian. My name is Niklas Bläsing. I'm a Director Consulting Expert in Germany, and I'm responsible for the development and expansion of data and AI in our strategic business unit in Germany.
My work is focused on strategic business process automation. So, we are advising and driving activities within the areas of strategy, conception, implementation, and scaling AI business processes.
Victor (01:31)
Thank you both for being here today. Sebastian, let me start with you. From your perspective, what are the key business challenges that agentic AI solves? And how does that contrast to previous automation technologies and what those technologies couldn't do?
Sebastian (01:45)
I think the reality is that chatbots and voicebots have let people down for two decades. They were too robotic. They were not human-like, and they couldn't fulfill enough end-to-end transactions to be really helpful to humans. And that was because they used to be completely predefined in a deterministic state machine.
That means some smart conversation designer would have to anticipate every potential turn of a conversation, every question a customer could have or ask and anticipate anything that could happen in the course of a conversation.
Now, agentic AI completely changes that. It uses a combination of conversational AI and large language models, generative AI to deliver truly human-like experiences in real time by actively listening to what a customer is saying and then creating that response and taking action on real time. And that really changes the game for everybody. For the customers who will receive better service instantly on the channels that they truly prefer 24/7.
And it also changes the game for businesses because these AI systems become truly helpful and then can become a huge cost saver in the contact center.
Victor (02:50)
I love that. And I can tell you, by personal experience, there is nothing more painful than having a robot answer the phone and having to navigate some very fixed navigation menu and literally begging to speak to a human.
Sebastian (03:04)
Exactly. In a way, it's reversing the structure between the caller and the agent on the other side. Because in previous bot generations, think of the classical IVR, press one for this, press two for this, the human caller would have to understand the bot mechanics. And now it's working the other way around. The machine is truly interpreting what the customer is asking for and acting based on that input, pretty much like a smart contact center agent with some years of experience would do.
Victor Foulk (03:28)
Excellent. Niklas.
Niklas (03:29)
In my point of view, decision making is one of the most important points because decisions are based on a real time context and data, which is now a real next step from the predefined scripts, which we had an automation before. So, we have a whole value stream end-to-end based on live data and live context, which we can use.
And this is a large business challenge which we can now address with agentic AI, which was not possible before. It is more dynamic; we have less human intervention, and we can coordinate our processes across systems to achieve outcomes rather than just execute any tasks.
Victor Foulk (04:13)
Yeah, I fully agree. You brought up the scale and data. And one of the things that the agentic AI capability is bringing to the table is Sebastian, as you said, more complex decision-making, the ability to navigate more complex dynamics and the ability to scale based on data and new data allows personalization and dynamic adaptation as use cases evolve or as business tasks evolve.
The static RPA of the past, very functional, very efficient, but it requires explicit human intervention in order to adapt it to new use cases or to evolving data. And I think that this is really a transformational time in terms of process automation.
Victor (04:56)
Okay, so with all of this new capability from your perspectives, what industries are leading the way in leveraging this agentic AI capabilities and what are the results that they're seeing?
Niklas, let's start with you.
Niklas (05:06)
When we take a look on the retail and e-commerce industry, we have a large benefit in using agentic AI because they can use it to optimize the whole value stream of inventory optimization, returns management and customer service.
So they can directly feel the result of customer satisfaction, lower operational costs and see that we can really have the whole value chain to see the customer is getting the order in the front end but everything which is happening in the back end and in the logistics part, everything is automated and they can directly see when will the next delivery appear and can do some better improvement of all the stuff which is available and there's already some nice use cases in place where we already have the first experiences at customers and nice success stories.
Sebastian (06:00)
Yeah, I think retail is an excellent point, especially because in retail, there's a ton of data available and AI agents can use that data to truly personalize experiences. So, when somebody calls or texts and when they're logged in, it's very easy for an AI agent to provide a journey that is aligned with previous purchases, with ongoing orders, or with anything that the customer could ask for. So I agree with you that e-commerce is a great example.
But in general, I would say that the industries most prone to agentic automation are those with very high service interaction volumes and with some degree of repetitive use cases, but they can be complex.
Aside from commerce, we see a lot of adoption in travel and hospitality. And that is because there's a high peak volume that is occurring because when people are in trouble regarding their travel plans, they all call at the same time. Everybody wants to rebook their flight right now because they know the demand is high and the availability is quite scarce. So, there's a huge peak load in these industries, and the complexity is also quite high so that we see a truly high adoption rate in there.
Another area, this is more of a use case angle that is almost new with agentic AI, is outbound calling in proactive services. Because that is something that is extremely hard to do with previous bot generations. And the reason is that if you call people and it can be with perfectly good intentions, you have no idea how they're going to react. Some of them may be happy about a machine calling them. Others will be upset, or again, others will be asked to receive a call back in five minutes or maybe a call back in five days.
Some customers may ask to talk to a human right away. You cannot anticipate all these things in deterministic chatbot machines, but agentic AI can, because those agents making those outbound calls, have clear goals and they are able to take in all the signals, everything that the customer is saying on the phone and also that is available, as a background about why these customers are called in the first place and create a hyper-personalized unique journey for that specific occasion. That is something that was previously completely impossible, and we see a lot of uptick in terms of proactive outbound AI agents for chat and also for voice.
Niklas (08:12)
I can add another industry. The Germans love their bureaucracy, and we have a lot of large processes which take a lot of time. When we focus on the public sector, we have a new field of play because of agentic AI. We have the possibility to assist the citizen service portals, which are building up in the digital transformation in Germany.
There is a huge benefit because we can help all the people within the back office sitting within the cities and all the citizens which are going through all the steps. There we find a new level of customer satisfaction because even the citizen is a customer of the city in this moment where we find assistance and if you can take it into business consideration for ticketing and case management you have a lot of AI agents in this field already and there's a large potential where if we have AI improving all the process and helping you to fill out all these forms which are really complex.
Victor Foulk (09:13)
Yeah, so I think in the US, what we've seen is very similar to what you've both described. Customer experience as a broad umbrella seems to be the most obvious early demand signal and the area where we're seeing early adoption, both in government and commercially.
The high touch aspects of engaging with not just customers, citizens, or any consumer of services. On the federal side, we're seeing a lot of agencies leveraging these capabilities to dramatically improve the efficiency. Because when it comes down to customer experience, it's not just having a good and factual interaction—they also want it fast.
They also want to be able to see and ask what's the status of my application? What is the status of this service that you're going to provide to me? And being able to dynamically field those types of interactions and integrate with a variety of different data sources is providing that much improved customer experience in government as well as commercially.
Niklas, you'd mentioned retail and supply chain and logistics. That's another area that we're seeing a large uptick. Commercially, the large retail and logistics entities are already deep and advanced in this and kind of leading the charge.
But other portions of industry and government are leveraging the deep integration capabilities, allowing agentic AI functions to access IoT systems, sensors either on the shop floor, sensors in the field, and being able to take action, most of the time with human in the loop or human on the loop supervision. But these agentic capabilities are able to interact in a reasonable way with the physical environment as well. We're starting to see that especially in asset management and logistics, that is a prevailing use case.
Sebastian (11:00)
Yeah, we do have a use case with Toyota, the car manufacturer, who has built an integration into the onboard electronics of the vehicles where when the engine warning light flashes up, an automatic call is triggered to the vehicle owner. And that is not necessarily the driver. It can be someone in fleet management who isn't even aware that one of the cars may have a problem. And that triggers a whole process up to scheduling a workshop appointment for the vehicle owner.
And that ensures that the vehicle remains its value, and a problem is solved even before it gets worse. And I think that is also a brilliant example, not just of how data is brought together, but also how outbound calling and proactive services can really mean a difference for the consumer.
Because many times those outbound capabilities are kind of negatively connotated with robot calling. But it's not always about, you know, upsell and just proactively trying to promote something. It can also be providing a meaningful service to the consumer. And I think that also ties back to what Niklas said about government.
You know, I would appreciate it if I got a call before, let's say, my passport would expire so I can take action, right? And that can be a differentiating service offering that not many companies today provide. It will be more important in the future to keep up with the service expectations of consumers.
Victor Foulk (12:15)
I love that. I think that when we talk about the state of the technology, it's incredibly powerful. But when you start talking about how these capabilities can allow us to lean in proactively to challenges in a way that we never could before, whole new value streams are going to manifest, not just in terms of efficiency and experience, but cases like condition-based maintenance.
We can actually leverage these technologies and tools to coordinate the activities of humans to maintain systems. And I think that's a really important aspect of the future of generative AI and agentic AI in that it's not just reactive and it's not just taking input from humans. It can actually be an active contribution to the workforce. And as we go forward, we're going to be looking at these agentic capabilities much more like human additions to our workforce, individual entities capable of activity as a part of the whole, rather than simple bits of code and software.
Sebastian (13:13)
Let me add to that. It is also a new era of human and machine collaboration because not every service request will be solved instantly, not today and not in the future. But when you first touch is with an AI agent, that can trigger a cascade of asynchronous communication where the AI agent checks in with other agents. They may check in with human supervisors and maybe you get a call back an hour later.
The whole idea of this agentic experience is that someone actively works on your behalf somewhere else and comes back to you with a solution. And that previously wasn't possible. And this async working collaboration is something that's entirely new and that will also shape the service experience of the future.
Niklas (13:53)
I guess that's the differentiator, because before you get a ping, a message on Teams that was the maximum of possibilities. And now voice chat, mail, everything is possible even across the channels. And to have a process communicating with you in the way you like to—that's a new era.
Victor Foulk (14:11)
Indeed. So one of the technical nuances that implementers are going to need to grapple with in some of these cases, the past 18 months of generative AI application development, there's been a hyper focus on latency and response time and the use cases you just described, in some of those cases, these agents are working behind the scenes, leveraging a lot of large language model integration, a lot of token consumption to make sure that the plan for achieving a task is the right plan, that the steps executed to achieve the task are done correctly and coordinating with other agents in a deep and impactful way. That's not going to be in every case instantaneous, but it is going to be powerful.
Organizations are going to have to grapple with a different mindset with some of the deep agentic integrations that are going to be different than past implementations.
Beyond that, what other unique barriers to agentic AI adoption are you all seeing and how are those barriers different from generative AI of the past 18 months, 24 months? And then you can talk about some of the recommendations you have for how organizations can overcome those challenges.
Niklas (15:23)
In my point of view, the complexity of integration is the largest barrier which we face in Germany right now. You need interfaces which are ready for agentic AI, that are able to have the seamless communication, because if the interface is not there, it's not making sense to have an agent because it has to enter the data sources and get data available.
That's something where we have to work together with the vendors to make the interfaces agent-ready that we get a full disclosure of their access of the metadata, for example, to get the agent running within the systems, and we have the possibility to get an access for a technical user like the agent.
On the other side, change management because most of the people are just working on the tech side right now and people want to see fast use cases. But we have to think about the people behind the processes and how do we catch up with them?
So, the next challenge, how to get all the people in one boat and disrupt the siloed workflows and get into the next stage in era of AI.
Sebastian (16:31)
I do think the biggest shift in agentic is really driven by the core of the technology. And that is a shift from deterministic mechanisms to probabilistic mechanisms. We've all seen in ChatGPT, you ask the same question twice, you get two different answers. That is in the nature of LLMs, right? And it is like that purposely. If you change any data point, you reword your input slightly, you will get a different answer every single time. And now chatbots and voicebots are behaving like that. That is a feature, not a bug.
But now, what does it mean? It means like you need to let go of some of the 100 % control that previous bot generations had. And there are still use cases where that is absolutely essential.
For example, we work with pharmaceutical providers who use chatbots for people to report medication side effects. You will never want that to be handled with a large language model. That is something for a deterministic process, for form filling where you just follow the protocol step by step, just like a human agent would do, right? There's no more time to small talk when it's becoming a life and death issue.
But now in other cases, LLMs are perfectly fine. They're good for answer retrieval, good for knowledge retrieval, and so on. So that is really great. But still, organizations have to let go and trust the AI to work.
But it's more than trust. It's also about implementing governance rules in those autonomous systems, defining the guardrails, making sure that there is no data leakage back into the model. Now, if you think that through, what does it mean for the people rolling out the bots? It also means testing becomes a lot harder, right? You cannot follow the same kind of playbook step by step and just check for goal achievement because the paths, they are not always the same. So, testing requires also AI to successfully test on regressions or test on bot behavior. And everything gets to Niklas point here a little bit more complex. It becomes truly an integration problem.
The change management that Niklas mentioned is something that we also see with our customers. It is a new mindset of bot building and leveraging kind of agentic workers in the organization. And that is a huge shift on how to orchestrate those experiences. There's also an increasing demand from the consumer side for agentic experiences.
Everyone, including the decision makers and companies now understand how good AI can be. They're all ChatGPT users, right? They're creating images for fun and they're using it for knowledge retrieval, for, helping them with the upcoming board meeting.
So, there is a pressure to do it right, to get it going, but stay within the guardrails even in a probabilistic LLM powered or GenAI powered world and balance between parts of the experiences that are hard coded, and those which are LLM driven and a little more probabilistic.
Victor (19:07)
Yeah, so I know that I asked about unique barriers, but one of the things that strikes me here as I'm listening to the conversation is the same barriers of technical debt and legacy systems complexity still presents a challenge today. Data and having control of your data estate is still a foundational and fundamental challenge.
Niklas, you mentioned being able to understand your APIs, not just exposed, but in a form that the agents can dynamically process and understand is important. And in a lot of network ecosystems, that represents a whole body of work that has to be done before agentic AI is even truly feasible in an organization.
Only then can you really get into deploying these agentic AI capabilities and then once you do, Sebastian, as you mentioned, ethical and regulatory concerns, data privacy and security, and then how do you actually measure and monitor the performance of these agents, recognizing that it's no longer a deterministic system.
And then organizations are going to really have to understand their AI use case library, their AI use case portfolio, and have a good framework for selecting their use case, how they will deploy it, and how they will measure its effectiveness.
So, what would you say to clients who are looking at these agentic AI capabilities and wondering, where do I start? Niklas?
Niklas (20:27)
Start small but think big because people have to find the first use case and measurable pain points because when you find something where you need some assistance you get the attention and once the return on investment is proven, scaling becomes a business-driven conversation.
Victor (20:47)
I love that. Really, it comes down to identifying the business needs, setting objectives, solving a problem, and measuring return on investment. It's not about the AI. Where do I start? You start with a pain point, and you build.
Sebastian (21:01)
We also see with our clients that it is always a favorable approach to start small and pick those use cases with a measurable impact. And the reason is that agentic AI is not an either-or. It is not something like a big launch where suddenly you switch your contact center over to agentic. No, you can start with individual use cases and just use it for something.
And depending on the majority of the customer, sometimes it isn't even advised to start with agentic. Maybe something like call summarization for the human agent in the contact center is the most impactful first use case.
For other clients of us, it is rebooking because rebooking is very repetitive. It's very complex. It is hard to do on the phone line. You need that visual feedback on the UI, and the chatbot can provide that. Maybe that is a great use case, even though it is not agentic to start with. And from there on, you can use the gaps and get internal buy-in to tap more and more into the power of AI.
Because just handing over from a process that is working, even though it's costly, to an LLM process that may be better, but you don't know yet, that is a very tough obstacle to overcome. But if you start small, in a dedicated market with a dedicated use case, that also gives a lot of security that you're on the right track.
And from there, you can expand to more complex use cases and hand over more responsibility over time from the humans to the AI agents. Because ultimately, what our customers are striving for is that AI-first contact center. But it's not something that you're going to get on day one. You have to start somewhere, and that is typically with those high impact use cases that are also good to measure.
Victor (22:32)
I think that highlights another area that when clients are asking, where do I start?
We have structured approaches and a detailed framework to assess an organization's readiness. And part of that is the workforce. There is no part of our workforce that is not touched by generative AI today and agentic AI tomorrow will be no different.
And so, when we talk about workforce readiness, and I've long said in the United States and I think potentially globally, the single biggest risk to national security is the workforce skills gap. The technology is evolving at a pace that the workforce struggles to keep up with. And it's more than just data and APIs. We have to evolve and upskill the human element to be able to properly integrate with these highly capable systems we're going to be bringing to the fore. In that regard, how do you handle the human element of this type of technology deployment, and how do you advise clients?
Sebastian (23:28)
Specifically in the contact center, what we hear again and again is that the biggest bottleneck currently is staffing the contact center with the right agents. It's really tough to hire for the contact center. And the reason is that the work is not easy to do. It's not often very well paid, but it gets harder and harder. Why? Because the easy stuff gets eaten by self-service. What remains in the contact center is the really complex task dealing with customers who are emotional, customers who are under time pressure, and so on.
What we do to get buy-in from those managers of the human workforce is showing them and educating them how AI can assist those humans in the contact center and make their lives easier. And it can be something as simple as a fully automated identification and verification, asking the customer for the name, the customer number, using the multifactor authentication to securely sign them in into the system. That is nothing that humans like to do and it's nothing that they should be doing.
But if you implement that with AI agents, you get that part of the process automated, which often cuts out already 60 or 90 seconds of manual handling time. But what you also get is the intelligent routing so that the human caller is directly routed to the right person in the shortest time possible. And that alone already takes a lot of pressure out of the service interaction.
What Niklas said earlier is absolutely true. The worst thing that you can do to a customer is letting them wait and then making them repeat themselves. But speaking to a human agent after a short bot interaction, and that human agent says, “Hey, Victor, I know what you're calling about. I'm on it. Please just give me a couple of seconds.” That already takes so much pressure and emotions from the conversation. And this is the kind of buy-in that we also want to get from the human service workers to understand and see how they can be working hand-in-hand with AI in the future.
Niklas (25:15)
Empathy is a big part because people already said we are not ready for AI services because humans are more empathic. But if you get shouted at 10 times, you're not that empathic anymore. So, I guess the bot is already in a good position when it can tell the [human] agent, “Let's take a short break after this call.”
So, the burnout factor of an [human] agent could be much better, and the AI helps the phone [human] agent to be empathic and the best answer giver for the client afterwards.
Victor (25:49)
That is a phenomenal example of where the power of this technology rests.
Sebastian (25:54)
Yeah, Victor, I think one of the worst things that AI could be used for is to increase the amount of deflection and stop customers from interacting with organizations. For a long time, the contact center has been viewed sort of as a cost center, something that you have to do, but nobody really loved it.
Now, AI gives an opportunity to flip that around and have more conversations with organizations. Make it easier for people to reach out. So, in the future, you could have basically your favorite company on speed dial and talk to them whenever you want.
Victor (26:24)
I think you just outlined a concept for antagonistic offensive agents. I love it.
Sebastian (26:29)
Totally.
Victor (26:31)
Yeah, and that's a wrap for today's episode.
We delved into the power of agentic AI. We explored the industries leading the charge, and we tackled some of the unique barriers to agentic AI adoption. So, whether you're just starting out or looking to scale your AI initiatives, remember that the journey begins with understanding your business needs and setting clear objectives from the start.
If you have any questions or topics, you'd like us to cover in future episodes, reach out and be sure to follow us and keep an eye on CGI.com for the next time.
