AI Agents for Retail

Hello, and welcome, everyone. Thank you for joining us today. We're excited to talk with you about how retailers are using AI agents to support their store staff as well as HQ employees. And part of this, helping those same employees get better information and more consistent information around inventory, orders, pricing, and help automate portions of business processes. My name is Cameron Caudill, and I'm a senior solutions manager at Sunrise Technologies. Sunrise Technologies is a global, award winning Microsoft partner that specializes in retail and consumer packaged goods industries. Sunrise has completed over four hundred and eighty Dynamics three sixty five implementations and has rolled out Dynamics three sixty five to over thirty four hundred retail stores. I am very pleased to have Quintin Hunkin from Microsoft join me today. Quintin, would you like to introduce yourself? Thank you, Cam, and thank you everyone for joining us. My name is Quintin Hunkin, and I am a senior solution engineer here at Microsoft focused on AI business processes. What that means is that I work closely with customers across various industries to prove the value of our product suites such as Dynamics three sixty five, the Power Platform, and of course, Copilot. And I'm here today because I have a deep passion for helping retailers of all sizes. And what's really exciting is that we're at a point where AI isn't something that's just happening around retailers. It's something that's happening inside of their business processes. And today, we're gonna walk through what that looks like in a very practical, hands on way. As we go through today's content, I would encourage you to use the live chat feature to get your question answered by our team in real time. Quentin, a question for you. In today's ever changing technology landscape and where we seemingly have a new model or a new AI solution released every other day, How are you seeing retailers react? In conversations that I've had, retailers really want to understand how they can make the most of their Microsoft investment and ecosystem and deliver value to their teams and help change the retail experience. Absolutely, Cam. We're in a moment of rapid transformation. Customers expect real time answers, where associates need tools to reduce friction, and back office teams are under pressure to operate leaner as as well as smarter. And at Microsoft, we're investing into this underlying AI architecture. Not just standalone features, but a cohesive fabric across all of the Microsoft cloud. And that's gonna include things like first party agents that are gonna be built directly into Dynamics three sixty five, our intelligence suite of CRM and ERP applications, Copilot Studio for that custom agent development. We have MCP servers, or model context protocol servers, that are gonna enable that secure, that governed access to business logic. And then we have orchestrated AI workflows that are gonna allow agents to reason, to take action, and execute work across multiple systems. This combination is really going to give retailers the flexibility to use pre built agents, or create specialized agents that are going to reflect own custom business processes. We're moving beyond just AI as a chat interface, and into AI as a full participant across retail operations. Those are excellent points. A few things that that I wanna make note of is that when I've been working with our customers on deploying agentic solutions, the time to value can really be decreased when there is a solid and sound data and structural foundation behind it, as well as when we the customer has a vision for how the agentic solution could work. When you have all three of those things, the time to value and the the time to get the agentic solution implemented really decreases. For example, one of the customer use cases that we're going to talk through today, when we deployed this solution, we went from sort of the initial conversations through design and development in less than two weeks with part time resources to get the customer in a phase where they were then able to go in and start testing. And depending on the the complexity and the scope of the agentic solution, that could even decrease even further. But, again, not all agentic solutions have to be custom. At Sunrise, we are creating Commerce Companion. Commerce Companion is a collection of agents that are aimed to work with and help the, frontline workers in a retail store. Answer those common questions like, what's the price of this item? What's the purchase history for this specific customer? What's my on hand inventory, or when is the next transfer order for this item coming in? All within natural language, all deployed right there for the store employee to make use of. Alright. I think our viewers are ready to see what this actually looks like. Let's jump into these agentic capabilities that are within the Microsoft ecosystem, explore how Sunrise has extended these in tools such as the Commerce Companion, and walk through some real world customer examples. We're also going to be sharing best practices to help you get started in your own organization. I agree. Let's get started. Retailers today are under more pressure than ever. High employee turnover, increasing expectations for store associates, and not to mention the rising cost for labor and products. At the same time, customers walk into stores more informed than ever before and with information readily available at their fingertips. That means frontline employees need to deliver clear, accurate, and fast information to meet the demands of today's retail environment. At Sunrise, we set out to create a solution that supports frontline workers with their day to day task in the store. That's why we built Commerce Companion. Commerce Companion is a collection of intelligent agents that work together to perform a wide range of tasks. It's built on an orchestration model where each sub agent is designed for a specific purpose. Together, these agents give employees direct access to critical information from Dynamics three sixty five and internal knowledge sources, including standard operating procedures and other internal documentation. Each agent connects directly with the appropriate data source to ensure information is accurate and up to date. Commerce Companion is built with an MCP server, allowing it to integrate with the agent platform of your organization's choosing. But when deployed with Copilot Studio, Commerce Companion can be delivered directly in Teams, providing a familiar and accessible interface for organizations around the world. You can see with a single chat, Commerce Companion provides answers to questions ranging from store procedures to customer purchase history to product information. For retailers, the benefits are clear: Answering everyday store questions, assisting employees with task, and creating an additional channel for training and onboarding, all while enabling better, more consistent customer service. Next, let's take a look at game changing features released by Microsoft for Dynamics three sixty five, first party MCP servers. MCP stands for model context protocol, a standardized way for agents and applications to interact with Dynamics three sixty five systems in a secure, governed, and real time manner. Rather than requiring custom integrations for every data source or action, MCP provides a common protocol that unlocks direct access to Dynamics three sixty five business logic and data. This opens agentic opportunities for every form, table, data entity, and data action in the environment. Microsoft has created an MCP server for Dynamics three sixty five Finance and Supply Chain, as well as d three sixty five Commerce, Dataverse, and Analytics. This standardized interface effectively transforms Dynamics three sixty five into a platform that's ready for agentic AI, intelligent agents that don't just read data but act on it, reason with the context, and execute business processes across systems. MCP enables AI agents to discover and navigate Dynamics three sixty five business models. It provides governed, secure access to ERP functions just like a human user would have access to d three sixty five with the right permissions. It also supports real time data and action execution across applications. With MCP, the Dynamics three sixty five environment can be added as a tool to the agent. This means developers spend less time wiring up APIs and workflows and more time building valuable agent experiences. Describing the agent and the task it should perform provides a rapid time to value for organizations using Dynamics three sixty five. The examples displayed showed how agents deployed with the Dynamics three sixty five MCP Server could be used to create customers, inquire about vendor and purchase data, as well as build a cart with Agent at Commerce. The opportunities for how to insert agents into business processes are endless, and we are very excited about the future with these tools in hand. Now that we've reviewed some of the tools available to build your own agents, let's take a look at one of the first party agents in Dynamics three sixty five that enables organizations to get started today. Retailers can often have thousands of SKUs and vendors that may have a wide variety of technical capabilities. Some vendors may leverage EDI while other communication can be very manual. The supplier communications agent is a first party agent built directly to work with Dynamics three sixty five Supply Chain Management, designed to support back office employees responsible for purchase order management and updates. The agent monitors your supplier mailbox and identifies messages that include potential updates to purchase orders, such as quantity changes, shipment delays, or revised delivery dates. The agent can be trained on vendor specific language and communication patterns, allowing it to accurately interpret supplier messages even when formats or terminology vary from vendor to vendor. When a relevant update is detected, the agent surfaces suggested changes directly within Dynamics three sixty five. Back office employees can review the proposed updates in context, validate the information, and approve the changes, all without leaving the d three sixty five experience. Once approved, the updates are automatically applied to the purchase orders ensuring that supply chain data stays aligned with the latest supplier communication. By operating as a first party agent within D365, the supplier communications agent works within existing security roles, business rules, and workflows. The result is a streamlined experience that helps back office employees stay focused on managing supplier relationships and less time combing through emails and updating purchase orders. The supplier communications agent becomes part of the team, executing an important task that pays dividends of having accurate purchasing data within the ERP. Next, we want to discuss two customer stories with you. The first is Jaypour Living. Jaypour Living is the largest designer and distributor of hand knotted rugs. Jaypour Living has created a network of more than forty thousand artisans in seven hundred villages across India, which weave the highest standards of ethics into every touchpoint of the brand. They distribute products through a variety of channels, e com, retail, and wholesale, but they have a large b to b focus with internal and external sales reps. Sunrise and Jaypour Living completed the initial Dynamics three sixty five finance and supply chain management project on April first twenty twenty five. Since going live, Jaypour Living has looked to continue to leverage their Microsoft investment and improve processes to gain efficiencies and automation. When looking into areas which could be improved, customer service was an obvious choice. The sales order agent, which the Japor Living team endearingly named Orty, streamlined the process of a customer getting responses to common sales order questions like what's my order status? What's the tracking number for this order? Are there order holds? Are there any cancellations on this order? What are the open amounts? And more. The project had several goals to drive design and adoption. A few to call out are the removal of data barriers between teams, streamlining the customer experience, and providing the sales team a self-service tool in an easy to use mobile manner in Microsoft Teams. This enables sales reps to interact with data as they're on the go, going into a customer meeting, or to provide a quick answer to a customer email or message that comes in. In a quick demo, you can see how variables are set for the order. The agent provides a summary response based on the order, and users can continue to ask follow-up questions. With the architecture behind the scenes, all order information is obtained in a single call to d three sixty five, and OpenAI prompts are leveraged within the agent to answer employee questions based on the data, company policies, and verbiage mentioned in the OpenAI prompt. At this point, you've heard a lot from me in this webinar. Let's take a minute and hear directly from the Japor Living team. I'm so excited for Orty. That's our new AI customer service agent that we partnered with Sunrise on after our original go live. So my goal as a leader is to dramatically reduce our response times. We wanna go from two to four hours to under an hour. We can't reach that goal if we are bogged down with simple questions like, did this ship? What's the ETA? Oortie gives our reps and field sales team direct access to real time D365 five data from their mobile device. So this frees our team to focus on higher value conversations and faster resolutions for our customers. Reps can just open up Microsoft Teams and get the answers they need with Orty. Sunrise team told us about the Copilot studio, about the agents, and a little bit about their capabilities. So we did a initial project for we called the bot an Ordy, which would give our team members inside Microsoft Teams status of an order because order status is probably the one of the biggest incoming inquiries from a customer service standpoint that we see. It takes up half of the time of our customer service team to answer those questions. And now with a system like that, a salesperson who's on the road can just chat in the order number and then Audi would give it all the information it needs. The speed of the implementation, was just amazed that how quickly it was done. Of course, we've got a lot of good feedback from our sales team. Hey. Can I do this? Can I do that? And the the speed of the change that we're able to do to to that system, I'm just amazed. Right? Because a lot of times what happens is when you do it implement something in IT and if you do wanna do a change, right, then it becomes a full change request process and then, like, okay. We have probably a month's of it to make those changes. This chatbot now, you know, sometimes you're a few hours away to make those changes. You know, we just have to wait, for offline business hours to make that change. But then in our test environment, we are probably making changes, very soon and very quickly, so which is which is amazing, the capability that it offers. Ordi highlights how Jaypour Living is building on its investment in the Microsoft ecosystem to drive continuous innovation. The Jaypour Living team has already taken ownership of Orty and has expanded its capabilities and is looking forward to expanding additional agentic capabilities into other parts of their business and their organization, such as an agent that investigates inbound purchase orders or voyages and provides updates to the appropriate teams. By extending its ERP with practical AI tools, Japor Living is creating scalable, efficient, and intelligent operations. This effort reflects its ongoing commitment to exceptional customer service. Our second story highlights how a global leader in licensed sports apparel partnered with Sunrise to deploy agentic solutions that support their finance and accounts payable teams. The customer began their Dynamics three sixty five Finance and Supply Chain Management rollout with the first go live in twenty twenty and continued to expand across regions. As the global footprint has grown, so has their investment in the Microsoft ecosystem, adding solutions like Invoice Capture and Copilot Studio along the way. Today, we'll look at two agentic solutions they've implemented. The first is an Invoice Capture agent. This agent works alongside Invoice Capture and Dynamics three sixty five to step in when invoice processing errors occur. Instead of requiring manual investigation, the agent interprets the error message, performs research, and automates the follow-up task, such as preparing outreach to the vendor to request a corrected invoice. When an invoice fails, the agent evaluates the error, and if certain criteria are met, pulls additional details from Dynamics three sixty five like the vendor contact information, name, and email address. The agent drafts an email to the vendor requesting a clean copy of the invoice and additional details as to what went wrong. This draft is then saved for review and further refinement by the AP team before being sent. The result is fewer manual touch points for the AP team and significantly reduced error handling time. The second solution is a multi agent accounts payable assistant designed to streamline common vendor and invoice related questions. Similar to the orchestration model we discussed with Commerce Companion, each sub agent has a clearly defined role. These agents communicate seamlessly to one another to deliver real time answers to AP questions, like what's the status of invoice one two three? What are the pending invoices for vendor a b c? And who's the contact? Based on agent relationships and instructions, orchestration occurs to provide the task to the appropriate sub agent and a response through the AP agent. The deep link agent saves time and clicks by providing a deep link URL that allows employees to jump straight to the relevant record in Dynamics three sixty five. No more opening forms and searching and filtering for the specific record. The agent helps the user and saves time. This entire solution is built with Copilot Studio and deployed through Microsoft Teams. Together, these use cases demonstrate how organizations can fully leverage their Microsoft ecosystem investment to build secure, scalable, agentic solutions that deliver real, measurable impact for their teams. I love seeing what customers and partners are achieving with these solutions, but it's important to recognize that getting meaningful, repeatable results from Magentic AI requires planning and a strong foundation. Cam, do you wanna walk us through some of the key recommendations that you have? Yes. Absolutely. Based on our experience of working with customers implementing AI solutions as well as implementing solutions internally at Sunrise, there are some areas of recommendation. Really, there's four that I want to call out. The first is an accessible data foundation. The second is to identify structured, repetitive work. The third is to identify the right task type. And the fourth is to identify an area where we can inject AI into the business process and not cause harm. Basically, we want it to be as cohesive as possible. So we'll we'll work a little bit more, or I'll explain a little bit more about that shortly. Awesome. Let's go through them. Having clean, structured data is vastly important to agentic solutions. And unfortunately for many organizations, this is the barrier to entry to even attempt to deploy an agentic solution. As you saw in the solutions and demos that we outlined today, customers that have a structured data foundation and accessible data and known business processes in those applications are able to quickly deploy agentic solutions and inject those into those known business processes. Alright. Let's talk about structured, repeatable work. When you're looking to bring AI into a business process, the hardest part is often, where do I start? Right? We typically recommend beginning with structured, repeatable work. The tasks that happen every day, that follow a predictable pattern, and carry measurable impact. If you can clearly articulate value, time saved each week, faster responses, fewer manual errors, it becomes much easier to build momentum across your organization. Right? Secure that buy in, and keep the project moving forward instead of moving to that next idea prematurely. At this point, we've all used AI to create draft emails, documents, and AI is very, very good at those things. That is a great use case of AI. Using AI for the right task is very important. Think about doing a home improvement project at home. Right? You need to have the right tool in order to do what you're doing. You wouldn't try to drill a hole with a screwdriver, as an example. Right? So agentic solutions and what we choose to use them for is going to be the same way. We need to use LLMs and agentic solutions for the things that they're good at, like summarizing, drafting, and reasoning over information that already is provided. When something requires more strict rules or advanced calculations, that is not the best use case of AI or an LLM. So we need to keep math compliance checks and rigid business rules outside, allow that calculation to occur, and then bring the results back in and let the agentic model and solution reason over that data and then provide some additional insights. What this will do is it will remove that brittle behavior, keep the very concise calculation logic external, and then allow you to continue on your business process with the AI and agentic solution. Alright. We've talked about this one quite a bit, but let's talk about placing AI into a business process. High impact AI doesn't start with systems, it starts with people doing the work. Look at where different roles lose most of their time. Right? Identify those teams, and then see where are they losing the most of their time across what system. Say, it could be Teams, it could be Outlook, it could be their ERP, or their CRM, it could be support queues, it could be handheld devices. And when you look and map real workflows, patterns are gonna emerge. Right? If we look at sales reps, right, they may be constantly checking multiple systems. What could potentially help them with that? An agnostic system like Teams, and have new copilot surface directly within that application. Or service teams answering variations of multiple questions. What could help them, potentially? Maybe a co pilot agent that's using knowledge retrieval. Or if we go and shift more to the back office, you may have finance teams that are reconciling mismatches. What could help them? Maybe some cross system co pilot workflows. This role centered approach is aligned with industry best practices. Start where that work actually happens. Alright, we've gone over a ton of excellent tips today. One more I'd like to add is to experiment with different AI models, and Microsoft is gonna give you that flexibility to do so, whether that's using models such as OpenAI or Claude or even your own custom models. That's gonna allow you to tailor each agent to the tone, to the reasoning style, and the precision of your business needs. Cam, what guidance would you give teams looking to deploy their first agent? Yeah. So there's four areas that I would recommend in order for an organization to get started and and to look for when getting started. The first is that you need to identify the areas of opportunity. So department by department or maybe cross functionally, if you have a cross functional team that is investigating these things, you want to look within your organization to find areas that are repetitive and have a high impact within the organization. And once you know that, you also want to try to quantify what that impact is so the more broad team can understand, the reason why this is a good candidate for AI and why we're looking to change this business process, and inject some agentic solution into that process. The second thing that I would encourage you to do is to start small and really understand that business process and then expand. So as you're getting started with AI and agentic solutions, you want to, like, be okay with singles and doubles to use a baseball metaphor. Don't try to immediately hit a home run, when you are starting. Home runs are great. I I love home runs. You can see my baseball signed in the background, but not everything has to be a home run. A lot of singles and doubles add up, especially across a business process that really could use some help. So you want to start small, get buy in, show individuals within your organization that this truly is having an impact, have those be your champions, and then continue to expand. The third is that you want to learn and also use new tools. However, this comes with a caveat because as I talked about earlier, there are new AI tools and models that are released seemingly almost every day. So with this, right, you want to rely on your technology partners around you to help guide you because there's gonna be so much stuff that you're not going to be able to keep up with it necessarily. So leverage that, but also make note in that once you have an actual working solution, if it's working for your organization, you don't have to immediately go and change it and try to use whatever is new. It's always good to go back and optimize as with business processes. Right? We're we're looking to optimize those. We're looking to continue to improve upon them. But at a in a given cadence, right, you don't need to go and look every month to see if it needs to be optimized or updated. You could do this on a six month or yearly cadence. Once you have an AI solution that actually is working for your organization, don't immediately try to just go and use the the new tool that just got released. The last thing that I would want to mention is you really want a solution to meet your needs. Finding a working solution, that meets eighty percent of what you're looking to do, eighty percent of the use cases, and then having the human go back and do the, additional twenty percent is still a win. Try not to over solution to try to get every single use case covered, because when you try to over solution and with AI or with with any tool, really, you can get yourself in a hole. And that could result in not as much buy in of that solution or of the tool that you are trying to deploy. So you want a solution to meet the needs. You want to maintain that end goal in mind of this is the impact that we're trying to make. And if some things still need to be reviewed or done in a different way, that's okay. You're look you're you're still making improvements to your processes. You're still making improvements within the organization and hopefully having some time saving there. And the last thing I would wanna mention is that that solution that you are defining, keep the employee and the end user in mind. You don't want them to have to, like, leave whatever they're doing, and go to a different application and do something else. You want to make that, accessible to the employee. So that those are those are a little bit of my thoughts and and and next steps of what I would recommend for an organization. Thinking about the entire business process is is very valuable. Like, an entire process, really. Considering employees and how they will interact with their tools in their day to day processes can sometimes be overlooked. Absolutely. So as we look to wrap up today's call, first, I want to thank you for your time and for your attendance today. I know that your time is very valuable, and I'm fortunate that you chose to spend it with us. I wanna echo that sentiment. I hope today's discussion helped you see how retailers can use AI to empower frontline associates and back office teams, get a clearer picture of what's available across the Microsoft ecosystem, learn from real world customer examples, and then understand where to begin in your own journey. Please continue to place your questions into the chat in order to continue the conversation, and you can also connect with us using the QR code on your screen. We look forward to chatting further with your organization about Dynamics three sixty five, Copilot Studio, and any of your agentic endeavors within your organization. Thank you.