Episode Categories:
Resources:
Join the discussion: Ask Jordan and Ray questions on Microsoft AI
Upcoming Episodes: Check out the upcoming Everyday AI Livestream lineup
Connect with Jordan Wilson: LinkedIn Profile
Try Our Free AI Prompting Course: Register for our free Prime, Prompt, and Polish AI Course!
Introducing Copilot Studio
In the age of burgeoning artificial intelligence (AI), tools that ease the transition of digital solutions from prototype to production are becoming vital. One platform opening new frontiers in AI application development is Copilot Studio. This innovative solution permits rapid prototyping and application development through natural language input, simplifying the overall development lifecycle.
Unleashing Business Potential with Custom AI Agents
Copilot Studio gives businesses the power and flexibility to create bespoke AI agents tailored to address their specific needs. Whether it's automating mundane business processes such as two-way invoice matching or executing complex autonomous tasks, Copilot Studio delivers compelling solutions quickly and efficiently.
Three Approaches to Leveraging M365 Copilot
Considering different levels of complexity and expertise, professionals can engage M365 Copilot at various levels. For simple tasks and data retrieval, lightweight AI agents prove sufficient. But for more complicated activities, businesses can harness the power of Copilot Studio for custom app development. Advanced developers can then access raw APIs and experiment with frameworks in Azure AI Foundry.
How to Start with AI Adoption
With Copilot Studio, the path to AI adoption starts with a clear use case in mind. Focus on solving a specific problem through rapid prototyping to determine the tool's investment worthiness quickly. This way, a business can quickly identify applications that will yield considerable returns.
Considerations before Deployment
Before taking the plunge into full production deployment, it's crucial to debug and ensure the robustness of the AI agent. Ensuring a seamless, trouble-free implementation will not only protect against unnecessary costs but also foster trust and confidence in the AI solutions deployed.
From Document Creation to Data Retrieval
In the world of AI, evolution is rapid. For instance, what began as an AI use case focused on simplifying document creation and research has seen advancements that enable agents to handle tasks like document retrieval and summarization, benefiting industries and teams across myriad spectrums.
Effectively Aligning Workflow with AI
Key to successful AI adoption is innovating workflows in alignment with AI application. Regular engagement with industry experts and participation in conferences can provide invaluable insights for driving AI integration and application more effectively.
Enjoying the Personal Benefit of AI
Witnessing AI in action brings its benefits to sharp relief. A tool such as Copilot Studio can handle writing, research, and task execution, freeing up valuable time for more mission-critical activities. With SharePoint agents, one can efficiently gather information from a plethora of sources, facilitating faster decision-making.
In summary, AI represents much more than just automated task execution. It's an opportunity to reshape how we interact with technology, both in personal and professional spaces. There's a whole world of possibilities within reach with platforms like Copilot Studio championing the AI application frontier.
Topics Covered in This Episode
1. Copilot Studio Overview
2. Use Cases for Copilot Studio
3. Three Layers of Using M365 Copilot
4. Shift in Work Dynamics
5. SharePoint Agents
6. Data Security and Governance
Podcast Transcript
Jordan Wilson [00:00:17]:
AI agents are everywhere. Literally, if if if you follow artificial intelligence, generative AI, large language models, the conversation over the last couple of months has really shifted toward agentic AI. And when I say AI agents are everywhere, I mean literally where I'm at right now. So coming to you, live from the Microsoft Ignite conference. We got a ton of new information on what Microsoft is rolling out, to its customers worldwide when it comes to agent and I'm, agents. And I'm excited, today to talk with a leader at Microsoft, in the agent's space. So what's going on y'all? Thank you for tuning in. My name is Jordan Wilson, and this is Everyday AI.
Jordan Wilson [00:01:05]:
We are your daily livestream podcast and free daily newsletter helping everyday people not just keep up with generative AI, but how we can all use this information and get ahead to grow our companies and to grow our careers. Before we get started, have to quickly give a shout out to our partners at Microsoft WorkLab. So why should you listen to the WorkLab podcast from Microsoft? Because it's the place to find actionable insights to guide your organization's AI transformation. Tune in to learn how to think more creatively about unlocking the full potential of the technology. That's w o r k l a b. No spaces available wherever you get your podcast. Another place you can go get more than 400 podcasts is our website. So if you are new here, maybe listening on the podcast, thank you for tuning in.
Jordan Wilson [00:01:51]:
As always, make sure to check out the show notes. Go to your everyday ai.com. We will be recapping today's conversation. I already know we're gonna learn a lot in the next 25 ish minutes, so we're gonna be recapping it all in our daily newsletter. So make sure you go sign up for that at your everyday ai.com. Also, we'll have the normal daily news there. Normally, we bring you the AI news live, but because we're at Microsoft Ignite this week, switching it up a little bit different, but we will have all that in the newsletter. Alright.
Jordan Wilson [00:02:18]:
Enough chit chat. It is time to dive into AI agents. And I'm very excited for our guest for today. So, please help me welcome, to the show. We have Ray Smith joining us who is the VP of AI agents at Microsoft. Ray, thank you so much for taking time out of your busy busy ignite schedule to join the show.
Ray Smith [00:02:40]:
No. Welcome, and and glad to be here, Jordan. It's definitely been a busy few days and lots of, as you say, agents everywhere.
Jordan Wilson [00:02:47]:
Oh, man. We're gonna get into that. But, Ray, can you just tell us before we dive in a little too deep, can you tell us, like, what is your role, at Microsoft, you know, working as the VP of AI and agents?
Ray Smith [00:02:58]:
Yeah. So I I'm actually relatively new to Microsoft here about 5 years, but I've been leading over the last year a project aiming with unlocking kind of autonomous agents. So the early day use cases of agents was it was great for summarizing text, you know, asking questions, discovering stuff on the Internet. So we postulated over a year ago that we're like, we think AI will be able to run tasks, we'll be able to orchestrate things, and get ultimately get worked on because that's really where it gets into the high ROI. It's like full a full autonomy. So we kick started a project, which then was released into early access program in May. So it was announced in in build. So these autonomous agent capabilities are build.
Ray Smith [00:03:39]:
And then this week, we're launching them into public access or public preview. So, yeah, there's been lots of GPUs being used over the last couple of days.
Jordan Wilson [00:03:49]:
The GPUs are running. Right? So let's look I mean, we'll rewind and, I do wanna kind of ask you about the the the recent evolution, but let's get into the good stuff. Let's get into the news here, because, at Ignite, we heard a lot in, Sadia's keynote, on Tuesday, as well as, you know, sessions throughout the week. But what is new, with agents at Microsoft? Because there's so much news.
Ray Smith [00:04:14]:
Yeah. There's there's so much. I even chatted with colleagues who are like, did you know about that announcement? Do you know about that announcement? There's so much. So, I mean, foundationally, it's a whole new paradigm. Like, AL is gonna transform how how we work, and it it it's effectively manifesting across a couple of key parts of the products or the AI stack. Starting with m365 Copilot as the the UI for AI, so the interface or the gateway into all these, AI, capabilities or agents, then Copilot Studio announcing, obviously, the autonomous agents, which is kind of where it's the local way you go build these modern day apps in the AI world. So when you think of agents, I just think of apps. They're just the modern day apps.
Ray Smith [00:04:58]:
They're built a little differently. And then lastly, there's the Azure AI Foundry announcements, It's all around that kind of, you know, platform as a service in the AI world, all those, agentic services to be able to support the pro code development to allow, developers fine tune models and really go deeper into the, agentic architecture. So kind of 3 key layers of announcements. There was lots of agents or these new apps announced across, you know, the modern works or across office, like interpreter agents, facilitator agents, and meetings. So lots of really exciting use cases where people can go, I can see how I'm gonna use this every day, or here's some really high ROI examples of how, my business can benefit from them. So so so much dimension in the list, but that's a a rough overview.
Jordan Wilson [00:05:48]:
Yeah. Even the, like, facilitator agent and some of the new, agentic capability in teams, like, I saw that in, like, the ability to, like, speak a different language, like, live in real time, like, my mind is blown. Like, I I I mean, maybe can you even talk about some of those, you know, not low hanging fruit use cases, but for, you know, everyday people, maybe for the nontechnical people, how is their kind of day to day work or how could their day to day work really change, with some of these more specialized agents rolling out across the 365 platforms?
Ray Smith [00:06:24]:
Yeah. There's I I probably put it into 2 different camps. 1 is, like, boosting productivities for humans on their kind of workflow. So bringing all these AI capabilities to drive efficiency or drive better performance. And then there's another camp of autonomous capabilities where it's actually running tasks automatically in the background and maybe coming back to humans. But the the interpreter one example you gave is, like, for for, you know, Star Trek fans out there, the universal translator, you know, immediately pops to mind when you're on these meetings. Being able to, in real time, translate languages is is something that everyone can identify with. Or the meeting facilitator, as was announced, similarly, someone just keeping track of actions, you know, time checks, then making sure the balance of conversation.
Ray Smith [00:07:07]:
So I I would say on the productivity across Copilot pages, across the various agents across Teams and Office, very much driving that productivity for humans. You will see some of the announcements on the autonomous agents from, key announcements with, like, Dow Chemical and partners like Pets at Home, McKinsey, the autonomous agents that they've built should dramatically transform how they run their business. So whether it's like from, you know, revenue protection with Pets at Home, whether it's engagement management from McKinsey or supply chain optimization from Dev Chemical, they're actually fundamentally transforming back office processes as well as maybe how they engage customers and partners. So there's there's so many different use cases. I'd say that's one of the challenges in terms of the autonomous agents because it's effectively a platform people can build and, you know, these custom tailored solutions for their own businesses. But the ones that I think the average person will identify is whether it's across, you know, Outlook, Teams, or the Office products where they're like, That's like that I could use tomorrow. So, they're they're probably more easily identifiable, use cases.
Jordan Wilson [00:08:16]:
Yeah. And those were great use cases. I like, I believe that was, shared. The McKinsey one was shared a a a couple of weeks ago. I think we talked about it on their show. Something like they cut lead time by, like, 90% or something like bananas like that by using autonomous agents. So so maybe, Ray, can you walk us through for maybe those that aren't familiar? Let's just talk about what the heck is Copilot Studio? Yep. And what is the difference between an autonomous agent versus these agents that, you know, kind of are available to, you know, people within different apps.
Jordan Wilson [00:08:49]:
Right? We kinda just talked about the facilitator agent or, you know, kind of pre built agents in Teams, etcetera. So, yeah, walk us through Copilot Studio and Autonomous Agents.
Ray Smith [00:08:58]:
Yeah. So Copilot Studio fundamentally is where you can go extend kinda m 365 Copilot capabilities. I bring a customized flavor to the knowledge that's grounding some of the the skills and tools, or bring in new actions. So it's either to extend them 365 Copilot or where customers can go build their own custom agents from scratch, tailored for their own business and use cases. And it's all kind of low code. So it's kind of like a rapid way to prototype and build these apps in this new world by just describing it because you're just natural language saying, hey. I want to implement the two way invoice matching process. I'm gonna, you know, receive invoices from email, and the POs are gonna sit in SAP.
Ray Smith [00:09:39]:
And it's able to then go through, parse all those instructions, pull in the relevant connectors, and it will effectively bring you through this app building experience in a really kinda low code, way. The beauty of Copilot Studio we hear from customers is it's that it's it's managed, it's it's managed solution, environments, you know, life cycle. So it takes a lot of that, going from prototype to get to production, you know, all the the challenges around that. Because a lot of people building early AI apps, they're saying, I'm playing with this endpoint. I've got a cool demo, but I don't really know how I'm gonna bring it into production. So all that, you know, kind of almost abstracting away some of that complexity around the model management, all the the kind of the guardrails and governance, you know, the security when you're bringing in information, making sure you have the right access to that information rather than, any data loss prevention and so on. So in short, Copilot Studio is just it's a low code app building solution, to build these agents, which honestly is gonna see an explosion. It's you know, we've seen explosions of apps through various technological breakthroughs.
Ray Smith [00:10:49]:
You know, needless to say, handsets are mobile. We're gonna see the same in in this new kind of agentic world.
Jordan Wilson [00:10:57]:
Yeah. So, you know, one question that I've already been been hearing is, you know, some people are like, okay. When do I, you know, go in and and use a, you know, as an example, a a templated agent? When do I just use, you know, an agent within one of these programs? When do I build, you know, kind of an agent with the, you know, no code, low code builder? Right? So what's your suggestion on, you know, as people start to reimagine their workflow. Right? Because I think that's almost you almost have to shift, right, how you think about work in order to really start using agents in the right way. But maybe what is your best advice for people on how to kind of shift that daily workflow and and where to go you know, where do they go to to find the right agent to use?
Ray Smith [00:11:41]:
Yeah. And it's, it's so it's a great, great question, Jordan, because it's you know, the way I kinda tell customers is you get started where you get comfortable in terms of skill set and what you kinda have access to. So I kind of talked about the 3 layers we have with m 3 65 Copilot. Anyone who's kind of, you know, an m 3 65 Copilot customer can go in and build their own agent. So a lightweight agent, as an information marker. And that can be on the the kind of the retrieval, so the basic rag, q and a, and it can shift into, like, automating tasks. So I would say that's where people can go to get started with the the very simple, the very kind of intuitive natural language. Then when you go into building your own custom app within Copilot Studio, you can shift into more, autonomous agents where you can then say add a trigger.
Ray Smith [00:12:32]:
So I want this to run when an email is received. I want it to trigger on an event in, like, SAP, Oracle, Workday, and all these systems, and, applications that we use across the business. So that kinda shifts more towards people who work in IT, who are comfortable with all those connections, and how these tools are built. And then the 3rd layer is really around if you're pro dev and you really wanna get access to the raw APIs, experiment with different frameworks, fine tune models, then you'll probably go to Azure AI Foundry. So each of those support an on ramp depending on your skill set, or, you know, what you feel comfortable with. And but they allow you to kind of, you know, the better together to kinda simulate or bring each of those kind of services and capabilities together. The you know, I think the key part is, what we hear from customers, is they probably sometimes jump in, and they're like, hey. We're building from these API or, you know, endpoints, and we're using these models, and we're fine tuning.
Ray Smith [00:13:33]:
And I was like, they just wanna start with the use case. What problem will it solve for me? And maybe I should start with a rapid prototyping ability, get it very quickly to 80% to go. This is now viable to invest, you know, more time and energy. So it's it's very easy to get a quick solution. I think where the you then spend the time is just debugging on the various scenarios and the examples to make sure it's robust to go into production. So prototyping fast, you know, prioritizing use cases of when and how you can use is probably the biggest learning we see from customers.
Jordan Wilson [00:14:08]:
You know, use cases are everything, I think. Right? Especially when you're, trying to move from from ideation to implementation. You know, and one thing I love, Ray, about being at these conferences is is being able to go around and talk with, you know, smart people like you. But even, you know, in the hub here at Ignite, being able to talk to all the product managers that are helping to build. And, you know, one thing I love asking is, okay, well, how are you using the product that you build? Right? So, you know, there's always the say, you know, you gotta eat your own dog food or drink your own champagne. Right? So how are you actually using, you know, whether it's you wanna talk about the 365, the Copilot Studio, the the the ProDev, right? Those 3 different tiers you laid out. But how are you using autonomous agents or CoPilot Studio in your day to day? Alright. I have to take a quick break to shout out our partners from Microsoft WorkLab.
Jordan Wilson [00:14:58]:
So why should you listen to the WorkLab podcast from Microsoft? It explores the questions business leaders are asking. How can they guide their organization's AI adoption journey? How can AI help them maximize value and create new products and business models? How should they help their teams reskill for this new era of work? And why is it important to be completely transparent about when and how you use AI? Find the answers on WorkLab. That's w o r k l a b. No spaces. Available wherever you get your podcasts. Alright. Back to the interview.
Ray Smith [00:15:32]:
Yeah. It's a it's a great one. So I I think I would have started on the journey similar to, you know, the listeners, which was, hey, need to write a document or I need to do some research. So that was, I think, the use case that everyone latched onto really fast. It says, hey. It helps me write a doc so, you know, we're in the engineering and product development, so we're constantly writing documents around planning and, you know, seeking investment and so on. So that's where most people start. When you then take all of those documents across many teams, then you're probably putting in money SharePoint.
Ray Smith [00:16:04]:
And we, you know, have the ability to have a SharePoint agent. So I can just go in and ask a question from all the planning docs across the whole product group, and I'm easily able to guess dependencies, you know, overlap across teams, ask all those questions, and it's hugely valuable rather than having to chase through various groups of people to get to those answers. So it it kinda starts off very much content summarization and creation into kind of retrieval or rag where you're just trying to get access to information, search the web, do research. Then it moves into actions and past completions. Like, the biggest one for me is if I'm double booked in various meetings, I'm like, I just hit following. Like, it's a new pattern. I don't necessarily decline the meeting. I hit following, and it stays in my calendar, and I'll get the full summary.
Ray Smith [00:16:50]:
I'll get all the actions. So I feel like I'm almost in that meeting when there's callouts, when there's actions. So this is the kind of productivity boost we're seeing from AI. In terms of some of the autonomous agents, I've seen my team build, you know, weird and wonderful agents that would, you know, whether doing, you know, book bashes where they would automatically have an agent that would reach out to the employees, ask for their pizza preferences, order the pizza, and it would arrive and place you via the API just in time for a kind of a get together or a book bash. So I've seen and it remembers what your peep how many slices you'd want, what your preference, aggregated, and then it would place the order for the set number of, pizzas. So, I see lots of interesting use cases. Some of them very custom bespoke, but I think the the productivity boost is, is is is absolutely real in the ROI. I think when it comes to customer examples, so the early access, program that we ran since around May, I've had many days with hairs on the back of the neck going, well, Dental Chemical built a whole series of agents around supply chain optimization and their freight.
Ray Smith [00:18:01]:
So I think seeing the customer use cases where they come at a problem, they apply it, and they they're able to, like, achieve, like, things that even I'm blown away by, daily.
Jordan Wilson [00:18:18]:
And let me give a quick shout out to our partners again from Microsoft WorkLab. So why should you listen to the WorkLab podcast from Microsoft? Well, it explores the questions business leaders are asking. How can they guide their organization's AI adoption journey? How can AI help them maximize value and create new products in business models? How should they help their teams reskill for this new era of work? And why is it important to be completely transparent about when and how you use AI? So find the answers on WorkLab. That's worklab, no spaces, available wherever you get your podcast. What's speaking of, SharePoint, I heard you mentioned SharePoint in there. And, I don't know. Personally, I'm I'm very bullish on these SharePoint agents. Right? Seeing the announcement and seeing some of the information and, you know, being able to speak with some of the product managers, yesterday on SharePoint and just being able to, you know, quickly, you know, select different folders in your SharePoint or the default SharePoint agent.
Jordan Wilson [00:19:27]:
I mean, to me, having a grounded agent that it it takes, like, 0 technical ability Yeah. That can change how we work with our files. Can you walk us through real quick, Ray, you know, a little bit about those SharePoint agents? I think they're fantastic from what I've seen.
Ray Smith [00:19:43]:
Yeah. The I think the agents that we will create from just a click of a button is really gonna be unlocked for, you know, pretty much everyone. So for anyone that has, like, OneDrive or SharePoint and they have folders, I gave the example of our planning documents, for example. But it could be, like, policies, standard operating procedures. It could be, like, lots of things across the business, and you're not necessarily able to connect with the SME. So, you know, in in every walk of life, oh, well, Mary or Jim will be back in on Monday and should be able to answer that question for you. And you're like, well, maybe we should be able to point it to a lot of policies or documentations or sources of information, and everyone should have access to that knowledge as quickly as possible. So in SharePoint, if there's, you know, a site, you just click and say, I wanna create an agent for this, this, you know, file structure, and then quickly grounded on that knowledge, and that information to be able to ask, questions super fast is is is really, it's a game changer.
Jordan Wilson [00:20:43]:
Yeah. I yeah. I mean, just just the ability to in a in a couple of clicks, to be able to create a grounded agent. It's it's mind boggling to me. Like like, you know, podcast audience, I'm just sitting here, like, chuckling and and scratching my head because it's it's pretty fascinating. But one thing I I do wanna talk a little bit more about, Ray, is the data side. Because I think that especially, you know, smaller, companies, medium sized organizations. Right? Maybe they've been on the fence the last couple of quarters on on diving into agents, but maybe they've seen, you know, productivity gains by using other large language models.
Jordan Wilson [00:21:20]:
But those models work with static data. And that's not how, you know, autonomous agents and and copilot studio agents work. They work with your data in real time. Can you tell people a little bit on, kind of the advantages of working with that dynamic data inside of 365 and also, you know, data security? Because I think that's something that people don't always necessarily understand or get right. So walk us through the data side of agents. Yeah.
Ray Smith [00:21:47]:
I think data and actions are probably the 2 of the biggest or most important, things that we will add to an agent. So we're giving it knowledge and we're giving it the ability to to take actions. But focusing specifically on the data side, the biggest concern customers have is, like, will data get leaked? Will people be able to access information that that that they're not supposed to have access to? And if you just build, you know, using large language models and you just drag at files and the maker then shares that with someone else, they're like, well, you didn't have access to those files in the first place. So you're now able to ask questions about documents that you shouldn't have access to. So the beauty of these dynamic sources like SharePoint, not just on the file storage, and across Office, is that it's kind of adheres to all those access, access policies and information labels. So at runtime, if you're asking a question that says, hey. You're Ray Smith. You don't have access to these documents.
Ray Smith [00:22:44]:
It won't pull those documents or those indexes into the grounding, at at run time. So it's making sure that all that, you know, protection, the policies, the governments is really adhered to. And that's that's that's a big deal, not just from files, but also in transactional data. So this is a whole wave that we've done for Fabric and Power BI over the years is how do we make sure even for analytical data, transactional data, and then, obviously, the files themselves within SharePoint that people have access to that information. So that's a that's a kind of really important point. Plus, you don't wanna have to keep changing sources by uploading or changing documents. You just wanna point it out a folder and say, things may change, and for that information as it changes to be to to be current in the, in the agent itself.
Jordan Wilson [00:23:31]:
You know, maybe this is an obvious question, but I think it's worth, you know, bringing up, especially when we talk about data and security. Because I and and maybe this was more of a year ago, but I heard so many, you know, companies say like, oh, yeah. We're not sure about, you know, Copilot or just AI and large language models in general. Yet, these same companies are using obviously OneDrive and and and SharePoint. Right? Is there a difference in in in how companies data is is handled? Right. You don't have to go into the the the granular, but from a high level, is there a big difference, right, for companies that already have all of their data, you know, in the in the Microsoft cloud, so to speak? Is is there a difference for once you start, using Copilot and agents?
Ray Smith [00:24:16]:
Yeah. I think the and, obviously, we support connectors to various systems and tools as well. So but if if it's in the Microsoft ecosystem, you've already kind of invested in those access policies and controls and governance tools that exist across Microsoft. So whether it's Purview for, you know, viewing, you know, your data, across what's fabric. So you bring you're bringing those, data sources and those governance tools, to bear very quickly as you're building these agents. So you can kinda have comfort in that the same thing you've invested in, those same, policies, those, same kind of, access controls will be followed through. And I think that's a key benefit for customers that they can just they can build. And we've been on this journey for probably the last, you know, decade or so with citizen developers and low code, so Power Apps and Power Automate.
Ray Smith [00:25:09]:
So the same way of building these kind of, solutions and the governance around those, is is is similarly followed through for Copilot Studio. So the same, you know, admin center and governance tools, to oversee, you know, what Copilots have been created, what's running, what's it accessing, and so on.
Jordan Wilson [00:25:29]:
Great great answer. I I hope you have a great answer for this one too, even though this one's a a tough one, Ray. But, you know, as we start going, you know, in business businesses at large take this at different, at different paces. But, you know, as we start using autonomous agents, right? Especially inside CoPilot Studio, how does this change the human role, right? With with these autonomous agents that can go through multiple steps, you know, connect to your data. Right? Which to me, it's like a lot of these are what we would spend a lot of manual time as knowledge workers doing. So how does the human role change as autonomous agents become more and more robust?
Ray Smith [00:26:12]:
Yeah. It's definitely a powerful leap forward in automation, but I I think we take a step back. We as humans for the last number of decades in the computer era have always been trying to automate more, whether it's scripting, macros in Excel, building, you know, apps and do software development. We've always been trying to automate more. And as we've automated more, we as humans have kinda shifted to higher value work and, you know, so are you know, we've moved to other skill sets. So with kinda AI and agents, particularly the autonomous agents, there's definitely gonna be lots more automation. So, you know, how does it differ from, you know, RPA, or so robotic process automation, is that we're now strapping a kind of a cognitive reasoning or brain on the top of those kind of skills, functions, tools, or automations, and it's able to do a lot more work. So, for example, you know, something comes in via email, checking policies, reviewing, assessing the the customer criteria before it goes in to executing a transaction.
Ray Smith [00:27:14]:
That can now be done by AI. What I see the shift in for humans is instead of a lot of that kind of frontline, sometimes mundane tasks, humans will shift to then managing these agents that are running these processes. So I would say agents are never gonna be perfect. So it's gonna have humans in the loop for exception handling, for overseeing, you know, all the run events, maybe for handling some exceptions if the AI spits it out and says, I I don't think I should proceed here because I don't have enough information. So humans will start to manage more and more of these agents, and it's definitely a skill set we need to embrace. You know, it's people would say, hey. I have 2 young sons that should they learn how to code? I'm like, yeah. But they also need to learn how to use the AI tools to really get the the most out of how they would code, for example.
Ray Smith [00:28:02]:
So I I would see similar shift on how we've seen any automation allowing us to do more and, other higher order tasks. And, obviously, AI agents is a big unlock for automation. And we as humans will shift to managing a lot of agents.
Jordan Wilson [00:28:18]:
Yeah. That's I I said that, like, a year or 2 ago, and I think people looked at me like I was crazy. But I said, you know, in the future, in the same way that we just interact with different, you know, softwares on the Internet. Right? Like Outlook or Word, etcetera. Might the future of work and I'm not gonna ask you to, you know, make a bold claim, but might the future of work in the next couple of years just be knowledge workers, really just interacting almost all day with with different agents? Is that is that a wild thought, or is that maybe where we're headed?
Ray Smith [00:28:50]:
No. I think that's like, at least my personal opinion is that's where we're headed. Like, I think we will all choose a kind of a personal assistance that was heavily personalized for me. So whether that's in my personal life, and it may even be, like, bring your own assistant to work that interfaces with your kind of work assistance, and then it will tap into, let's say, for example, m 365 Copilot is this gateway into all these agents across your business. So I I would definitely see a world where, there's 100, if not thousands, of agents across the business, and we as knowledge workers are interfacing, via an assistant of some sort. And, you know, even going further, I was like, I think it could be redefine the Internet. Like, you're not necessarily gonna go on and type www. We're gonna just, hey, talk to our assistant and go, I wanna order these shoes, and it will go get that company or brand agents.
Ray Smith [00:29:45]:
It will remember my account, the details, and what shoes I ordered last time, and it will say, yeah. They have them available in stock. So I it'll redefine how we live in our personal lives, but and also how we work, professionally, you know, interacting with these agents.
Jordan Wilson [00:30:01]:
Alright.
Ray Smith [00:30:01]:
That's my little plan.
Jordan Wilson [00:30:03]:
No. I love it. I love it because I'm I'm I'm right there with you. Right? I've always said, like, there's probably gonna be more, more agents than than humans at some point soon. But but, Ray, we've we've talked about a lot. Right? You you kind of walked us through this transition we're going from from, you know, AI being about productivity to now taking on autonomous capabilities. You walked us through kind of like the 3 tiers that, companies and enterprises, can start, dabbling in agents. But as we wrap up today's conversation, what is maybe the most important takeaway that you want people to know when it comes to Copilot Studio, Autonomous Agents, everything that was announced and and where we're going?
Ray Smith [00:30:41]:
Yeah. I think that fundamental takeaway is this is a fundamental step change in how we work. We've talked about business transformation over the years. This is gonna fundamentally transform how we live and work. I think we've talked with that on the on the on the pod today. I think my advice I would give people is just bring focus. Focus on the use cases and get dabbling with these tools and capabilities to start thinking about how you can adopt them, how you can transform, how you work, and, prioritization and a bit of focus on those use cases. One of the early learnings from the early access program was start small.
Ray Smith [00:31:20]:
You know, people would get access to this blank page, and they'd say, I wanna build an agent that's gonna run my whole company or my whole department and or even just an entire process. But it's really a process is made up of many steps, so policy checker, invoice processor. You know, you it may be like a series. So you build these atomic agents to do parts of the process, And over time, you'll stitch those agents together, and it'll then complete the entire process fully autonomously. So focus on the use case, start small to build those skills, build a center of excellence around these capabilities, and prototype fast.
Jordan Wilson [00:31:56]:
So much great information there, Ray. This was a great conversation. Thank you for helping us all make better sense of all of the new capabilities that we got, you know, here both at Microsoft Ignite and everything that's rolling out across Copilot and Copilot Studio. We really appreciate your time and expertise for coming on the show. Thank you so much.
Ray Smith [00:32:17]:
No. Thank you. And, hey, as
Jordan Wilson [00:32:19]:
a reminder y'all, that was a lot. There's so much new, so many new capabilities, with both autonomous agents and everything available in Copilot and Copilot Studio. So if you haven't already, please make sure to go to your everydayai.com. Sign up for that free daily newsletter. We would be recapping everything that Ray just shared with us. So thanks for tuning in. Please join us tomorrow and every day for more everyday AI. Thanks.
Copilot Studio: Autonomous AI agents for everyday people
Ready for the buzzword(s) of 2025? 👇
↳ Autonomous AI Agents.
Ready for your cheatsheet for Autonomous Agents?
↳ We gave it to you today, courtesy of one of the worldwide leaders in the space -- Ray Smith.
Ray is the VP of AI Agents at Microsoft.
(You know.... the current company leading the world's development in .... Autonomous Agents.)
Ray took time out of his busy schedule to join us at the Microsoft Ignite conference.
This is one convo you legit cannot miss.
Don't say we didn't warn ya shorties.
Think about it…. those repetitive tasks you hate? They're about to vanish. If you follow this guide.
Want proof? McKinsey slashed lead follow time by 90%.
Need to process 10,000 invoices by morning?
Tired of copy-pasting data between systems?
Ready to automate your entire department?
Miss this, and you'll spend 2025 watching competitors build AI workforces while you're still copying and pasting from ChatGPT.
Let’s get autonomous y’all. 🦾
1 – The Three-Tier Agent Architecture That Changes Everything 📚
Microsoft just democratized agent creation across three distinct layers, each targeting different technical abilities.
Ray broke down the three tier approach to agentic implementation.
The first hits Microsoft 365 users, enabling instant agent creation from any SharePoint site. Zero code required.
The second unleashes Copilot Studio, where business units can build autonomous agents that trigger on emails, monitor SAP, and execute complex workflows.
The third opens Azure AI Foundry to developers, offering raw API access for custom agent development.
Try This:
Start with your most-accessed SharePoint folder. Create an agent today – it takes two clicks. (Alright, maybe 3.5)
Point it at your company policies. Task it with cross-referencing three months of team communications against those policies. Measure response accuracy against your current manual process.
Target: 85% accuracy in the first week.
2 – Teams Just Became Your Universal AI Command Center ☎️
Picture this: Your German team member speaks. Your Japanese colleague hears it in real-time in their native language.
That's live now.
(Yes, actual mind blown emjoi.)
The new Teams facilitator agent doesn't just track action items. It balances speaking time, flags off-topic discussions, and ensures meeting objectives stay on target.
This isn't about replacing human interaction. It's about injecting AI into every collaborative moment to multiply its impact.
Try This:
Deploy the Teams translator in tomorrow's international meeting.
Measure participation rates from non-native English speakers before and after. Track comprehension through follow-up task completion rates. Set a baseline by recording current international meeting participation metrics tonight.
Who knew… using AI is actually a way to create more empathy between humans.
3 – The SharePoint Security Model That Makes Legal Teams Sleep Better 🔐
Every new SharePoint agent inherits your existing security framework.
No new permissions needed. No additional security reviews required.
It just…. Works.
(Side note, we talked to the SharePoint team at Ignite and there’s some REALLY good things coming.)
These Sharepoint agents read only what humans can access. They respect every label, every policy, every restriction already in place.
When they pass information between departments? They maintain those permissions dynamically. In real-time. At enterprise scale.
Try This:
Select your most sensitive SharePoint site with cross-department access restrictions. Create an agent. Test it with users from different permission levels. Have them request information outside their clearance. Document every response. Verify against your current permission matrix. Complete this security audit by end of week.
Your existing Microsoft security stack just became your AI governance framework. Without adding a single new policy.