Dans cet épisode du balado De l’IA au RCI de CGI, Victor Foulk, vice-président, Technologies émergentes, pour CGI Federal discute avec Andrew Donaher, vice-président responsable de l’IA, des données et de l’analyse pour CGI au Canada. Ensemble, ils discutent de la complexité évolutive des chaînes d’approvisionnement mondiales et du rôle transformatif de l’IA agentique.
De la conformité réglementaire à la logistique intermodale en passant par l’automatisation centrée sur l’humain, ils présentent leurs perspectives sur la manière dont les agents intelligents favorisent la résilience, l’agilité et la durabilité. La conversation met l’accent sur l’importance de la visibilité des données, de la gouvernance responsable de l’IA et de la gestion du changement centrée sur l’humain alors que les organisations se préparent à l’avenir des écosystèmes de chaînes d’approvisionnement intelligentes.
1. L’IA agentique refaçonne les écosystèmes de chaînes d’approvisionnement modernes grâce à une approche globale et centrée sur le consommateur.
Les chaînes d’approvisionnement ne sont plus considérées comme linéaires, mais bien comme des écosystèmes complexes et interdépendants. « Les citoyens, les parties prenantes et les entreprises considèrent désormais leurs chaînes d’approvisionnement comme un écosystème », explique Andrew. Les perturbations externes telles que la pandémie de COVID-19 de même que les changements de politique commerciale ont renforcé la prise de conscience ainsi que les attentes des consommateurs et des acteurs du secteur public. En conséquence, les organisations repensent leurs chaînes d’approvisionnement dans une optique plus intégrée et centrée sur l’humain.
« Nous avons connu tellement de perturbations… et je pense que les consommateurs sont plus exigeants. Lorsque nous adoptons cette vision des écosystèmes de chaînes d’approvisionnement centrés sur les citoyens ou les consommateurs, je trouve que cela correspond très bien », explique Victor.
2. La visibilité complète des actifs et l’intégration de données sont essentielles au déploiement efficace de l’IA agentique.
Les chaînes d’approvisionnement modernes nécessitent une visibilité en temps réel et consolidée des stocks, de la logistique et des systèmes financiers. « La capacité à intégrer les systèmes de gestion des actifs, les systèmes financiers et les systèmes logistiques […] est un point de départ essentiel », explique Victor.
L’IA agentique excelle dans cet environnement, en automatisant la prise de décision et en améliorant l’agilité grâce à l’exploitation de données intégrées de haute qualité. Cela permet d’automatiser des processus complexes tels que l’acquisition et le réapprovisionnement, au-delà des capacités traditionnelles de l’automatisation robotisée des processus. Récemment, explique Andrew, « nous avons découvert que nous pouvions tirer parti de l’IA agentique » pour optimiser les règles et réglementations afin de classer correctement les marchandises, ce qui en facilite ainsi leur transport transfrontalier.
3. L’IA responsable nécessite transparence, surveillance et supervision humaine pour la prise de décision.
À mesure que les systèmes d’IA deviennent plus autonomes, il est de plus en plus important d’assurer la transparence et de maintenir la supervision humaine. L’IA agentique doit soutenir les experts, et non les remplacer. « Tout est conçu pour aider les experts à être plus efficaces et performants. La supervision humaine est essentielle afin que l’expert sache ce qu’il examine et qu’il prenne la décision finale », explique Andrew.
Des modèles de gouvernance clairs tels que l’intervention humaine et les structures d’approbation sont essentiels pour déployer ces technologies de manière responsable et sûre. « Vous pouvez automatiser une grande partie des processus, mais aux étapes clés, vous avez vraiment besoin d’un être humain qui a le pouvoir de décision pour accepter un résultat susceptible d’avoir un impact sur un consommateur ou un citoyen », ajoute Victor.
4. La transformation de la main-d’œuvre est essentielle à la pleine réalisation de la valeur de l’IA agentique.
Adopter l’IA agentique requiert plus qu’un simple déploiement technologique. Il faut également investir dans la formation de la main-d’œuvre. Lorsqu’elles sont mises en place avec une gestion du changement réfléchie et un renforcement des compétences, ces technologies réduisent le stress du personnel et augmentent son engagement. « Nous constatons une réelle diminution du niveau de stress de la main-d’œuvre, car ces personnes ne sont plus submergées par des tâches sans valeur ajoutée », explique Andrew. Toutefois, si elles sont mal gérées, ces technologies peuvent créer des tensions et aggraver le déficit de compétences. « Lorsque l’adoption ne progresse pas rapidement, l’adaptation du personnel [aux nouvelles technologies] reste une source de stress », explique Victor.
Explorez d’autres épisodes du balado De l’IA au RCI et découvrez comment l’IA transforme les entreprises et les organisations gouvernementales. Visitez cgi.com/fr/intelligence-artificielle pour obtenir des conseils, des ressources et des mises à jour sur les stratégies fondées sur l’IA.
Victor Foulk (00:00)
Welcome, everyone. I'm Victor Foulk, Vice-President of Emerging Technologies for CGI Federal, driving innovation and AI for the public sector. I'm excited to be hosting today's discussion with Andy Donaher, Vice-President and Canada's lead for AI, Data and Analytics. Today, we're going to dig into the complexities of modern-day supply chains in the era of AI and unravel some new ways of thinking about those complexities in a citizen or consumer-centric fashion, empowered by modern agentic AI principles.
So, Andy, you're a frequent guest on this series, but for those who may be tuning in for the first time, why don’t you give us an overview of your mission here at CGI and your role in driving supply chain solutions?
Andrew Donaher (00:38)
Sure, Victor. Thanks for the intro. Hi, everyone. Very excited to be here today. I’m a Vice-President at CGI for AI in Canada. We're very much focused on delivering that value, and one of the things that we're seeing in the market today is how can we help organizations realize the value from AI. There's a lot of pressure right now economically, demographically, across Canada and around the world on inventory optimization, logistics optimization, flexibility in the supply chain.
And it really behooves us to make sure that we're leveraging these new technologies to help organizations be able to operate in an increasingly complex and an increasingly time-sensitive world. So, very much looking forward to the discussion today.
Victor Foulk (01:25)
Excellent. So, given the disruptions that we've seen over the past, say, five years—from our COVID response to the blockage of the Suez Canal, and shifts in political policies around trade, how are you seeing mindsets shift with respect to supply chain resilience and agility?
Andrew Donaher (01:42)
I don't think I've ever seen in the world more people more aware of their supply chains. I think that's the number one thing. When you start to look at the externalities, citizens in general, stakeholders and corporate organizations are now looking at their supply chains as an ecosystem.
It's no longer a straight line of a truck on a line from A to B. It is looking at their supply chain ecosystem holistically and understanding how to balance and leverage those things. Even the average citizen now is much more aware of where their fruits and vegetables are coming from, where their clothing is coming from, how is that being delivered on time and expecting things to be able to be delivered from overseas on XYZ platform in hours.
And those expectations are continuing to evolve and expand, and they're making more demands on their supply chain as well. So, organizations need to think about this holistically in terms of alternatives to supply. And so, that's what I'm seeing.
You're a practitioner as well, Victor. Interested in your opinions on this.
Victor Foulk (02:50)
I have to agree with you wholeheartedly. I don't think our citizens or consumers have ever been more aware of the complexities of the supply chain ecosystem around them. And I think that's driven largely by the fact that we have seen so much disruption in the not-so-distant past.
I think that consumers are actually demanding more. And when we take this view of citizen or consumer-centric supply chain ecosystems, I think it maps really well. As you know, I'm responsible for innovation in our federal government market and we're seeing the exact same mindset shift in terms of government services. And then, the supply chain of government services is being applied with a citizen-centric focus, and it allows us to look more holistically at the broader mission ecosystem in a way that a product or particular service-focused view of the supply chain ecosystem just wouldn't provide.
Andrew Donaher (03:50)
So, I wanted to ask you a question. We're both practitioners. I'm super interested about some of the work that you're doing around how you're using agentic AI and how we're enabling visibility and flexibility from the systems, leveraging agentic AI and AI.
And, I think, everybody would love to hear a little bit about how you're doing that.
Victor Foulk (04:09)
Sure. For AI in general, data is at the center of everything, right? Being able to integrate asset management systems, financial systems and logistics systems to approach this concept of total asset visibility is a key starting point. And we've got a variety of programs, some very large ones with the Marine Corps and Defense Logistics Agency and others around federal government where we provide these total asset visibility services with solutions.
And when you have visibility and you have data and you couple that with workflows, you gain efficiencies. And when we apply these modern technologies, such as the agentic AI platforms that we develop, you can begin to automate these more complex processes that exceed the capabilities of robotic process automation, and you can begin to move faster and more reliably with various aspects such as acquisition and contracting.
Consider the consolidation of acquisition that's going on in the US market right now, where commodity acquisition services are being consolidated to the General Services Administration. Where you used to have contracting activities across the US federal government, those processes are now consolidating.
The data is going to be centralized, and there is opportunity for agentic AI implementations to conduct acquisition processes, such as being able to monitor inventories at a particular agency's location, consumables and being able to do things like automatically procure replenishments and distribute those to the agencies.
Those are all things that are in the realm of possible now. It's really enabled by having a consolidated view of data and then being able to securely implement technologies that act on that data.
What about you? What are the most promising use cases you're seeing in agentic AI for the supply chain? And talk through some of the impacts.
Andrew Donaher (06:05)
You're 100% correct in terms of the data. One thing is taking into account externalities. So, if you're looking at intermodal alternatives or flexible warehousing or commercial delivery lines, one of the things that we're seeing is the focus on the breadth of the data, not just the quality and not just the understanding. Those are super important. You've already mentioned those. So, I want to just add on to it the breadth.
So, by being able to leverage agentic AI and having agents look for wider components of data or additional variables or feature engineering to help us endeavor to take externalities into account is really helping us to look at how we can tweak the edges of these things.
Another one of those is quantum computing. Another one of those is trade and compliance. So, one of the things that we're looking at is how are we going to leverage agentic AI beyond the data, but in a regulatory capacity.
And because of the sheer amount of data, we're finding that we can leverage agentic AI to be able to look at those rules and regulations, look at those classifications and help in understanding, because we all have a human bias. We all think, “Yeah, I know that rule. I know that classification.” And yeah, you're an expert, but maybe you don't know about the one over there from a couple of years ago or the one up here that's just new last week. And so, we're fighting a great deal of value in that, in supporting those humans to make sure that they're optimizing that delivery.
And then, the other one is maybe not from an agentic perspective but being able to take advantage of the complexity of the data by leveraging quantum computing. A lot of people don't know that quantum computing is a real thing right now. I've done a couple of projects myself that are in an operational capacity with an organization called D-Wave.
And when you're starting to look at the complexities of a supply chain, and really optimizing your intermodal capabilities, your warehousing capabilities, that's where those quantum solutions really come into play.
Victor Foulk (08:33)
Yeah, I can't agree with you more. I remember the excitement in the last conversation you and I had around our generative AI platform and the first time we had an agent successfully communicating with both AWS and Azure Quantum Services. I think the realization we both had was that this really, truly is a pivotal time in the coalescence of these technologies.
We are evolving to a point where agents, like AI, are going to have to be managed differently. We've said this on previous podcasts. It's not just a block of code. You really do have to manage it, kind of how you would manage a human augmenting your workforce. And when you start to think about it that way, all the examples that you listed start coming to play.
You have, not just agents that are processing and executing the workflow to make a particular thing happen, but independent agents that are overseeing the process to ensure compliance, to ensure security. You have other potential agentic applications, which I really love that you mentioned, where you can have agents that are responsible for looking at the broader context of data and enumerating additional elements of your supply chain visibility that you didn't have already.
The potential for acceleration and scalability there really makes this period in technology evolution an incredible thing to be watching. We talk about speed and all the capability that the emerging technologies bring and placing the citizen and the consumer at the center of this supply chain ecosystem. It really emphasizes how critical the ecosystems are and, I think, the potential impacts of improperly applying this technology in those ecosystems.
So, from your perspective, how do we ensure that the agentic AI makes the splash and delivers results fast. But not just fast, also responsibly and safely?
Andrew Donaher (10:21)
So much to unpack there. But I think the two key things to that are number one, the transparency. It is understanding where it came from, understanding cosine similarity and the truthfulness and groundedness and all these different types of metrics that help us to understand how much confidence we can have in the answer.
Number two is that you know what you're looking at. So, this is all designed to help the subject matter expert be more efficient and more effective. It's not going to make a lay person an expert. So that human in the loop, so that the expert, knows what they're looking at and their making the final decision is critical.
One of the examples I always use is if you put me in a Ferrari and you put Michael Schumacher in a Ferrari, you get two very different results. It's the same kind of thing. Giving a non-expert an intermodal optimization alternative is a little different than giving it to some other people. So, that's kind of how we're approaching it.
Victor Foulk (11:20)
Yeah, and one of the ways that we're talking about it here, is really looking at it as a technology-augmented workforce. And when you think about building a team and constructing a team to go off and execute a particular mission, you have folks that are responsible for elements of the decision-making process, elements of the execution. But there's always someone in any org structure who is ultimately responsible. There's approval authority.
And when you're implementing these systems, you can apply a significant amount of automation, but at the right points. You really do need to have a human that has the decision-making authority to accept an outcome that could potentially have an impact on a consumer or a citizen.
I think that that human in and human on the loop type of governance model is still going to be critical as these technologies continue to infiltrate the supply chain, if you will.
Andrew Donaher (12:10)
Yeah, and I'm interested in your thoughts on this and what you're seeing. One of the things that we're seeing is around flexibility in the workforce by being able to leverage these agents. So, as the demands increase for more flexibility, more autonomy for workers, citizens, and consumers, that puts a lot of pressure on the people in the workforce associated with these supply chains.
So, one of the things that we've seen is that, as we're giving people these tools and agentic AI assistance in different capacities, we can actually see the workforce's stress levels decrease because people aren't as inundated with having to do the non-value add work or the unproductive work they can focus on what they need to do.
And there is a paper written ages ago now, it was a year and a half ago, I think. It was called Generative AI at Work. And in that paper was one of the first peer-reviewed studies. I believe it was done by MIT or Stanford in this space around generative AI. And it talks about yes, the productivity gains and all that.
And one of the things in that paper that it talked about was the increased level of engagement, the decreased stress, and the increased level of satisfaction that employees were having when they use these types of tools. One of the things that we're seeing right now is you can see people’s shoulders dropping.
They're relaxing because they're not having to do all that. And so, that's having a great impact on people's productivity because they're actually able to do more and feel more engaged in what they're doing. So, a little bit of a byproduct.
I was just wondering if you're seeing some of that because you deal with some pretty high stress situations too.
Victor Foulk (13:47)
We do see the same thing in pockets, and the reason I say pockets is because one of the greatest national security threats we have for every nation across the globe is that workforce skills gap. The pace of technology development is far exceeding the pace of technology adoption and the skill, the upskilling of the workforce.
So, where you have appropriate workforce upskilling, where the adoption of these technologies has been done appropriately with the right organizational change management, 100% agree. We're seeing exactly that result.
We're seeing the opposite in some places, where the adoption is not moving at speed, where the workforce skills gap is still a pervasive problem. We're seeing technologies like this moving into the workplace, creating additional stressors. So, I think that there's a bifurcated case there.
The value and the stress relief and the performance enhancements are 100% there, where the technologies are implemented and the humans are brought to the place they need to be, from an upscaling perspective. But, it is very much a challenge across many companies, across many even governmental organizations, where these technologies are starting to become present, the workforce adapting to them is still a little bit of a stressor.
Andrew Donaher (15:02)
Yeah, that's totally fair.
Victor Foulk (15:04)
I think that kind gives some of the perspective that we're bringing to the market around the things that organizations need to be thinking about right now, that we are augmenting a human workforce, and that human workforce is going to need to be transformed as well, that the human and the technology ecosystems have to come together.
And both are going to be different than what they are today in that unified state. And I think, organizations need to be thinking about that human change management aspect just as much as they're thinking about responsibly and safely deploying these technologies at speed.
But from your perspective, what do you think are the top three things that organizations and agencies should be thinking about right now to prepare for agentic supply chain ecosystems of tomorrow?
Andrew Donaher (15:46)
Number one is move. Start in a controlled way so that your organization can understand. You can never underestimate a human's hatred of change. It's one of my life models in work and you have to figure out where your organization is in that change acceptance process.
And so, when you initiate these types of things, that's going to give you an understanding of how willing or how optimistic your organization is for that change. Can they take it on board and at what pace? That's number one. Start.
Number two is mitigate your risk. Look at areas of opportunity for you and your organization, whether it's with back-office workforce support, whether it's on intermodal optimization, whether it's on commercial incentive optimization. Any of those areas across your supply chain are critical. Figure out the right one for you.
Start small but get to production. Do not do a POC in the back office. That's not going to go anywhere. You have to get to production to truly understand how is this going to benefit the organization? The world is littered with tests in the basement. It's of no value. We have to get to production and help our clients realize a little bit of that value initially. And then from there, expand and expand the pace. No more science experiments in the basement.
Let's get to production and let's help our clients realize this value, and that really helps us to understand too where those change management challenges could be. That's my experience.
Victor Foulk (17:16)
I love it. Start now, start small, move. This technology is real. Let's do real things with it.
Andy, is there anything else that you want to cover?
Andrew Donaher (17:26)
No, I was going to actually just ask a question outside this. I'm super interested, what would be your summary? What would be your piece of advice as we close out here in the last couple of seconds?
Victor Foulk (17:39)
What he said.
Andrew Donaher (17:39)
Ha!
Victor Foulk (17:40)
I really can't, I cannot agree with you more.
This is not a technology still in science experiment realm. This is commercially available, broadly-available technology that it’s ready to transform missions today. If you are not already deeply embracing the technology and preparing for your agentic AI future, well, the best time to do it was yesterday. Second best time is today.
Hey, Andy. This has been a great discussion. We have tons of exciting insights into the future of evolving global supply chains. They're not just chains but truly integrated in complex ecosystems. And we know, agentic AI is the new engine of intelligent supply chains.
And I think it's important to also note that, speed without trust is risk, and responsible design is critical as outlined in a variety of the use cases we've discussed today. And with that, my friend, thank you for joining.
And for you listening, thank you for joining us. Check out our podcast archives and be on the lookout for our next exciting episode.
Andrew Donaher (18:38)
Thank you, Victor.
