Power Platform Meets AI: Smarter Workflows for M365 Users
Hey, everybody. Welcome back to another episode of Make Others Successful, a podcast where we share insights, stories, and strategies to help you build a better workplace and get more out of your Microsoft three sixty five subscription. I'm Matt. I'm here with Mike and Mikaela, and we're gonna talk everything Power Platform and AI. So this podcast is gonna be a little bit different in that we are gonna have kind of a front half and a back half.
Matt Dressel:The first part, we're gonna talk a little bit about how you might think about a business process and how you might think about automating it or improving it with technology. And then once we kinda talk about that and talk about in the context of Power Platform generically, how you might solve that problem, then we're gonna switch it around and say, Cool. How should we think about solving that with AI? Let's get right into it. Mike, do you want to set us up a little bit with the story about the business and the process that we're trying to automate?
Mike:Sure. So, I'm not going be very specific here. We're going to try to keep it generic. Think of a relatively simple scenario like an IT request process. So maybe a request for the use of some software or a request for a new hardware device or something like that.
Mike:And maybe you're managing that process with a shared mailbox or just emails to the IT team. Like, hey, I wanna use this thing or can you, you know, allow me to install this software on my machine or something like that. And typically stuff like that, especially in a larger organization has to go through some sort of approval process. And then while you're waiting on the approval you have maybe no idea what's going on and then you follow-up with another email and you say, hey, did you get my request? Right?
Mike:Maybe there wasn't wasn't even an acknowledgment. So there's all kinds of problems with that process. And then ultimately once it's like actually approved and then you can go install it, somebody needs to let you know about that and somebody probably ought to document the fact that we approved it and that you installed it, right? And there's all of those other follow on things that need to happen. And so that's a pretty common problem because if you have an organization with any number of people, you're managing all sorts of different things from all different people because they all do different things.
Mike:And it would be a lot better if you could track that in a more formal way. And so there are ways we can do that inside the Power Platform. You're about to say something.
Matt Dressel:Yeah. You talked about the process. Let's talk about the common pain points that people fall into in that business process and how what we might want to solve. Mikaela, do you want to talk about that a little bit?
Michaela:Yeah. So, I think if you're running this out of a mailbox, like you said, Mike, who has access to the mailbox is one of our first problems. We need to make sure everybody who should be in it is in it. We're tracking things based on maybe subject of the email. It's rather disorganized.
Michaela:There's no standard maybe form we need to collect information about all of the, you know, equipment that we're reviewing and its use case maybe. And so, it seems very non standardized, disorganized honestly, if we're doing it out of a mailbox, we don't have a source to go to.
Matt Dressel:So, those scenarios, a lot of the things you talked about are really pain points for the people receiving the requests, right? So, they're like, hey, I get these requests in different formats and I have to deal with that. Or, you know, maybe somebody marked it as red and then I look at it again and don't realize it and it gets lost. So those are pain points on the team that is performing the process. Talk to me a little bit about what those things cause problems with for the end user, the person who's experiencing the process.
Michaela:Yeah, so maybe I'm a user that's submitting a request for something, I might not even know what these folks need to know about this, I might not know where to go to ask for this information and to request something from these folks. So, you know, I I may even not know the words to describe what I need. So if I were to submit something like this, I would want something guided and something helpful that gives me context of what they're looking for from me.
Mike:Okay. So so when I think of the problems, I think of it's could be slow, It could be a black hole, something gets completely lost and I never hear back. I have no idea if anybody received my request in the first place. From a reviewer or processing standpoint in dealing with those requests, let's say you need more information from the end user, right? What's the process for How
Matt Dressel:do send it back? How do you send
Mike:it back? Say, I need more details, right? Or how do you ultimately find a similar request that maybe has been approved In past. In the past. Like, oh, we did this before.
Mike:I know we did this before. But I don't know who we did it for, right? So all sorts of things that could be pain points on both sides of that.
Matt Dressel:And in this one, I think the one that you guys are talking about, it's actually like a multi step approval. So maybe I'm the person that receives it and I do one part of it, now I have to hand it off to somebody else. How do I do that? Do I email that to them to get them to start it? Right.
Matt Dressel:How does that work?
Mike:Which there's all kinds of problems there with expectations about like the users actually understanding the process soup to nuts and following through and doing it the right way, which is a training challenge, right? There's all kinds of problems with that. Yep. Which there are tools now fortunately that you can solve those problems with.
Matt Dressel:And just to kind of give some perspective about what we have seen generically in the industry, a scenario like this is ripe for improvement, right? Like this is something that could be a process that might take a month, two months even, in a very unorganized fashion or even a semi automated fashion. But we've seen those things drop down to under a week Yep. In in regards to being
Mike:just able to
Matt Dressel:by doing some of the things we're gonna talk about. Is that is that right?
Mike:Absolutely. Yeah. And there are variations of along the path to digitizing something like this. So what we're talking about is like stone age. Yeah.
Mike:Like shared mailbox scenario. Ten years ago, you might have been using something like InfoPath and SharePoint workflow if you're in the Microsoft world for something like that. But even that now is being phased out, if not already, it's going to be if you're using it. And so there are better tools for that. And even in that scenario, there were challenges there as well.
Mike:It's kind of an old process, didn't have all of the all of the bells and whistles as it was very integrated. It wasn't wasn't very
Matt Dressel:integrated all on its own. Yep. Okay. So anything else we want to talk about from a setup perspective and and kinda setting the stage for what the business problem is?
Mike:I think the only other thing that I would say is that's a pretty generic problem. We're trying to describe something that's common to a lot of organizations. But if you have some other problem that's maybe a little bit more specific or unique to you, our hope is that this challenge maps onto your problem. Yeah. Yeah.
Mike:We're talking
Matt Dressel:about it in a very specific way to try to help people understand. Because it's hard to talk very generically about these things and and not mention any sort of business case.
Mike:Or it's hard to talk specifically about 200 different things.
Matt Dressel:Yeah. Right? Yep. Yep. Because they're all a little
Mike:bit In a thirty minute podcast.
Matt Dressel:Yeah. Yeah. Okay. Cool. So now, we're gonna move into the other part of kind of this first section, which is let's talk about how you solve this problem.
Matt Dressel:So we know what the business case is or the usage scenario is. I have an IT request process. I get that in. Someone needs to review it. Maybe they need to go out and review terms of service or whatever that might be for the software.
Matt Dressel:They're going to pass it off to somebody for approval. It might be two or three different people. They either approve or reject it, send it back to the submitter, and either they start getting the tool or the software or they
Mike:Or goes to procurement.
Matt Dressel:Or yeah. It goes to procurement, that kind of thing. Right. So that's the process. Let's talk about solutions.
Matt Dressel:So, Mikayla, you want to take point on that?
Michaela:Yeah. So, I if I think just within Microsoft, we have so many different tools we could leverage. And I think that's the fun aspect of all of this is there is some room for creativity. So, say we want to use SharePoint, that's, you you know, we have SharePoint lists, we can add rows in there and create a table that's going to track all of this for us. We could add a status column so we know where things are at, at a glance, we see everything right there.
Michaela:We could use Microsoft Forms to collect the info from folks, and then maybe we stick in a SharePoint list, maybe we're using Azure DevOps. There's so many different ways you could track all of this. My personal favorite, of course, is going to be Dataverse with a model driven app. And so, we can go from something that's free and quick to build all the way up to something that is polished and kind of handles a lot of this process for us and has more automation tied into it, and is maybe a more mature solution than something that is, you know, a little faster to build, but it gets our users where they need to be sooner.
Matt Dressel:So, that's great. There's a lot of free tools out there. I know Mike has this blog, and we have a podcast that talks about choosing between the free tools and the paid tools. Can either of you help me identify how I might make the choice between which one we wanna use? And in my case, in the scenario that we're talking about today, we're talking about something that is mission critical, it's really important to our business, and we get like thousands of these things a year.
Matt Dressel:What should I be choosing? What should I pick?
Mike:I mean, think in that scenario where you get a thousand of them, it's kind of a no brainer and I would encourage people to go back and watch my video. I did a little math lesson there. Generally speaking, the cost of licenses for a Dataverse connected app are worth it if you're if it's mission critical or if you're dealing with, like, thousands of requests. Like, that's kind of a no brainer. There is a cost associated to how many users you have using the app, but that's pretty simple math that I show you how to do in that video.
Matt Dressel:Sweet. Okay. So now that we know the business process that we're dealing with and the technology we want to choose, it's going be Dataverse and some of the paid tools, who wants to talk through how you might use these tools and for what purpose you might kind of fit the different tools that you have at your disposal.
Michaela:Yeah. So, I think that's kind of the beauty of the Power Platform is it integrates so seamlessly into all of these Microsoft products. So, maybe you're communicating in Teams and you need to reference different things that are in Dataverse, you can have a flow that will post something to someone in Teams or you're using approvals within, you know, Microsoft Power Platform and you're sending someone an approval, approve, reject, and you can link that back to a Dataverse row, and then, you know, you can integrate that into Outlook as well, and so it can tie everything together.
Matt Dressel:Okay. So, core is the Dataverse, which is where all your data is going to be. You might do some things with teams to kind of communicate. What about the reporting? I know you guys talked about being one of the benefits is being able to report on the data.
Matt Dressel:What tools and how would that look in this solution?
Mike:Well, one of the keys there is that in this scenario we're talking about Dataverse, which just to point out to everybody, that's a well structured relational data source.
Matt Dressel:Yep.
Mike:So think of traditional like SQL You've Server or something like that, got some rock solid data that you can pull and hopefully it's well structured because you've somebody that knows a little bit about what they're doing and putting those tables together. Then you can connect directly to it with a tool like Power BI and you could publish that Power BI report right up to a workspace that people can access or you could even embed a report from Power BI in a model driven app or in a Canvas app to give users some insight into what's actually going on or to give admin users the ability to, like, track how long it takes for something to go through the process or how many requests are currently in process and how many have been approved and things like that.
Matt Dressel:You mentioned a Canvas app. Why would I use a Canvas app when I can just use a model app? I, you know, if I use a Canvas app, could do it for free. What's the what's the point? What's the value?
Mike:Well, so the Canvas app will give you a little bit more ability to be very prescriptive in how you let your users interact with the data and engage with the data. So you can be that way with a model driven app, but it's a little bit more data centered. And so you're gonna get forms and they're mobile friendly, you can access the model driven app on a mobile device. But the Canvas app is gonna give you, like if you really want a wizard like ability to guide somebody through the process of filling out a form or a request, it'll give you that capability. And it'll have a few other, I guess, you know, they call it Canvas app for a reason.
Mike:It's like you're painting on a canvas, give you some more ability to kinda control the look and feel of that app for your end users. And if you want something that's even more geared toward mobile format, right, a Canvas app is Canvas gonna app. Okay. A better option.
Matt Dressel:So if it's really important for my end users to submit this thing in a mobile app or if I want to have more control, really make the wizard experience nice, Canvas app. Even though I'm still working with Dataverse, Right. Canvas has
Mike:You're still going to need the Dataverse license because they're accessing Dataverse. Sure.
Matt Dressel:But Okay. Okay. Talk to me about the approvals. We haven't talked about that at all. What is this approval process that maybe before I was sending an email to a person, getting an email back, maybe storing it in a Word document, how does that work today?
Michaela:So, you can use the built in standard approvals, is what it's called in Power Platform. And so, you can send someone an approval request where you have maybe a title and some information about it, and you can send them a link to get more info on this request, and then they have a button they can press, accept or reject, you can, you know, customize it as much as you want. And so, it's all built into maybe an automation and a process where then you receive their input, and then you can process that itself. You could also use something like the tasks table within Dataverse to
Matt Dressel:Okay.
Michaela:Do this as well, and they can respond on those too.
Mike:I think on the approvals front, I would add that it's a standard, tracked, secure, documented method. Right? It can't really be tampered with if you do it right as opposed to your own homegrown like, I'm gonna put something in a table or I'm gonna send an email and track a response to an email or something like that. It's a it's a more standard way of creating and tracking approval requests. Is the is the approval process,
Matt Dressel:you mentioned tasks as well as the approval process, standard approval process, is that, you know, what you guys see most of the time and and that's the the tool in the tool belt you pull the most for that type of of thing?
Mike:It's what I like to use. It's my first go to. There are scenarios where it doesn't quite fit. You can customize approvals so you like, you can have the the simple approve deny or accept reject option. But you can also come up with your own custom options.
Mike:Right? Put it on hold, for example, is a potential response which could have a specific outcome in your process depending on what you need to implement. So you can customize it to a great extent. What we deal with a lot is stage, multi stage approvals. There is a mechanism within an approval itself that you can make it multi stage, but it is not dynamically configurable as as easy.
Mike:Like once you start that approval, let's say in Power Automate and you've got those stages, they're just what they are.
Matt Dressel:They are what they are.
Mike:You can't inject somebody into the second stage of the approval depending on the first person's response, for example. So in those scenarios we come up with custom processes to deal with multiple stages.
Matt Dressel:Cool. Okay. So I think we have painted a picture. So in this new process, instead of an email, they come in through a Canvas app. So this user comes in and they submit a request in a Canvas app.
Matt Dressel:That Canvas app walks them through the things they need to get. They fill out the information. They click submit. It goes into a Dataverse table. That Dataverse table starts to trigger some flows to say, hey, we got a new request.
Matt Dressel:You need to start looking at it. You're the first step in the process. Go do this thing. They start to approve that, manage that all the way through to where it's either approved or rejected. And then you've got Power BI reports and Teams messaging and some other things going on to kind of orchestrate that process and monitor that process.
Matt Dressel:So Power BI to monitor what's going on and Teams etcetera to go communicate about that and and allow people to kind of collaborate on that process. Sound about right?
Mike:Sounds right. One other benefit of the Canvas app, if you choose to use it, is you can use the push notification feature.
Matt Dressel:Oh, right into the Canvas app?
Mike:Right into the Canvas app. So, somebody's got the Canvas app on their mobile device, they can receive that push notification.
Matt Dressel:Feels like a modern built in app. It's close. Yeah. Yeah. Okay.
Matt Dressel:Cool. I think that's really great. Let's talk about that. We got this new scenario, we built it. We spent the time.
Matt Dressel:We we got this app built. It was super easy because everything's low code and super simple. And what what have we gained by just doing that? Without any AI, we're just doing this, you know, modern Power Platform implementation. What are the things that got gained out of that?
Mike:So we've closed the loop or closed the holes on a lot of the things like the process was slow. Nobody knows where my request is, so the black holes are gone. I think the biggest gains though are we have all of the data. We're tracking it in a structured relational data store, which is something that we probably didn't have before either because it was in some free solution or it was in a shared mailbox or in multiple people's mailboxes, right? But now it's in one place where we can actually report on it, quantify the results.
Mike:So now we can actually measure whether or not the process is getting better. If there's still a problem in the process, we can pretty easily find out where that problem is by simply being able to measure those things.
Matt Dressel:Cool. Okay. We've gotten some efficiency gains. It was obviously better than where we were before. But when we started, AI wasn't available.
Matt Dressel:We didn't have Copilot. We didn't have any of these features. Now that I have Copilot, let's start talking about some of the ways that you might think about applying that to this problem in a way that it kind of takes it to the next level. Closes some gaps maybe you didn't think you could close before or maybe, you know, weren't something you thought about previously before you had this technology and this capability?
Mike:Yeah. So there's a few areas that I can think of now that we have Copilot. And it's some simple things like, instead of the user using a Canvas app, we could give them access to an agent where they could have a conversation and say To use request. Yeah, I'd really like to use this software. It could help them gather the information to fill out the request because we know we have a structured form, we have data that we we have inputs that we know that the reviewers need and so they could have a conversation with an agent that could then go retrieve information about that device or that software and attempt to fill the form out for them, like that would be one area.
Mike:We can even help with a pre submission check. Like if they pick a piece of software that we already know about that somebody said, nope, you can't use it, we can be like, hey, no, this is denied right off the bat, right? So those would be
Matt Dressel:areas Or you can submit it if something's changed but Or if it was denied a year ago.
Mike:Right. So there are some of those areas that we, once we have that data that we didn't have before, we can allow Copilot to have access to that data and pre answer some questions.
Matt Dressel:How diff how is that different than building into a Canvas app logic that says, oh, somebody put this name of this software in, I'm just going to go look through the things and I'm just going to do it myself. What's the benefit of using AI? Because I, you know, I can imagine I'm a developer, I can imagine a where I could develop a solution like that.
Mike:Yeah, so I would say first that is possible and that's something that is already being done, right? So if I'm searching for a piece of software, well, I have a database, a catalog full of software, look it up. Mhmm. Right? And oh, that looks like the one.
Mike:Mhmm. But then there are gonna be other scenarios where, you know, with software there's always new versions of the software and the EULAs might have changed, so that's something that often gets reviewed. So there are like various requirements there And that's something that AI can analyze, for example, the documentation of the about that software or about those devices and tell you if, oh, this this is actually different. This is a completely different request because something has changed about that software. Maybe it's a different version.
Mike:Maybe the documentation for the software like that you supplied is different or that the bot found out there on the Internet is different. And that can help in the review process as well to highlight we think this is different, right? This isn't the same thing.
Matt Dressel:Okay. Okay. What's what's another area, what's another part of that process that AI would is just waiting for AI to kind of transform and help?
Michaela:Well, I think of the next step after the user submits their request, it's going to undergo this review process. And so, can AI pre review it of highlighting things in the EULA that we might disagree with or might be concerns, things like that? Or can it maybe suggest an alternative? So, we have this other product that you might want to use instead. Things like that to help now the review side of things, the back end side.
Matt Dressel:Okay. So reviewing the EULAs, trying to do some of that analysis for the end user. Not necessarily doing it completely, but like pointing to some things that maybe they want to review rather than having to manually review the entire EULA or manually reviewing, you know, the features or whatever might be on on the product.
Michaela:Exactly. Yep. And that's a major concern that a lot of folks have right now is we still want the human involvement even if we have AI as a tool to enhance this.
Matt Dressel:So, what would you guys say to someone who's looking at this and going, well, I already pay for Microsoft three sixty five Copilot or maybe ChatGee PT Pro for all my users, can't they just go use that and say, hey, tell me about this EULA and give me, you know, recommendations of how this doesn't apply? What's what's wrong with doing that? And then I don't have to spend all the money on this integration with my app.
Mike:Well, they can, but I would say that they're missing out on all of the the valuable content that you may have collected in that system over the years. And what that really points to is whatever those other tools that they might use are aren't using the same grounding or aren't using common grounding that needs to be used for all of the requests that go through this process. Right? Otherwise, it's 800 different processes. Yeah.
Matt Dressel:So everybody who's using it might do it a little differently Right. Or might not have access to all of the information that
Mike:Guaranteed they won't have access to to all of the information. Okay.
Matt Dressel:Yep. Excellent. Excellent. Okay. So what what else?
Matt Dressel:What's another area that AI could potentially help with this process? For let's say, the person who's administering this and managing this process?
Mike:Well, once you have this data, like that's just one more way that a tool like Copilot can be used is to go and do an investigation, find out what's going on with that process. You could obviously you could have some Power BI reports that tell you how long it takes on average to complete the process. But you could have a conversation with an agent that
Matt Dressel:Understands your data.
Mike:Understands your data and you could simply say, how long does it typically take for this to happen? How long did it take in January? How long does it take today? Right?
Matt Dressel:Yeah.
Mike:And get those questions in a user friendly, a conversational way, which is useful.
Matt Dressel:Yeah. Okay. Okay. Cool. So I think we've articulated some ways that you could improve the process.
Matt Dressel:Let's talk a little bit about whether or not that is what you should do. Right? So should I be looking at every business process like this and say, AI is the best thing, everybody's talking about it, let's go apply it?
Michaela:Yeah, I think that's a question that a lot of folks are starting to ask, because it is so exciting, people want to use it, but is it always the best use of the tool? And so, I think if you have something that is one to one, where you have data that you're going to just input into something, you don't need to analyze it, transform it in any way, is AI the tool to leverage there? I don't know, it comes with some cost to it. There's licensing aspects, depends on how much you use Co Pilot agents that, you know, if you use it more, you're going to pay more. And then, it's also the, like, information that it hands back to you.
Michaela:Is it quality? Is it trustworthy? You know, do we have to still review what the AI is handing us? And so we need to figure those things out.
Matt Dressel:And does that provide a benefit? If I still have to review it anyways, what's the real value? What's the real return on investment?
Mike:You know, I think that's the biggest one is uncertainty or risk, perceived risk and liability associated with taking action on the responses that you got from the bot. And depending on the like an IT request process, probably not life or death. I can think of some other scenarios that might be life or death that you might not want to trust the robot. Yeah. Or you or or you can't afford to trust the robot, so you gotta review it anyways.
Mike:I think there are areas though that it helps, right, like high level areas. Decision support, process automation, right, those are some big ones. I mean, assisting the user, but you just need to know where the user needs to draw the line, right, what they need to look at. And honestly, in an agent, you could implement logic that says, I don't know what to do here. You make the decision or escalate it, right?
Mike:Yeah.
Matt Dressel:The confidence of whether or not the quality that you have. Right. Yeah. There's a bunch of things you can do. The only thing that I would add to that is that the more precise you need it to be, the more expensive it is to do.
Matt Dressel:Right? If you need it to be 100% accurate every time, you should look at that as being, you know, a more expensive thing to do. It's just it's going to take more time, energy, etcetera.
Mike:That's what I call the Mat Bot.
Matt Dressel:Not for sale. You can't subscribe nor would you want to. So let's wrap this up. Obviously, if you have a business process, the first thing you should do is automate it. If you're still working in email and Teams messages, etcetera, there's not really much that the AI is gonna do that's transformative to
Mike:your business. Not on its own.
Matt Dressel:You're gonna wanna put that into, you know, into Dataverse, into Canvas apps and Power BI and use all those things. From there, you can add AI in for things like smarter workflows, less manual work to get these things started, you know, proactive insights, so analyzing something ahead of when you might Those look at are all things that we've talked about. And you really shouldn't just assume that the business process is going to benefit from AI, from what you were just saying. You really should look at that and assess whether or not there's going to be value to that. Does that make sense?
Matt Dressel:Did I cover everything that we kind of talked about?
Mike:I think the only thing I want to add is I want to encourage listeners out there, if you are listening, to take what you've learned here and think about what workflows in your organization this might apply to. So whether or not you have a process that needs help and is in the stone age and needs to grow up a little bit before you can apply AI, or if you have something that's already in place, you've got some data that you're tracking, whether or not you can apply something like Copilot to it. Like, think of those areas where you can apply it. And we'd love to hear back from you if you have any
Matt Dressel:Yeah.
Mike:Cool examples.
Matt Dressel:Send us send us some examples. We're always looking for interesting ones. Well, this was fun. It was a great time. Change of pace.
Matt Dressel:Mitch isn't running the show this time around on the podcast. Hopefully, did a good job. Mike, Mikaela, thank you for the for the time and hope everybody enjoyed the podcast. Thanks everybody. Thanks, Matt.
Michaela:Thank you.
Mitch:Hey. Thanks for tuning in to make others successful. If you enjoyed this episode, we'd love to hear from you. There's a couple ways you can do that. One, we'd love for you to rate our show on your favorite podcasting app and then if you have feedback or topic ideas or suggestions for future episodes, head over to bulb.digital/feedback and let us know.
Mitch:Your input is super valuable and we'd love to hear from you. Don't forget to subscribe and share this podcast with others who care about building a better workplace. Until next time, keep making others successful. I'll see you.
