Good morning. Good afternoon, everyone. I'm very glad you're all joining me here today for my session. I'm Angela Huffman.I'm a senior director for commercial BI and analytics at DHA Express, which is the largest logistics company in the world. And I also bring a PhD in AI and machine learning long before it was even called AI. In the next fifteen minutes, I would like to share what I have learned from building agentic systems at scale for a global logistics company. I will talk about real deployments, real results. I'm not here to sell a vision. I'm here to share what actually works and what doesn't work. So expect lots of insights and, a bit of experience from that journey.

But before I go into the content, I would also like to hear from you where your organization is standing when it comes to AI adaption. So with that one, let me, publish a call a poll to get more insights about where your organization is standing. So are you actually looking at copilots and, chatbots? Have you started with automated workflows? Have you looked at AI agent agents from end to end process perspective or still in a beginning phase. So please let me know, and then let's look jointly at the results. Giving you a few more seconds to share your view votes. So I see lots of you are already working with Copilot with ChatGPT or other LM solutions. Great to see. I see few few of you have already looked at automation workshops, but also end to end processes with using AI agents.

And there are also, there's one honest voice, which I appreciate about, and I'm not sure where you're standing. And I think all of that is, is right. And, so let's, let me share what you will hear from me today. So today, it's about four pillars. I will start with the great shift, I'm seeing, a blueprint to look at the processes and the work for you. I will talk about safety and trust because for me, that is the core pillar, for every scalable solution. And I will also touch what the workforce evolution, means for the people, the skills, and how you can set up effective, working models. So and in the end, I also have one promise. And we have something concrete for you to do do on Monday morning to take away after today's sessions.

And with that, let me go a bit deeper into the content. So I'm pretty sure yes. All I see in that logic. That is the logic most organization follow. So we start with automation that is making our work faster. It's looking at, all the different elements of a process, looking at how that could be, automated step by step. Then we had a wave of copilot and assistance coming in. So that was making your human work smarter. And now, agenda guy, comes in, and that is really redesigning the entire work entirely. Notice the difference. The first two areas, are about enhancing and optimizing what already exist. The third one is really new area, which is questioning whether it should exist at all.

And there are also studies confirming that, thirty third three times more enterprise apps will be agent powered in the upcoming three years. And we also see 80% of companies use already AI. So the volume isn't the problem, the devs and the app adaption is, is it? And as you can see, most of the organization that, was also the feedback you gave to me are still in the copilot area. So developing AI on top of existing legacy processes. But the core transformation for me is looking at the processes itself. And, with that, let me share what it means for me. So, I'm pretty sure if you work for larger organizations, you have seen processes, like that. So classical, old fashioned ways with tons of different steps, handoffs, approvals, waiting, rework, all of those, coming together. Many, many enterprise processes are designed, like that.

And the reason is, we as humans, we work sequentially. So it's really, reflecting how we we work. And, to be honest, I'm from Germany. So we are really mastering the art of building processes, overcomplicated, very bureaucratic like that. So now if you start with AI, for every single step in there, you won't go faster. You're only, improving on broken process. But to be honest, you can't really automate your way out of this process. So what I would like to do, is to look at the intent. So start with the question, what is it that I want to achieve? So what is the outcome? What is the business perspective, in it? And if you start from there, what is needed for reaching this outcome?

So which steps are really crucial, to have and, which processes would you need to follow to achieve that step. So instead of looking step by step, really going with end to end, what's my attempt and what is the outcome I want to achieve? And then the question isn't really where can I add AI or where can I, add a chatbot on top? But, it's more the question, why does this process exist in its shape? And then, that is where the transformation start. So what I'm asking you, is to really look at redesigning existing processes. And I have a framework for you to, to look at. So in the zero based process redesign, one question is driving everything.

That is, if an agent existed from day one, would we have ever built the process the way it exists today? And I'm pretty sure for a lot of legacy processes that we have today, your answer will be no. So most enterprise processes are really designed for the human world on paper. Sequential steps, reporting, not designed for autonomy. They were not designed for real time signals. They were not designed to self correct. So let's start from zero. And let's start kind of threw away the current process, but start with the intended outcome. So what is it, the business problem, that we want to answer? What is the outcome that must exist? And from there on, build your way. So the second part is then map AI and agentic capabilities. So define what agents can own and solve autonomously versus what needs a human in the loop.

So, afterwards, design and rebuild the entire entire workflow with autonomous decisions, loops, and handoffs. And last one, embed feedback groups. Because in the end, we are building self optimizing, systems from day one, not as an afterthought. So, if you apply those, human decisions become agent decisions. Sequential step become parallel tasks, and your workforce can also run a self optimization, systems. That is the outcome. To be honest, it's not an easy goal because we are building about, we are talking about processes that exist since years and decades. So it's rather a leadership decision. It's not an IT decision. So the core question, wherever you start, is always with agents being available. Would I really build and design it this way, or is there another way around? I have one example for you, and that comes from our core logistics word.

Have you ever experienced international shipments? I'm pretty sure you have. So when we are talking about cross border logist logistics, customs classification is a core piece. It's very menu. It's lots of paperwork. Every country has a specific format, specific requirements, every own way of declaring customs, lots of lookups, checks, delays, and rework of it. So it's really a bottleneck for millions of of shipments in our network. And when we started here, we didn't ask how can we make the customs team faster, But we really asked what does custom, customs classification need from a human perspective at all. So today, and we launched that last week, you can take a photo of a product, and automatically, you can identify the content. We are generating the custom codes. We can generate it depending on the country you're shipping it, and we are validating it. So in the end, it's ready to ship. It's faster.

It's more accurate, and it's at a scale that no human team can, can match. So, it's really starting with what do we need for customs, decoration, making that as fast and user centric, as possible while removing, menu steps not needed. It's in first phases. So we also need to be honest. Solving cross border logistics is not overnight, but that's the vision we are heading to. In the end, you only need a picture of what you're sending and all the rest, all the steps of paperwork, all the required decorations are done, for you. So with that example, I also want to touch what is important to make it work at scale. And for me, that is really the, safe AI component, in it. I've seen lots of vibrant coded, solutions in large scale organizations, and that include a high risk in itself.

So, if you want to apply safety by design, there are existing frameworks to use. And, I'm sharing one example for trust, risk, and security management developed by Gartner. So in this framework, there are six core elements. And let me go through them one by one. So the first one is control. So your agent should operate in a defined, action space with real time policy enforcement. The next one is identity and access. So same for your human users or agent actions need to be authorized based on rules, permissions, and delegations. The third one is the human in the loop. So for sensitive and high or high impact decisions, ensure there's a risk based, escalation, in place and a human oversight in it. The next one is full observability. Because in the end, we need end to end traceability over decisions, actions, and outcomes.

So every single step, every decision made by the agents need to be logged and monitored. The next one is data governance. And that is important, and this is building up on existing data governance. Same for you, for human users as well. So controlled access depending on the roads to data, to content, to really prevent leakage and misuse of the data. And the last one is having a continuous red team in place. So doing the stress test under edge conditions to ensure you can always step in. So with this framework, all the elements, to ensure trust, risk, and security management, are here. And for me, overall, SafeAI is not a feature. It's really the foundation on the ground that is needed to convince both your CFO and also your CSO to approve what you build on scale.

So it's nothing to build afterwards. It's really to be considered right from the start in the design of your application. Now going over to the next part, the skill evolution. So what does it mean for all the people here in this group? I'm convinced there's a shift in the job descriptions. Few years ago, everyone was talking about prompt engineers. Is that any longer, required with LMS becoming smarter and smarter? It's shifting towards becoming agent architect. Same with IT operators shifting towards orchestrating AI solutions. Compliance lead, transitioning towards, owning the guardrails for AI applications. So with all those roles, evolution, for me, I see coding skills are getting less important. It's really about having domain expertise, knowing the business environments, being able to think in systems, and build on scale.

There are two elements if you want to master that transition, which for me are important to have. So first of all, adaption is not a technical question. I'm always seeing lots of IT teams focusing on the right platforms, the tool, the algorithm. But in the end, 70% of transformations are people, processes, and culture. So that is the focus you should set in there. And most organizations do it backwards. So, they focus over models, about infra, and then wonder why adaption is done. There's a second element, and that is about the agents when they are entering the workforce. They should be part of your org chart. They need exactly the same process as a Newman employee. So you need onboarding. You need defined rules, risk racing models. You need performance reviews, and also an end of life life cycle to ensure, we have a retirement process in place.

All of that, what we have for human employees are equally important when you look at our digital workflows. So having both in place is really the key to master this transition. And now as announced, as an action for you, and that is your ninety day blueprint. So when you go back and start, logging into your machine on Monday, let's look in the first phase at one of the friction processes you have. And I'm pretty sure each organization has them. So identify the outcome beneath. And then afterwards, look at, redesigning that. So, look at zero based process redesign. Delete everything that is not required. And then afterwards, pilot how that should be set up if you build it with agentic AI in place. Include, the guard rates and observability that I shared before to ensure safe AI from day one.

And with that one, you know, let's all become the architects of, AI adaption. So it's not about the prompt. It's really owning the architecture and the shift that we see coming from AI. Thank you very much for joining today, and I'm very happy to take your question. Let me start with the questions I see in the chat, and thanks to to see, a few of them already popping up. So the first question is about how long this process takes to start development, for the shipping experience use case. And an honest answer, it's not overnight. So from running the first workshops to go through the process with, designing how it's, looking like, that took some weeks to get all the considerations, in place. Then building solution, really ensuring it's coming with a consistent architecture is another one too.

So an honest answer from me is six months for having the first MVP, for concrete process redesign in place, at least when you're talking about high impact, high scale legacy use cases in place. I hope that was answering that question. There's a second one from Anastasia. Thanks a lot for submitting. So you're very interested to learn a broad process redesign with AI and go into this field. So best ways to learn and to get there as a project manager. I think you have quite a lot of skills already. So as a PM, you know the processes you're talking to. You know the business in side. So, I think for me, it's really ensuring you get an understanding of the technology side, the, current capabilities, and be able to consult on, what is feasible. There are a lot of upskilling programs, available. There's a lot of materials, especially for project managers, towards to become more, efficient.

And what I would suggest is that you partner with your IT or tech teams and look at the first use case together. So really running a zero based process redesign workshop jointly, learning how IT is approaching that, focusing on the business outcome, and building a product team, to make that happen. So that is my recommendation for you. More questions, that are coming up. So existing frameworks for observability and data governance. Data governance, for sure, is not a new topic. That is existing for every large company. And, the principles in there when it comes to roles, responsibilities, decision making, and also, capabilities. That is equally true for AI agents. So first of all, look at existing data governance. And then for observability, technically, lots of solutions are now on the market, but it's also ensuring, you have the full insight of what you're actually monitoring. What is there?

Which KPIs are important to review and, what, you should focus on? So really looking also from a business perspective, which KPIs are crucial, and, ensuring that you have a login as well. There's a technical question coming up from, Maria about which technologies we are using. And we are a very large organization. So, to be honest, I would say that probably every solution on the market is tested. However, they are not production ready for us now. So at the moment, we're using, Copilot from Microsoft. We are using Cloud. We are using self built, solutions, and, the platform capabilities are involving. So in the pipeline, there are a lot of pilots ongoing for which are the best technologies. I said for me, it's important to have the business focus.

So, instead of having 20 different platforms in place, really focus on what is needed for the specific use case. And can that be already materialized with the existing ones? And then the last one, more insights into the logistics process. Marian, I would recommend that we, maybe connect outside of that. I can share more about the process in detail, how we set it up, what is important, and the different elements, of it. So I'm very happy to connect and, to share the experience we made. Thank you very much, everyone, and it was a pleasure having you here. Please enjoy the conference. It's very great to have all ambitions women in it, and, let's stay connected. Enjoy the rest of this conference. Bye bye, everyone.