Name: Workday’s Agentic Ecosystem: Exploring the Technology That Will Power The Digital Workforce
Uploaded: 2025-11-05T19:16:11.603Z
Duration: 40 min 35 s
Description: Workday’s Agentic Ecosystem: Exploring the Technology That Will Power The Digital Workforce

Transcript for "Workday’s Agentic Ecosystem: Exploring the Technology That Will Power The Digital Workforce": Thank you for joining to today's webinar. This event is part of our looking forward with Workday webinar series, which was designed to give you insight into how our organization can do more with the Workday platform. I am Scott Erickson from Workday Illuminate product marketing team, and I will be your host today. Before we start, a few housekeeping items. Everyone will receive an email within twenty four hours with a link to the webinar replay. Please ask your questions in the q and a section of the console, and we'll answer them via text or live at the end of the presentation. Otherwise, we will reach out directly with our responses after the event. Joining me today are Griffin Anderson, principal enterprise architect, and Kelly Imo, senior principal product marketing manager. Just a reminder, we are discussing some features, functionalities, and ideas that are forward looking and are subject to change. First first, I wanna show a 3,000 foot view of the Workday platform. While not everything is listed here, we hope you can get an idea of where today's topics reside in the platform. This would be less about our core products and their features and more about the underlying technologies that the Workday platform is built off of, from Illuminate to Workday build and some of the important industry protocols and integrations with agent platforms and identity providers we are supporting. As we take a more narrow view of the agentic technology, our goal is to help start demystify agents. We will do this by explaining the different types of agents and show you that Workday is creating an open ecosystem where a Gentic AI can be built or leveraged in many different ways. We understand the agents won't have a single source or a single home, so you need a platform where agents of all flavors have the ability to connect in, work together, and give the enterprise a holistic view of the activity and value all in a secure manner. We will use this session to discuss the technologies of and around the three core types of agents, Workday agents, customer built slash partner agents, and third party agents, as well as the technologies that bring everything together in a single view. First, the foundation. Workday has been building, leveraging, and refining the Illuminate architecture for years. This is the core foundation of AI in Workday. Second, we will dive into the Workday agents and discuss the robust platform that Workday developers use to build agents called AgentForge. Third, we will move up the stack to extend and custom agents. Customers can leverage Workday Bill to create powerful agents and applications that meet their exact needs and unique requirements, amplifying the value of work the Workday platform can offer our customers. Next, we will discuss third party agents by way of agent interoperability. We will help you understand the two core industry protocols, a to a and MCP, that are quickly becoming the standard for agent interoperability. Lastly, we'll look at how and why all these technologies can come together in a single pane of glass with agent system of record. And with that, I'm gonna hand it off to Griffin Anderson to discuss the core Illuminate architecture. Griffin? Yeah. Perfect. Thanks, Scott. Hey, everyone. As Scott said, my name is Griffin Anderson. I'm the enterprise architect here at Workday and get to share with you our illuminate architecture. So Scott shared this high level view of the Workday platform. What we're gonna do in this first section is drill into that little highlighted box in the illuminate slice here. So that's where our enterprise LLMs and AI technology architecture lives. So a pretty simplistic box on this 30,000 foot view, but there's a lot of really powerful technology, going on there. So let's look at that. So at the core of how we're delivering Illuminate or AI features within Workday is our centralized inference service or CIS. So this allows for, as the name implies, a centralized point for deployment, management, monitoring of our different, AI models. And you can see there's a couple different consumers of this CIS. The first are our customers. Right? So our delivered AI or eliminate features coming from the tenants, those are all hitting the the CIS as we get inference requests from our different eliminate features. We have customers that are building out custom applications with Illuminate infused through extend. Those are coming in through the AI gateway and hitting that CIS as well. But the CIS is also a place where our developers can leverage different models that are available to them. So as product teams and, developers are building out new solutions or experimenting with new solutions, they can leverage that CIS to, choose the best model and the best approach for their features without necessarily being AI experts. We have, a lot of AI experts within our product teams at Workday, but through this approach, you don't necessarily have to be an AI expert to start building out AI features. So different consumers of that CIS. And then as we move to the right on this slide, you can see we have different approaches to our models as well. So if we start at the top, Workday's been delivering Illuminate features for many years. Our road map has accelerated over the last few years like like most, enterprise companies, with the advent of the LLMs, but we've been in the space for a long time. So we have a lot of kind of traditional ML features. So machine learning features that do things like anomaly detection or categorization, regression testing. So think things like expense protect, journal insights, payroll and time anomalies. All these are using traditional ML models that that we create and deliver to the Workday application. And then another applicate or another approach that we may use is something called retrieval augmented generation. So this is a really powerful and secure way to leverage an LLM and pass it additional context that is that is relevant to the request at run time. So a great example of this is our upcoming self-service agent. That agent is using policy documents, knowledge based articles that are stored in a vector database. And then when a user asks a question, the, we're gonna find the relevant policy documents, relevant knowledge articles in that vector database, and pass those to the LLM with the original prompt as additional context. So at the point of run time, you can share that information with the LLM to get a tailored response that is very customer specific without, training that model on any customer information. So it provides a very secure way to pass that context and be very, very tailored to your specific organization in a very secure way. And then, ultimately, at the end of the day, regardless of the approach we're using, we share these models to a model catalog, and that model catalog may be, patchy packaging up that model as a Kubernetes cluster and putting it in our architecture, or we may leverage, third party models as well that are, hosted on AWS or GCP. One last comment here before we move to the next slide. If we would have looked at this slide just a few months ago, there would have been kind of a fourth fourth category here. So as we all know, this space isn't moving incredibly quickly. We used to have a fourth kind of bucket where, we were doing a lot of custom models or fine tuning of foundational models. That was a good approach, but what we found is the accuracy and reliability of the foundational models was starting to converge with what we were doing from a a fine tuning approach. So from a scale, benefit of scale, benefit of cost, we were able to eliminate some of that fine tuning and just leverage the foundational model as it is. Let's look at the next slide here. So let's talk for a moment about data contributions. So as you may know, the AI fact sheets, as well as when you opt in to eliminate features, share exactly what, data is contributed to training, if any. Right? So this is very this is very feature specific. But if there's data shared with the, shared with the training environments, and this is an automated process, so there's jobs that run on a regular scheduled basis to pull that data from your tenant. And then there's two paths that that data could follow. If we follow the top line of the slide here, that's where we have some human intervention. So this goes to our machine learning development environment, and we have data scientists working in that environment to massage or transform that data. And then they're leveraging that to train our different models. As they update those models, those go into the model the model store, model deployment, and then ultimately to that inference, environment. The other path that this may take oops. Let's go back one more. The other path that this may take is an automated path where that, customer contributed data is fully automated, and it goes to an automatic, an automated fine tuning environment where those models are trained and tuned in a fully automated fashion. There's no human involvement there, and then, follow that same path. And then the third is the third party model. So you see those in the lower right corner of this slide. There's models for things like language translation that we're leveraging in our knowledge article translation or OCR that we're using in supplier invoices or expenses. We're not gonna go create those models. Those models do not need customer data to train. There's third party subprocessors that are do that doing that extremely well. So we use Google for both of those examples that I shared. Those models, you'll see there is no customer data, going into training of those models. Those third party subprocessors are responsible for training their those models on their data. We're just sending that data at run time through that that, API to get the output there. So there's no training of those third party models using customer data. So last thing I wanna share before I pass things back over to Scott is where is all this happening? So where is this infrastructure actually residing? So the the rectangles here are our, tenant environments, public and private cloud. And then our machine learning development environments there, the blue circles you can see are very closely aligned to those tenant environments globally. That provides us low latency. Right? We want those to be close by to our tenant environments. And then we have two main machine learning development environments. You can see those with the red circles here. So those are in Virginia and The United States and then Frankfurt, Germany. We chose Germany very intentionally because of some strict data residency requirements in Germany that that allows us to, conform to. But because Germany is also in the EU, it also, applies to the days of residency requirements of the broader EU. And then the kind of light pink, circles here are those automated training environments where there's no human interaction. We have those pipelines fully automated to deliver those features, and you can see we have those set up around the globe as well. This allows us to, again, have data residency requirements. So an example is in Singapore, the models in those automated, training environments are totally isolated. There's no data sharing with other automated training environments, and that allows us to kind of isolate those models regionally to meet those residency, restrictions there. So I'm gonna pass things back over to Scott. Hopefully, that was helpful to hear a little bit about our our Illuminate AI architecture, and we're gonna get into AgentForge next. Alright. Thanks, Griffin. So, yeah, Workday agents built on AgentForge. What is AgentForge? AgentForge is Workday's internal agent building platform. It brings industry leading open source platform called Langchain into Workday to allow agentic AI development on top of our Illuminate architecture and the wider Workday platform. Lang chain includes a number of tools, Lang Smith for observability, as well as LangGraph, the runtime environment, the agent builder, and debugger. Using Lang Chan for the agent forge platform gives Workday value in three key ways. Workday is able to incorporate the full Illuminate architecture and the whole Workday platform into AgenTic development seamlessly so that agents can be built across Workday. Two, as LineCham further develops their platform and tools, Workday can take advantage of those advancements. And three, it gives agent developers a strong framework to build agents and also allows them to take advantage of already developed agent agents, resources, and tools inside of AgentForge to expand the capabilities of their own agent. It is a true force multiplier when it comes to agentic development. The agent for platform agent forge platform gives Workday a strong foundation for agentic development and ensures Workday agents are powerful, comprehensive, and usable across the entire Workday platform. But as a customer, that's only part of the story as this technology, the Lang Chan platform, isn't only used for agent forge. This is extended into Workday's upcoming low code flow wise agent builder, which we speak about in the next section of this session. But first, it would be good to clarify a question that is often asked. What is an agent? There are really three progressions of technology at play to get to agents, each progression building off the foundation provided by the previous. First, you have ML technology. This is the process of using large data model sets to find patterns and make predictions. Think of it as image recognition, OCR, scanning receipts, finding anomalies in financial reporting, processing expense reports. Each ML model is going to be somewhat unique in its use case, and it will learn and refine itself as it's used. Eventually, ML technology led to natural language processing, NLP, which is a subset of AI. So next, you have AI technology. Think of that as a chatbot. It's usually associated with an LOM taking a question or a prompt and generating an answer for that user. It can leverage supplied context like rag, and it can reason to some degree, though it's largely a probabilistic outcome. It looks to produce the most likely text in response to a question, and today's found foundational models do a pretty good job with this. Lastly, you have agents. Agents you can view agents as a deterministic workflow. They use LLMs, sometimes more than one model, but they really do more than just take an input and give an output based on probability. They leverage additional context through tools, reasoning about what the next step in the workflow should be, and in order to reach a defined goal. The main upgrade from traditional workflows is that AI agents can be used can use the advanced capabilities of LLMs at different points of the workflow, whether it's to generate code, reason about domain specific problems, or perform information retrieval or sophisticated search. They are more deterministic in their process. They are not trying to find the most likely outcome, but rather the most correct outcome. The LLM on its own isn't going to be able to do this for more complex challenges. It requires an iteration of steps using the LLM's reasoning along with external tools and resources, building additional context on its own. Of course, as we build agents, one of the most important aspects is security, as in some cases, they can take actions across the system in in a way that requires robust security frameworks. This is done through a multilayered approach. Every agent is assigned a unique nonhuman agent system user, ASU, account for its operations. This provides a specific identity for precise auditing, authentication, and access management. ASU accounts are created automatically upon agent registration inside the agent system of record, and the account cannot be used for human UI sessions. The agent's capabilities are explicitly defined by its skills, not by a broad security group. Each skill is restricted to a preapproved list of tasks, like specific API calls, preventing the agent from performing any action outside of its stated purpose. A new agent admin role determines which employee security groups are allowed to interact with a specific agent. This ensures that only authorized users can invoke an agent similar to how domain security policies work for people. Delegation agents act on behalf of the user that are constrained by the user's own permissions, preventing privilege escalation. Ambient or background agents have their own permissions but cannot be directly invoked by users, only by secure automated system processes. All network calls to and from an agent must pass through the agent gateway. This gateway serves as a central point for enforcing security, tracking all agent activity, and managing authentication. All agent actions are recorded in detailed audit trails. Crucially, these logs capture both the agent's ASU identity and the identity of the human user who initiated the request, providing full accountability. Every Workday delivered and third party agent is required to undergo a formal review by Workday security team before it can be deployed. Now that we've dug into the foundation of Illuminate and Workday agents, I will hand it off to Kelly Imo to discuss how customers can build for themselves on the Workday platform. Thank you, Scott, and welcome, everyone. I'm excited to dive into how you're going to actually be able to build agents on the Workday platform. But before I do that, I wanna begin with a broader view. So at rising North America in September, we announced Workday Build. Workday Build is not a specific product or SKU. It's a comprehensive program and strategy designed to open the Workday platform to harness the power of Workday's applications, Workday's data, and their AI for our developer community to build unique apps, orchestrations, and now agents on Workday. So Workday Build is that developer platform, and it's supporting a growing developer community and a robust partner ecosystem. As such, Workday Build is designed to unify Workday's developers' tools and programs. You've probably already heard of Workday Extend. We have over 1,300 customers that have built over 3,000 applications, unique applications using Workday Extend. So we're expanding the concept of Workday build now with both extend apps and orchestrations and agents. And that's what I'm gonna dive into next is our offering to be able to build agents on the Workday platform, and that's Workday FlowWise agent builder. It's coming in the 2026, and it will enable you to build, deploy, and manage custom agents directly within Workday, From simple, conversational experiences to complex workflows, your agents will be secure, context aware, and enterprise ready. Let me go a little bit deeper on Workday Flowey's agent builder. So you may recall that back in the August, we acquired Flowise. And we brought we acquired Flowise to bring in a proven, innovative, low code visual builder for agents and an open source technology platform, which Scott referenced, based on the Lang chain foundation, really designed for developers to create this wide array of agents from simple to complex. And now what we're doing is we're kicking off the work to integrate FlowWise, the FlowWise technology into the Workday platform. We're going to be adding it into the core of Workday build within our developer site to give our customers and partners enterprise grade tooling that leverages the rich context, data, and governance of Workday. Workday FlowWise agent builder is going to become a key part of the build experience, and it will be part of Workday Extend Professional. So now your developers will be able to build both custom apps and agents that work together and deliver transformative HCM and finance experiences while securely interacting with your Workday data. But before I go into more details on what it looks like and what is our plan to integrate Flowise into Workday build, I wanna first share and explore the core innovations of Flowize itself for developers and IT leaders. So let's explore what's so innovative about the Flowize technology. First of all, let's take the perspective and the agent developer. There's three areas that stand out. First is Flowize is a visual builder for building agents. It provides distinct visual builders that allow developers to create various AI systems, as I mentioned before, from simple conversational assistance and single agent chatbots all the way to complex multi agent systems and workflow orchestrations. Your developers will build agents through this low code experience, and those agents can incorporate flows, simple and complex, and, and even contain complex logic like routing and looping. It also includes support for developing within a rich AI ecosystem. So this is through integrations and modularity built into Flowise. Flowise supports modular building blocks and and integrates with over a 100 tools, including various LLMs, embeddings, vector databases, and external services, enabling users to connect to those diverse data sources and extend the functionality of their agents. Developers will be able to build agents that connect to hundreds of data sources and tools and build AI workflows that are deeply integrated across systems, data, and other agents. These connections are built are delivered as templates, and that reduces the time to build for a developer. Lastly, Flowise supports developer tools, APIs, SDKs, such as SDKs for TypeScript and Python, and embedded chat widgets. So now developers can bring their custom built AI applications and agents into existing systems. It also supports self hosting and enterprise grade deployments. So through the API SDK and widget support, developers can build agents that surface in chat and web pages, agentic experience platforms like Google agent space, and collaboration apps like Slack and Teams, just to name a few. Let's also look at some of the powerful capabilities of the Flowise technology that resonate with IT with IT leaders. And I'm gonna explore some features for control, extensibility, and scalability. First of all, from a control perspective, developer teams that are using Flowize have tools built into Flowize that give them fine grained control over what they build. So with Floweyes, developers can build and controls that ensure agents behave in the ways that ensure they will perform, that they'll utilize resources effectively in line with key policies across agent to agent interactions, and very important, ensure human in the loop behavior that's critical with agents. I mentioned extensibility before, but from an IT perspective, the broad extensibility support for developers enables developers to leverage these, over a 100 templated connections that are already supported, and that really increases productivity for the developer. And lastly, Flowise is designed for agent scalability. Flowise agents can be deployed to support multitenancy and distributed queuing, ensure performance under load, and support secure deployments with SSO and role based access control built right in. Lastly, I wanna touch on something really critical to Workday and to all of you and and really build at the core of Flowise, and that's support for responsible agent development. First and foremost, agent interactions need to have human in the loop. Flowise is designed to build human touch points directly in agentic flows. This enables steps such as human approval, review steps, and the ability to build feedback into the agentic flow. Second, observability is built directly in the Flowise platform. So developers and DevOps leads can use third party application performance management or monitoring tools to monitor Flowise agent step by step instructions and ensure high performance and efficient use of resources at run time. Third, Flowise includes support for event hooks for third party APM tools to monitor specific metrics that are critical here, such as latency characteristics of agents in production, performance of scale, or cost of agents running in production, such as token utilization and ongoing quality of the output of those agents. So these are core capabilities of the Flowways technology that ensure that agents continue to meet enterprise expectations, perform at scale, behave responsibly, and meet cost and ROI profiles throughout their life cycle. Now I wanna take it a step further and share with you our vision and direction that we're going to bring this amazing technology into Workday build, into the Workday platform, and deliver the most secure and trusted foundation for building custom agents. Agents built with Flowways agent builder will provide a key element in our Workday agentic technology stack supporting our agent ecosystem. Agents you build with Workday Flowways agent builder will be secure, managed, and interact in a trusted way with Workday platform capabilities. Agents built with Workday Flowways agent builder will be registered in the Workday agent system of record, which was announced earlier this year. So you'll get visibility visibility and analytics on their behavior, their performance, and their ROI in the Workday context. In addition, agents built with Workday Flowways agent builder will interact securely with core Workday platform elements such as Workday APIs, and they can also surface and become part of Workday extend apps and will be able to surface in the Workday user experience. A key part of making this happen is the Workday agent gateway, which was announced in June. It and the Workday agent gateway is powered by a universal language for AI called the model context protocol or MCP. You can think of MCP as a secure way for your agents to talk to all the other systems and tools that you're already using, and you'll hear more about that from Scott. And now we're taking that further. Over the next year, we will be adding hundreds of new APIs designed to power agent interactions and standardizing them to speak the same language. So for our HR, finance, and IT leaders, this means a more reliable and secure flow of your most critical data, and it means a unified foundation for your agentic interaction, saving time, cutting costs, and increasing readiness. Now let me give you a little glimpse of what this is going going to look like. Building an agent with Workday Flowize agent builder that interacts with the core Workday platform. Let's see how agent building will surface as part of Workday build when we release this capability in the 2026. First, you'll build custom agents via the Workday Flowize agent builder low code tools that will be integrated within the Workday build developer experience. This will enable you to leverage integrated Workday platform, tools, and templates to quickly stand up an agent flow. Once you've built that core flow, you can then configure the agent behavior using the Flowise agent builder low code canvas. Flowise agent builder will craft the agentic workflow from your instructions, and you can then select the tools that will support the agentic behavior you're looking for using secure, trusted, prebuilt connectors. As I mentioned earlier, the Workday Flowwyze agent builder low code building experience will enable you to build simple conversational AI interactions or more complex flows like you see here that include multi agent interactions, and MCP is core to support this capability as well. If you're building a multi agent system, Workday Flowwyze agent builder will provide integrations to seamlessly access and connect to other agents you have registered in your agent system of record. Here, you see an employee recognition agent flow that will connect to an existing Workday agent to ensure a spot bonus is applied for a worker. Spoiler alert, it will be the performance agent. With Workday FlowEdge agent builder, you'll also be able to build agent flows that connect the agents you're building with external systems. Here you see once the spot bonus is added, a message can be sent to a worker or manager via Microsoft Teams. This could also be Slack or the Workday UI. So what I shared at a high level is a view on how the power of the Flowwyze low code capabilities for building agents will be combined and integrated with the Workday platform. For developers who are building HCM and financials agents and apps on Workday, you'll have a powerful integrated developer experience that brings together all the secure Workday platform elements, like MCP, APIs, UX, and Workday extend application development tools and components for building powerful custom agents that will run securely on Workday. So just to bring this home, we are actively building towards this experience, and we expect to bring the Workday Flowwise agent builder capability integrated in Workday build to all of you in the 2026. Until then, I also encourage you to start exploring agent building today with Workday open source offering. Now with that, let me turn the floor back to Scott. Thanks, Kelly. Now I wanna take the discussion to the idea of third party agents. We heard Kelly mentioned MCP. So let's take a look at how customers can bring these agents into Workday with MCP and a two a. First, we need to understand what the problem is with agents. Getting AI agents the data and tools they need is a challenge today. To interact with external data sources, you have multiple APIs and data shapes, APIs that can change without notice or documentation, creating a fragile fragile ecosystem of integrations. It's like asking someone to speak 10 different languages to get a simple job done. It creates a lot of stress and complexity. The same problem exists for agents that want to communicate with other agents. If agents communicate in a different protocol or format, it creates friction for those agents to communicate. So while MCP connects agents to tools, a two a connects agents to other agents. This takes stress out of AI by providing a simple standard way for agents to work with data and talk to each other. That way, development focuses on what matters most, helping your business. So why does all this matter to you aside from being hot buzzwords? Some of the core benefits are real time accuracy. No more feeding your agents old reports or outdated information with these new standards. Your agents can access real time data and provide up to the minute insights. It also provides business context. The agentic solutions we're building understand your unique organizational structure and business rules. This allows it to accurately complete tasks and business processes for you moving from insights to actions. Your agents can actually do things rather than just giving information, and finally, everything is done in a secure by design. The agentic solutions we're building will inherit your existing security models so you can trust that your data and systems are protected. So what does a two a specifically enable for your organization? It allows different systems across your tech stack to communicate with other with each other through a standardized, obfuscated protocol. This communication allows you to take action across your ecosystem, enabling your employees to do so with their within their natural workspace rather than jumping across different applications for the same task. A two a also enables the use of the right agents for the job. Different agents will have different functions and expertise, and a two a allows for finding the most appropriate agent for the task at hand by enabling discovery of agents by agents through standardized agent cards that give details about what an agent can accomplish for you. All of this creates a more connected and efficient ecosystem for your organization, making your employees more effective by allowing them to focus on more strategic priorities. MCP is a game changer for AI integration. A common analogy for MCP is that it's like a USB c plug for interacting with your data. It is a standardized format that AI agents know how to talk with. This simplifies development, reduces complexities, and ensures that your agents always get the data they need in a consistent format. Because MCP is an open standard, any LLM or agent that supports it can easily integrate with our Workday offering. This interoperability allows you to have flexibility to integrate Workday with the best AI tools for your business. And in the initial release of our MCP server, we will use Workday's public APIs as the tools that AI agents can interact with. So how do a two a and MCP work together, and why does that why does implementing both help the bottom line? We talked about the specifics of how they help individually. A two a allows for agents to work across your organization stack, choosing the right agent for the right task at hand, making your workflows less manual and more real time. MCP allows for intelligent use case specific actions, quicker development of agents, and adaptability. Together, they form a powerful partnership. This combination allows your workflows to be more efficient, removing siloed workflows, allowing for workflow execution within Workday. However, it is most natural for your employees. Workday is building both approaches into a scalable and secure and reliable with adaptable automation. Lastly, I wanna discuss how all the different agentic technologies can come together in a single pane of glass with the agent system of record. As we talked about with Workday agents, Flowize agent builders, third party agents securely connecting in through agent gateway via a to a and MCP, we find the agent system of record at the center of these technologies. So what are the benefits of managing agents with the agent system of record? When you look across the key capabilities needed for employees and agents, whether it's human or AI, the needs are somewhat similar, life cycle management, workforce optimization, governance, and compliance. We are focused on keeping humans at the center of this, always, using AI to amplify their potential. The human to AI interaction is important now, and it will be imperative in the future. But what does agent system of record offer that supports a dispersed agent agentic ecosystem? As the trusted system of record for the entire employee life cycle, Workday now extends the centralized control to agents. The agent system of record unifies the agent life cycle into a single tool, providing the total agent control necessary to manage, deploy, audit, and retire agents. Leveraging Workday's security framework and deep integrations with leading identity providers like Okta and Microsoft Entra, Microsoft Entra, the agent system of record addresses compliance risks by ensuring identity and data security are uniformly applied within and across all fragmented data sources, giving leaders the confidence that critical people, financial, and operational data remain safe and compliant. IT and business leaders need a single source of truth to manage, measure, and understand the impact of agentic AI. As the cornerstone of Workday's agentic stack, the agent system of record provides enterprises with the solution to securely and consistently measure the combat the combined effectiveness of AI agents and the human workforce, answering how agents and people work together across your operations and organization. By integrating with major agentic development platforms like Microsoft Copilot, Amazon Bedrock, AgentCore, and Gemini Enterprise, along with industry standard protocols like a two a and MCP, agent system of record offers robust interoperability of agents regardless of where they live. This allows all agents across the enterprise to gain the advantages from agent management, security, and analytics within the agent system of record. Just to review what we talked about today, we explored how Workday Illuminate is built into the core of Workday, allowing AI to be leveraged across the the platform and allowing Workday to iterate forward with the latest AI technologies from a single location, speeding up innovation in a manner that doesn't get in the way. We looked at at the different flavors of agents, how the different flavors of agents are built into the platform, showing the agentic use cases can be met in a variety of different ways. Lastly, we showed how you can bring these agents together in a single location, life cycle management, deep analytics of agent value, shedding light onto realized ROI across your AI investments.