Leveraging AI

278 | How To Build AI Agents: A Step-by-Step Playbook for Business Leaders with Jim Spignardo

• Isar Meitis, Jim Spignardo • Season 1 • Episode 278

Share episode

Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.

0:00 | 52:06

AI isn’t failing companies. Poor implementation is.

In this tactical, no-fluff session, you’ll learn exactly how to build AI agents that solve real business problems, from drafting winning RFP responses to eliminating internal knowledge bottlenecks.

We’ll walk through a real-world example of how a 500+ employee global organization built AI agents that reduced proposal effort by 80%, improved response quality, and transformed how teams access internal expertise. You’ll see what worked. What didn’t. And how better data discipline made all the difference.

Jim Spignardo, Director of Cloud Strategy & AI Enablement at ProArch leads AI adoption across a global Microsoft partner organization, overseeing AI governance, internal enablement, and real-world Copilot deployments. He’s built over 30 practical agents across sales, consulting, and operations — and he doesn’t just talk about AI strategy. He operationalizes it. At scale.

You’ll leave with a clear, step-by-step framework to:

  • Identify high-ROI AI agent opportunities
  • Structure and clean your data properly
  • Design effective instruction sets
  • Avoid common implementation mistakes
  • Decide when to use agents vs. notebooks/projects

About Leveraging AI

If you’ve enjoyed or benefited from some of the insights of this episode, leave us a five-star review on your favorite podcast platform, and let us know what you learned, found helpful, or liked most about this show!

Isar Meitis

Hello and welcome to another live episode of the Leveraging AI Podcast, the podcast that shares practical, ethical ways to leverage AI to improve efficiency, grow your business, and advance your career. This is Isar Metis, your host, and most companies today are looking for ways to build AI agents, and if they're not, they will be sometime very soon. Now that being said, most companies do not really understand what's involved in building AI agents that will work effectively, consistently, and safely within the environment while leveraging the resources that they already have, such as the data, the knowledge, and so on. And that is exactly what we're going to focus on in our episode today. Our guest today, Jim Spi Bernardo is the director of cloud. Strategy in AI enablement at Porch Arch, which is an IT services and consulting company that helps other businesses with their IT services. And as you can imagine in the past two years, that means a lot of ai and Jim is the one that has been leading it. Now. In addition, he has a couple of decades in the IT space, mostly in the Microsoft universe. So he comes with a lot of experience way before AI on how to implement business processes with an IT through an IT lens in an effective way, which makes him the perfect person to teach us this. Now, this means that Jim, in his day to day, he helps other companies identify opportunities for implementing AI agents and then helping them put together in place all the stuff that they need, such as data alignment with their business needs. Data safety and security aspects, all these kind of things come into play on what Jim does on the day to day, and today he's going to walk us through basically that process end to end. what to do in order to build an agent effectively and to make this even more interesting. The specific use case that we're gonna use to. Give us an example. He's gonna present a process on how to write effective proposals based on RFPs. Now, even if you don't do RFP work and you just write proposals, this is still a very good thing for you to learn because it's basically saying, how can I deeply understand the requirements of my clients using ai? And how can I connect that to the knowledge base that I have inside my company in order to write a proposal that is going to win me the business? Which I assume most of you want to know, because that means you can win more business with less effort, which is usually what business is trying to do. and because I think both these aspects, knowing how to write and create agents in an effective way, as well as winning more business by writing better proposals with a deep understanding of the requirements of the client are both extremely valuable in today's era. And so I'm very excited to welcome Jim to the show. Jim, welcome to leveraging ai.

Jim Spignardo

Thank you so much, ISAR. I appreciate it. It's great to be here.

Isar Meitis

we have a live audience, so we have people here joining us on Zoom. We had a technical issue on LinkedIn, so if you are trying to join us on LinkedIn, I apologize, we'll try to figure it out for, uh, next week. But if you are with us live, so first of all, thank you. Feel free to introduce yourself, in the chat. Say where you're from, uh, what do you wanna learn today, kinda like why you joined the session. And if you have any questions, please write them in the chat. I will bring them up to Jim in the right time. If you're not here with us live, the question is why not we do this every single Thursday at noon Eastern and you can come and join the cool people and be able to ask questions instead of just listening after the fact. So feel free to come and join us. There's a link on how to do this in the show notes. So with two clicks, you can sign up for this, have it on your calendar, and join us whenever you can. Uh, but now, Jim, the stage is yours. show us the magic.

Jim Spignardo

Appreciate. Thank you very much, ISAR. uh, I'll go ahead and share my screen, and kinda get things rolling here.

Isar Meitis

for those of you who are just listening, by the way, you, we are gonna tell you everything that's on the screen, so you don't have to be worried about that. This is a screen heavy kind of, uh, thing. But if you do want to see this, uh, you can see the episode on YouTube and or on Spotify. There's now a way to watch the video. So either way you can do that, but if you cannot, that's perfectly fine. We're gonna explain everything that we're doing.

Jim Spignardo

Yeah, thanks. I appreciate that. And so, ISAR made a real, interesting comment there. And we've been on this journey at Pro Arc probably almost two years now, and. The way we kind of got started was really looking at the use cases, the high value, low, um, low effort use cases that we could deploy in our organization. And we, we rolled out our licenses of copilots copilot over many months once we've established use cases for individual personas and roles. And one of the ones that we came to as soon as we got to sales was responding to RFPs request for proposals. traditionally as an organization, it is something we absolutely dreaded. Uh, usually came with a very long document of requirements of, of, scope in deliverables and, uh, asked a lot of information about our organization. And although we had responded to many in the past, we had mixed success, maybe about a 50% success rate, but what it really did was took away from a lot of people's time because these things usually. Came out of nowhere, you know, on a Tuesday, and expect to be able to respond by Friday to this, and, and put together something that was actually going to have impact and win you the business. And so in a lot of cases, we were becoming, uh, a little shy about responding. And so we knew there was an opportunity here to leverage ai, specifically copilot, to help us with this. what we really needed to do though, was start organizing our data. And so we started pulling together various sources of information that we've used in the past, other RFPs, various, information about our organization as well as some, uh, websites that actually talk about what makes up a good RFP. And so that was kind of the starting building blocks. And what I'm gonna show you today is kind of how we began that process. I'm gonna jump around a little bit. I'm gonna show you initially, I, you can kind of get started. From a no code perspective of just telling, copilot what you wanna do and show you how it starts to build the framework for that agent. And then we'll kind of switch over to the one we've actually built to show you what looks like in the background, what kind of sources it has, and then we'll actually put some, I'll put a, uh, a simulated RFP into the system to see how it actually responds.

Isar Meitis

I, I wanna pause you just for one second because you said two things that I think are really, really important. one. Which I really like, and too many companies do it the wrong way. As you said, we started giving licenses to people after we identified use cases for them to use. Correct. most companies get licenses to everybody and then they try to figure out what to do with them. Uh, correct. So what Jim is saying is the right way to do this, find the right use cases for the right departments, for the right people, train them on how to use the licenses to do these things. Hopefully give them some kind of a working environment so they can just use it and then develop from there and then give them the licenses. So that's one thing that is, you kind of said in a sentence in the background, like, no, this is actually pure gold. Uh, the other thing is, the way I tell people, the way you need to look at building agents is where you have business bottlenecks and you described the perfect bottleneck. It is driving business to the company if you do this successfully. But if you. Consistently struggle with doing this. It means you're investing a huge amount of efforts of the company for maybe a 50% success rate, right? This is the ultimate thing to automate, right? It's gonna, by definition, it's gonna drive new business, and it's not gonna do this by consuming more time that may not be, uh, useful for the business. So you're winning on both sides. And so both these things are fantastic. Even before, even before we get started.

Jim Spignardo

No. Absolutely. And that's, that's been our intent all along in our journey, right? We, we need to justify because this is not cheap. This technology's not cheap of at least the, the stuff that's good and worth using. Um, and so, um, we need to make sure that we're making wise investment and, we've done this repeatedly across various groups. I'm in charge of tracking our adoption. And, and so, you know, this has been a very successful model for us.

Isar Meitis

Awesome.

Jim Spignardo

Uh, if you're following along and you have a copilot, uh, license and you wanted to kind of maybe, see what this is all about, I'm in the copilot chat app, logged in with my M 365 account. Uh, this interface is also accessible in the browser. It's also accessible from teams. If you're in there. Um, so it's a similar interface no matter, no matter where it presents itself.

Isar Meitis

Yeah. And by the way, you can do exactly the same thing in Chachi pt, in Claude, in Gemini. Like, it doesn't really matter. Whatever we're showing you now in copilot, you can do on whatever platform you you're using for work.

Jim Spignardo

Yep, absolutely. There's definitely a lot of cross, uh, capabilities here. So what I'm gonna do, go ahead and I'm gonna go ahead and click my little sidebar here. And within this area you'll see there's an agent, uh, section called for agents. And you'll see we have a lot of them. Our organization, I think currently has 42 agents deployed.

Isar Meitis

Oh, wow.

Jim Spignardo

And, some of those, I built some of those other people in the organization, organization built. Um, so it's been kind of a real, uh, uh, uptick in, in, in the usage of these. But the quickest way to get to being able to build an agent is you just go ahead and click this new agent button here. And Microsoft has kind of changed the behavior a little bit as it relates to these no code agents. It used to throw you into the describe screen versus the configure screen. If you're very new to this, I would say flip over to the describe screen.'cause what ultimately you're gonna be doing here is using AI to help you write the instructions for your agent for you by simply explaining what it is you want to do. So I'm gonna come in here in my chat window and I'm gonna say, uh, I want, a an RFP response bot that can, let's see, that can, consume RFPs and, provide a, proposal output. Now, hopefully, based on this, and this is very vague, I would say we'd probably wanna spend a little bit more time providing some context sources of where we wanna pull information from. And I'll show you the one that we've actually completed. Now it functions, but that's really all you need to do to get started here. the other thing I would tell you is another method that I use sometimes when developing agents and getting an instruction set is I'll actually go to chat, whether it's open AI chat, chat, GPT or copilot itself and say, I'm building an agent for this platform. This is what I wanted to do, create me a really good instruction set. And oh, by the way. Also ask me some clarifying questions that may help me make this a better agent. So

Isar Meitis

this is, this is how I build all my agents, exactly what you're explaining. Now I start, and I actually do more than that. The way I do this is I start in a regular chat and I try to do the thing. So instead of telling it, I'm building an agent, I'm actually trying to do what the agent needs to do. So I'm like, okay, I have this, these four RFP documents.

Jim Spignardo

Yep.

Isar Meitis

And here is two proposals that we've written in the past based on previous RFPs. Uh, one of them was successful, the other one was not. Here is the proposal that actually won the RFP before, because if it's a public, uh, RFP, usually they will give you, show you who won and exactly why. Yeah. So we can put that in there. As well said, I wanna build a new proposal that will use the new RFP based on what learning the other RFPs. Uh, let's work on that together. And then you work through the steps. And then when you get to the final outcome, like, okay, this is a pretty good proposal that was written, or analysis, or whatever step you are in. I'm like, okay, now I wanna turn this. Into an agent or a custom GPT or a project, whatever, it doesn't matter. Depending on the platform you're on, please write me the instructions and they will write you incredible instructions that you can never write on your own. Yes. And then that becomes my first draft. There's still fine tuning afterwards. But that is to me, the best ways to actually do the thing. Yeah. You're trying to do in a regular chat and then ask it to convert it in. Once it's working and you went through different lefts and rights and turns and kinks, this is where I ask you to actually turn it into instructions.

Jim Spignardo

Yeah. and that's actually the same process I'll use if I have someone who's reaching out to me within our organization, says, Hey, I wanna build a, an, uh, an agent. I'll, we'll sit down and I'll say, first, show me how you would do, you'd do this. Without an agent, and because I wanna see it mind mapped out, we wanna map every process, every type of output they're looking for. We then take that and we, like you said, convert it to an instruction set, have them test it, and we go back and fine tune it, um, through, you know, limiting certain things or adding certain other variables or criteria. But yeah, I, I, I, that method is, is also one I employ as well, so. Yep. All right. we're ready. Just see what happens here. So we're gonna go ahead and, hit the, uh, the, uh, submit button. Give it some, give a couple of seconds here to think about it. I'm sure it's gonna probably ask me some from additional qualifying questions or, so we'll give it a second here. Depending on if, co uh, if Microsoft fed the squirrels today will depend on how quick this is.

Isar Meitis

Yeah.

Jim Spignardo

we always have a backup with live demos. We never wanna just, uh. But, uh, I'm confident we'll get a response here in a moment. And the nice thing is you kind of work in this canvas on the left hand side, and then the right hand side, it kind of starts to output and build what you're actually creating.

Isar Meitis

um, yeah, so for those of you who never used, either copilot, agent builder or custom GPTs, they kind of look the same because they came from the same source. You're working on the left and you're seeing the output on the right.

Jim Spignardo

Yep.

Isar Meitis

And so you are basically in real time, can see the output of what you're doing. And you see it updates as it updates and you can test it still here in the quote unquote development environment. And so you deploy it only once it's ready, which is very helpful.

Jim Spignardo

Yep, absolutely. And you can see it, it set it up, it gave it a name. We don't have to stick with that name if we don't want to gave it a description. It did give it a fairly limited instruction set at this point, um, but also created some conversation starters or some, some suggested prompts. Um, and then kind of determine that, uh, based on the capability, uh, it was going to enable data analysis and code interpreters so it understood those might be valuable skills for this agent to have. Now you'll notice down here, we could go through the process of uploading an RFP document or paste its content, or we could put, put in some information about our company, all things that we've done in the one that we've already built. So I'll kinda show you that in a moment. In a moment. But I wanted to flip over to the configure tab and demonstrate what was pulled in from just that single line of prompt. So you'll see I have the name, I have a description, and down here we have a very, extensive set of instructions. The purpose, it's general guidelines, the skills that we'll we will employ to essentially accomplish the task, and then some step-by-step instructions. Right. Review the document provided by the user, extract and outline the objectives, requirements, timelines, all things that are typically in a good RFP, um, and then organize that into specific instructions and then present, uh, the proposal to, to the user for review and making adjustments. There are some error handling in here, and this seems to be the default now. Microsoft is doing a much better job of creating these instruction sets from a statement. Now, um, when I first started using this, it was not as pretty, it was not giving you all these kind of sectioned areas. It didn't always include an error handling area. Now it seems to do that as well as some things like feedback and examples, all things that will really improve the output of the agent. So you'll see in here, uh, if there's anything incomplete or unclear, ask the user for clarification or additional documents. Uh, incorporate user feedback. Uh, and here's some examples, right? Here's an RFP. I've identified there were following requirements, blah, blah, blah. And then filing the follow, the follow up. In closing, you also have the ability to give it source document documents. So those could be internal websites, external websites. It can be, documents within your environment. It could be, things that you upload. This is actually something that's, been, uh, added since, uh, this technology came out, is the ability to bring documents that might be outside of your M 365 data state. Now, the one thing I would say about that, if you're linking things. Linking to things that are in your environment. The agent will always use the latest information because it'll go and grab that. So if there's revisions to it or for, or if there's new changes, it will continually be dynamic. If you're uploading content, you have to be aware that if that check content changes, you have to upload new versions to the agent for it to be able to read it. So,

Isar Meitis

yeah. Two things I, I want to add, or, or maybe clarify to people who have not used this in the past. One is you talked about conversation starters. Conversation starters are basically buttons that show up when you use the agents that give you a way to start the agent. Now, you don't have to have them, but if you're gonna deploy it in a, deploy it in a business and other people other than you are gonna use it, they may not know what to do because there's no instructions

Jim Spignardo

might

Isar Meitis

not be obvious. And so what this does, because it said, oh. click here and upload your RFP to get started. And then you're gonna click here and upload RFP because that's what I told you to do. And you can have many of those, and you can, even in the instructions, if you're putting a few of these, you can use them as different entry points to the process. As an example, one of them is you already have a draft. I already have a draft, I wanna continue working on it. So click here if you already have a draft. And then, then the, the agent will pick up in a different step in the instructions because in the instructions it will say, if the user clicked on, click here instead from the instructions start here, and then it will know how to do these kind of things. So this is for the conversation starters, but the other thing is more critical because one of the cool benefits of using copilot as a way to build these kind of agents is you can connect it to your SharePoint, you can connect it to a specific SharePoint drive, which means it will look to the drive, what's in the drive every time it gets started. So if you upload a new document, if you delete an old document, now it has a new set of documents, a new set of data to work with, versus if you upload files into the agent, they're quote unquote hardcoded inside the agent. The fact that you, in your SharePoint now has version three, it doesn't know that because it's looking at the one file that you uploaded. So from a best practices perspective, what Jim said, that's not necessarily a bad thing. You just need to be aware of that, right? If you're uploading a document, that's the document it's going to use. If you're giving it access to a SharePoint, uh, drive, then it dynamically will look what in that SharePoint. the other thing which I'm not sure about, uh, maybe you know, is does this provide a different level of. Number of documents or total volume in megabytes that it can have access to. So if I give it a SharePoint drive, do I get more data it can look at, or is it the same as if I upload the 20 files manually to the agent?

Jim Spignardo

Yeah, I, I haven't seen what, I have seen some limitations about the number of documents you cannot upload, but I have not seen any limitations as it relates to, I just pointed out a library, and that library may have 20 top folders with a, 600 sub folders with thousands of documents. that's really why when you can, it makes more sense to try and go that path because a

Isar Meitis

hundred percent,

Jim Spignardo

yeah.

Isar Meitis

Okay. Perfect.

Jim Spignardo

Absolutely. as I was mentioning here, you know, you can, you can add all this content in, you can upload content, you can also have it use the, internet. I would say I, I tend to lean away from this, especially if I'm building an agent that I know is gonna have a specific purpose and I want to control it. to the degree that I want the output to be looked very consistent. When you allow the agent to search the internet, it can pull in all kinds of other data, that may be or may not be accurate. So if you're trying to eliminate, eliminate, uh, or reduce the number of hallucinations, I usually make sure I don't have that on. But again, depends on the purpose of the agent. The other thing I always like to do is make sure I say, only use the specified sources. What that, what that does is essentially limits the agent to only the things that I provided it, and it can't go outside the bounds of that and try and find other data. That may be related or relevant or try and make things up as well. Um, and then the next one, really, this is kind of more towards whether or not the context of who you are in the organization would actually be beneficial to the agent, Microsoft. Uh, because, copilot lives inside your tenant has the ability to understand relationships with what you do for work, but also how you work with others. So sometimes if that can be used to a benefit in an agent, I will turn this on. Um, sometimes where that comes into play is. They can know my role and automatically answer based on the role that I play in my organization, which gives me answers that a bit more tailored to me. But the problem there is that what if I didn't want it to answer in that role? Well, then you kind of have to go, Hey, assume I'm a, I'm a different role at this point. I'm gonna leave that off of the purpose of this. the other nice part, which just recently, um, Microsoft added was the ability to change this icon but actually have it generated by ai, which is kind of cool. You don't have to go out and create a co uh, create a. A, a cool badge for your agent in another system or in copilot and drag it in here. I will tell you it's a little finicky. It doesn't work every single time. Uh, sometimes it makes the agent error out, but what it will do is look at the, the instructions, the description, the name, and come up with some sort of badge that represents that. I do find that that's very helpful when you have a lot of agents and people are trying to visually recognize them quickly, instead of giving you this kind of default, agent logo, which can kinda get lost in a sea of other ones. Do you see it start to materialize here? We'll give it a second to populate. and then we're gonna probably bomb out of this and go into the actual live one that we have. So it's got a cute little robot, right? It's, um, and, and I, I've actually learned something about how visual, uh, image creation works. It's kind of like building a castle from just a pile of sand, and that's why it kind of does this thing where it starts to assemble the bits. So you can see it's cute, it has a little check mark, it has a little document behind it, and now if I click apply, I can go ahead and add that in there. It will update my agent, and you'll notice there's the error. So that doesn't always work every single time. Um, in that case, sometimes I'll just screen scrape the one there and then just upload it. the other thing too is after you've created your agent, you have the ability to share it to your organization. You can either share it to the whole organization, you can share it to a subset of individuals. Either the group or Indi, individual user accounts. So I'm gonna actually,

Isar Meitis

so here's, here's an interesting it question for you, uh, related to that, which I truly don't know the answer, but I'm curious, let's say that this agent is using a specific, folder in my drive.

Jim Spignardo

Mm-hmm.

Isar Meitis

And let's say that I share this with the entire organization, but some people do not have access to that folder on their personal logins. will the agent prevent them from seeing that information? Did Microsoft close that loophole? Or if I gave them access to the agent and the agent has access to a folder, you're gonna get access to that folder even if you're not allowed to see that folder.

Jim Spignardo

Yeah. So if that content exists in M 365. Maybe that's your OneDrive or Teams or SharePoint. Uh, the agent can warn you that the people you're about to share it with don't have access. And you can, if you have the privileges right, you have to have the ability to assign permissions. it can do that as part of the agent, agent creation. The thing you have to keep in mind, uh, is that if you do do that during the agent creation and you then remove the agent, it doesn't remove the, the permissions.

Isar Meitis

Got it.

Jim Spignardo

If that content exists somewhere outside of M 365, it cannot just give you permissions to my C Drive, for instance. Right?

Isar Meitis

Or, but, but it doesn't work the other way around. Meaning, if I, let's say I do not assign permissions to the people who do not have permissions.

Jim Spignardo

Yep.

Isar Meitis

Would they not be able to use the agent? Like what's gonna happen then? Like would they be

Jim Spignardo

Yeah, they, they wouldn't, right. So it would got

Isar Meitis

it. Because the agent, so it does work this way. It protects the data structure and access from a company perspective.

Jim Spignardo

Yeah, absolutely. Yeah. And, and, and so that's something you need to think about if you go ahead and share it. And again, that they've just added that recently where, kind of similar to when you share a link and Outlook, it will say, Hey, some people in this,

Isar Meitis

yeah.

Jim Spignardo

Email may not have access. It's the same way. It'll say some people may not have access. Do you want to give them access? That wasn't there before. So people were setting up agents and then people were going in and going, this doesn't work, because they didn't have access to the files where they were located. But it, so it does still respect your permissions and again, you can't give permissions if you don't have that privilege though.

Isar Meitis

Got it.

Jim Spignardo

So I'm out here now in my, my agent, uh, catalog and you can see all these agents that we have and built. And what I wanna do is pull up the RFP response bot that, um, we actually have for our own organization. And so I do see that one right here, but I know, I think I actually have a second one, so lemme go ahead and type that in. Yeah, this is the, the newest one I have. So I'm gonna go ahead and find that in my list. What I wanna do is go in and edit that,'cause I wanna show you what's occurring behind the scenes. And so it immediately jumps to the configured tab and you'll notice that similar instructions, um, although it's a little bit different because it has some things in here about tone and style and, I actually used the second method that I talked about to create this one, which was to collaborate with copilot and tell it what I wanted to do to build me the instructions. Which I then brought into the agent, and so that's why I'm getting a little bit different, uh, output. I think things in here about compliance and ethics per parallel processing capability where we can handle multiple RFPs at the same time. Um, we also have some limits and transparency, uh, response optimization, and then continuous learning and feedback. You'll notice if I go down a little bit further here, this is where I have given access to these resources. So within our organization, I've given it access to our services SharePoint site, which goes in and talks about all the services and solutions that we, can sell and, and, and, uh, position with a client. We also have information in our sales and marketing website, which probably is going to fill in a lot of those sales pitchy parts of the RFP. We also have a document that specifically talks about RFP response sections. This we built over time responding to RFPs, and it's essentially, chunks of data within this Word document that typically get filled into an RFP. We also have an A PDF. About what are winning formats and strategies, right? So we wanna give it the ability to say, Hey, when you create that first draft, make sure you're relying on this, these this best practice knowledge about what a winning RFP looks like. We also gave it our location for all of our previous RFPs, and then we also gave it a couple PDFs for best practices for winning a proposal. And then mapping your ideal RFP response. I have up here allowed it to search all websites, which I think I actually, in, in hindsight, I'm gonna turn off. And you notice that I do have the ability to use specified, the only, the specified resources. Um, and over here you can see my, my starter prompts essentially. So what I wanna do at this point, and, and I, I've said before I can show you, this is where you would share it and you can see anyone in your organization, specific users. It does give you a little bit of a blurb about permissioning. Okay? It'll respond, uh, with knowledge that each user has access to. Okay? So it is permission bound to the user unless you change it. So keep that in mind. So let's drop out of this and go directly into the agent. And one of the other things, to be aware of, at least in the Microsoft ecosystem is when I'm in a chat. It could be in a chat, in, in any application or, or just general chat. If I wanna call up an agent, I don't have to necessarily open up the agent directly. If I type in the at symbol within the chat window, that is going to give me, um, a list. And of course I'm in the, an agent right now, so let me show you in a new chat. If I go ahead and hit the at symbol, it's gonna bring a list up of all the agents that are available to me to call up. So I don't, I can pivot any time into using an agent. in a regular chat. And like I said, that works in any of the Microsoft applications as well. PowerPoint, cope, um, outlook, word, Excel, you name it.

Isar Meitis

Two cool things about this, first of all, most people don't know that, and that's sad, right? Because it's a very helpful thing. Uh, so again, you can, you can use this from anywhere, which is very helpful. But the other really interesting thing is it means you can build multiple agents for a single process. So let's think about the RFP for a second. You can build one agent and analyzes the RFE, another agent that analyzes the data in your company that aligns with the requirements of the RFE. Sure. And a third one that actually writes the proposal outline and a fourth that writes the details for the proposal, like you can do it this way. The benefit of that is, well, two benefits, one is you have more control over the process because the in-between steps you can stop, go back, finesse, fine tune, et cetera before you continue. But the second is you can be much more granular instructions for each and every one of the steps, which will mimic more what you're actually doing in real life. But then the reason that connects to what Jim said before is you can have a regular chat and then at the first agent and say, let's start with the research. Here are the RFP documents for this RFP, and then we'll do the research. And then it comes up with an output and then you add the second agent that now knows what happened in the conversation because it's the same context window and this is how you can keep on going. so that a symbol is not. For being quote unquote lazy instead of trying to find the agent somewhere. It also is very useful because you can use multiple of them in a single conversation.

Jim Spignardo

Yep, absolutely. Yep. Good. Good point. Absolutely. And, and here's the rule of thumb I usually, put around agents. if you are in the process of building an agent and there's more than one primary task you want it to complete, it's a good, good chance. You probably want to create a another agent, right? We could have definitely built this as multiple agents, although typically it's a, we found that with our users, this is the best process for them. But we have some examples like we're building one for our HR department right now. There's a subunit of that department, which our talent acquisition department, and they said to us, what we want to, we wanna be able to answer questions like they have and we said, you know, no, we're not gonna do that. We're gonna basically just link to that agent when it gets to that topic and go, is there a question about talent acquisition? We will hand off to that agent so that their agent can answer it. And so we're not duplicating effort in our organization either. All right. So I have my first, I, I've chosen the, uh, initial, prompt here, which is the RFP analysis. I'm gonna go ahead and, um, click the plus button and I'm gonna upload a, um, a synthetic, uh, or simulated RFP that I asked chat, HPT to create for me. it's a school district. let me see if I can bring up, I actually don't have, yeah, I do have that. So this is what it looks like. It's a PDF file. it's pretty sparse. I know a lot of RFPs are a lot more detailed than this, but it created this fictitious school, Redwood Valley Unified School District gave it an address, gave it an RFP number, proposal date, due date, demographics of the school, you know, the purpose, the current network environment, their data center, their objectives, the scope of work. So, I mean, it has all the elements you'd be looking for it. Usually they're much more detailed than this. But, uh, for the purpose of this, um, demonstration, this'll do. So we now have the, the, PDF for that RFP, and we'll go ahead and, uh, click the, uh, submit button. And, uh, now you'll notice what it's doing here with copilot. Now, because it has a model router, it can actually figure out whether it needs to think deeper or it can answer fairly quickly. So you'll notice it'll see it responded fairly quickly because it probably had a decent, uh, bit of information. So below is a structured RFP. Read it ready analysis and draft response. If you wanna reformatted into a full proposal, executive summary, summary or compliance matrix, we just gotta let'em know, right? So here's, uh, the, the key highlights, uh, the purpose. My screen is frozen. There for a moment. It's going on. There we go. Just got stuck for a second. some of the gap risks and, proposals or, or, or for the proposal considerations. this, this part here, it said, uh, for the RFP compliant, the content is structured so it can be pasted directly into a proposal, which is cool. Keep in mind it can actually put it right into a Word document as well. There's a little bit of an executive summary draft, then goes into pro a's technical approach aligned to the RFP talks about, you know, what we would do. uh, as it relates to that specific part of it talks about the, the wireless network assessment. The security architecture review, talks about any cloud and SaaS connectivity. Then it also talks about the deliverables. Here's our vendor qualifications. Again, this stuff actually came from stuff related to our teams internally. So, you know, we have A-C-C-I-E level network architect. We have multiple ci, SSP engineers, we have Azure and cloud network specialists, on and on and on. And then it actually wanted to,

Isar Meitis

and again for people, just a second, for people to understand. Yeah. The way it knows that is because you gave it access to your company data through the relevant SharePoint folders, uh, that has this kind of information. So it's not making it up, it's actually pulling from the data that you gave it. And this is where the real magic happens, right? You need, going back to what Jim said in the beginning, you need to know which data it needs access to. It's exactly this. Like you need to plan ahead and say, I'm gonna create a folder or a sub folder somewhere that has exactly the information this agent needs no more nor less. And I'm gonna point it to that. And like you've seen from Jim, this could be five different folders. It doesn't have to be one that has specific aspects of information that will help it execute what it needs to execute. And because we told it to only use this data, that's the only source of truth it's going to use. And hence it knows how to do these kind of things down to the level of what certifications and how many people with those certification exist in the company to be able to comply with the requirements of the RFP.

Jim Spignardo

Yep, exactly. Exactly. And he was even able to go down here when we started working on pricing models to know that we actually do penetration testing. And, access point lifecycle design, right? So the, these are things that, uh, it may not have been mentioned in the RFP, but we can state that to the customer. Hey, as an optional add-on, we can also do these things for you as well.

Isar Meitis

Yeah.

Jim Spignardo

Goes through a compliance checklist, right? Which basically, um, is checking the draft to make sure it, it, it actually addressed all of the elements of the RFP, right? So it wanted to go through and say, let me check my work, let me ensure that all the things that are laid out there, because I'll be honest, we have answered RFPs where there was a section that we thought we addressed, but to the client, they thought, well, you didn't address it to the extent that we thought you would. So by doing it this way, we get much more consistent responses and we get, the AI to be able to check our work for us. So I. and then down below we get kind of these outstanding questions for finalization. So there might be some things that we want to a additionally add to this. So the one is, do we want, it's asking us, we want to have a complete proposal in a PDF or word format. do you wanna set of narratives, sections to paste into an existing template? or, and, or do you want a compliance matrix and executive summary only? You'll notice. Then I also get some suggested prompts, right? Based on the questions it's asking me, do we want the narratives only or do you wanna generate the compliance matrix and executive summary? What I'm gonna ask it for is, and I didn't see this, can we build out a task list and associated based timeline approach? Let's see what it gives me there. This is typically a lot of times what, what a customer wants to see in a, um, in an RFP. So it did just came out there and said, okay, assuming this kickoff date again, we can adjust those if necessary. Uh, we could complete by July 15th. Uh, the record recommended phased approach is aligned to a 60 day window. Phase one would be the project initiation. It knows, you know, our project management philosophy because again, we've feted documents and knows how we run our project management office. and what that would look like. Down here, we have deliverables out of that phase one, which is the project plan, a documentation request log, and a stakeholder engagement schedule. Under the network infrastructure piece of this, notice this information is coming right out of the RFP. I don't think we've showed you that, but. There was some specs in here about the high school had a 10 gig network that MS School had a five gig network. So on. Again, the deliverables, again, the wireless, uh, heat mapping exercise we will be doing. Even talking about, you know, which environments we're gonna look at that typically have density issues. Phase four, we get into the, so cybersecurity, architecture and compliance.

Isar Meitis

Yeah. I think, I think the, the phases themselves are less relevant. Like, but, but the, the clear thing is, is that it knows how to approach it from a systematic way, right? It knows how to read the RFP, understand the requirements, understand what the company knows how to do, and then put the two together into a very well structured format. Mm-hmm. The other thing that I will say that I do a lot when it comes to these kind of things, like when you're trying to write. A really long document and usually I respond to an RFP is not gonna be five pages, it's not gonna be 10, it's gonna be 50 to 500 sometimes. Yeah, sometimes. Yeah. and the right way to do this is kind of like to follow the process that Jim defines. So you start with an outline and then you go, okay, let's work on section one, 0.1, 0.3, and then you start working on that with the ai. And then once you get that, okay, let's go to 1.1 0.4. And, and this way,'cause if you ask AI to write the full document, uh, let's say it's supposed to be 150 page of answer, you're gonna get six pages, sometimes 12. The only way to get it to dive to the level of details you need is actually to do the step-by-step process of let's create the outline. Let's understand the components we need. Now let's start working on each and every one of the components separately. And as you do more of this, you will get a feel of what the level of granularity you need to. Impose versus what it will know how to do on its own.

Jim Spignardo

Yeah, absolutely. Yeah. And, and that's, that's probably that, that, that extends to a lot of different types of collaboration that I do with ai. And you go section by section, you kinda work out the details, you work out the kinks, you have it, analyze it over and over again and then say, okay, I'm happy with that section. Maybe copy that out, put it in a Word document and move on to the next section. And, you know, it sounds pretty labor intensive, but I will tell you it's a lot less labor intensive than it was in the past for us. you know, if I think about what on average it would take and how many people it would take to be involved in an RFP, you know, it's probably no less than three people, usually four to five people, over the course of 10 days, putting all this information together. We recently responded to an RFP. In less than a day and a half because we were able to pull all this information together so rapidly and get a good first draft. And then, like we always tell people how to human review it and go through with AI and say, okay, we don't like that section. This section reads a little inauthentic for who we are. And, and you can even say, Hey, can you make this seem more on brand? And then we'll do that and we'll change it up for you. But, um, you know, the, the amount of time and that it takes us to draft now is draft. It's probably down by 75, 80% of what it used to take. and then again, our win rate, um, has dramatically gone up. we probably, again, we win. We won about as many as we lost. I would say we're now closer to winning three quarters of our, our, RFPs. And sometimes it's not because of ai it's because we're not qualified enough to do the work. Yeah. Yeah. Or, or there was somebody already had some sort of, unknown advantage over us that we didn't know understand.

Isar Meitis

uh, two things. One, a bit of an interesting add-on to what you just said. And the second is, uh, is a question. So the adding a little bit on a lot of the work in the RFP is really evaluating whether the company actually has the expertise, the skills, the knowledge, the background, the past experience to actually win this.

Jim Spignardo

Yeah.

Isar Meitis

That by itself used to take days of work.

Jim Spignardo

Sure.

Isar Meitis

And then you may get to the conclusion like, okay, we're actually, we can't win this, let's not bid. Right. But you already spent the 2, 3, 4 days, two weeks, depending how big the project is to get to that conclusion. Where now this part of it, if you built your data set correctly, first. In five minutes, you will know whether you should invest more time in this or not, uh, without investing any, uh, serious resources. And the other thing is usually these resources are the most capable resources in the company. So it's not only you need to invest these two days, these are the two days who are the most senior project managers with the most experience with, like, these are the people that can evaluate whether you can do this or not. Uh, these are the people you don't want to get out of their day job because they're the one that generating the most amount of revenue to the company. Yeah. The question that I have in the, che PT universe, when I do this, I use Canvas to write the answers. The benefit of a canvas is that it's a fully editable document that I can add, text myself, I can highlight text and as AI to work with this and so on, which makes the iterative process of creating an outline and then working on 1.1, 1.1, two, two, and so on. Very easy to do. I don't believe that's possible in the copilot universe, but I'm asking, I'm not saying Yeah,

Jim Spignardo

not to the same extent. But what you do have the ability to do is to, I'm looking here, you can actually, uh, send this to pages, right? And, um, I'm, I'm not seeing that option right here, but if, uh, if you send it to a page. The page does open up a canvas, which you can then edit things in and more easily copy from the, from the chat window into that. Um, the feature you're talking about with, um, chat GBT Yeah. Is you're just create for coding too.'cause you can see the changes being made while you're asking for the iterations. Um,

Isar Meitis

so, so your best practices suggestion is working two windows or two monitors, one that has the chat, one that has a pages version of the first draft. And then you go back and forth and

Jim Spignardo

Yeah. And, and, and the nice thing about the page when you send, send the output to a page is that after you're done editing it, there is a button in the top that says, send this to word.

Isar Meitis

Yeah.

Jim Spignardo

Right. So you can take all of that, that ongoing, kind of workshop benching work stuff and say, now I'm done. Put it in word and we'll send it off for approval. Or, you know, last draft or last review by human. So

Isar Meitis

great stuff. Quick recap and summary of everything we did, and I'll let you say some final words. first is you need to know which data you have, where you have it, what's applicable for what purpose. And again, now generalize it. Forget about RFP for a second. That's true for any aspect of the business, right? Yep. So having a solid understanding of what information you have in your company, what of it is good versus not so good versus bad. labeling it correctly and putting it in places that will make sense to humans and agents, was always true, but it becomes way more important right now because if you don't have that, everything that Jim just showed you is not possible, right? Because you won't have the 12, five folders and six documents that it gave the, the RFP, agent to be able to do the things that it knows how to do. So correct number one. You need to have your right data intact. Number two, as we said in the beginning, you need to figure out the use cases. And as we said, the right way to figure out use cases is things that can help you grow the business while hopefully saving you time. One or the other would still probably be a good option. But in this particular example, it does both. Like it reduces the effort. Any generates more, uh, revenue.

Jim Spignardo

Yeah,

Isar Meitis

so it's, it's a perfect fit to be something you wanna build agents for. The next thing is, is really how you build the agents. And as Jim and I mentioned, the best way is to just have a conversation about it. Just talk to the AI in a regular chat to say, this is what I'm try to do is information I have, uh, this is the goal that I'm trying to reach and just work iteratively. And then in the end said, okay, now write me the instructions. and then the last thing is really. it's never really done, right? This gets you 90% there, but they're like, oh, we should have added this. go and add this. Like every time you learn something you didn't do before, you can go and add that additional thing to the agent. In many cases, I do this in the chat itself, so I'm like, oh. I, I would like the agent to also do this. Can you add me a set of instructions to the current instructions that will add this new functionality capability? Check whatever you wanna do, and it will write you another paragraph or two in bullet points and we'll tell you exactly where to put it in. And you copy and paste it, it in the instructions to just keep making it, better and better. Uh, and then, and then like Jim said, you can share this with relevant people in the company. So you're not the only one using it. The only really annoying thing that Microsoft did, but I understand they're trying to make money, is the other people need M 365 licenses as well in order to use the agents that were created, which to me is really annoying. Like it would've, it would've made more sense. There is

Jim Spignardo

a little bit of, a little bit of a footnote to that, by the way.

Isar Meitis

Okay. I'm, I'm, I'm curious,

Jim Spignardo

so if, if you do not wanna license your users all up, uh, and still wanna take advantage of agents. Microsoft does have a pay as you go model

Isar Meitis

Ooh.

Jim Spignardo

Where your, where you now can buy token packs. And so if you, if your agent is associated and, and set up for the pay as you go model, um, this is great for frontline workers, field workers. Yeah, yeah, yeah. they can use the agent every time they use it. You get charged a penny for each message you can buy. Also now buy, packs that are at a discount so that you can assign those as well. And so now you, you're not in that model where, I don't know if they're gonna use$30 worth of copilot every month. Right.

Isar Meitis

So

Jim Spignardo

it's

Isar Meitis

a much, now I have 300 employees that go$30 is not a big deal. Yes.$30 time. 30,000. That starts being interesting. Yep. So that's awesome to know. I did not know that, Jim, this was absolutely fantastic. I'm sure. People will learn a lot both from the mindset as well, from this specific example. I think it's a very powerful example that, as I mentioned, even if you're not doing RFP work, you're probably writing proposals. So don't call it an RFP, call it customer requirements. Right? Yeah. And it's, and you go through the same exact exercise and you can do this. Sure. I can tell you that I haven't written a proposal on my own for at least two years and I keep on upgrading and updating the process. So the process that I have right now is way better than I had even a quarter ago. but it now writes better proposals than I ever did. By a big spread. Yeah. By a big spread. And he does it in about five minutes when it would've taken me two to three days. Yeah. And so there's really zero reason not to do these kind of things. That's all I'm saying. It's, there was the old joke, you know, in the, in the, in the project world of faster, better, cheaper, pick two. You don't have to pick anymore. You can do faster, better, and cheaper. That

Jim Spignardo

is true. I didn't think of that. Yeah, that's true. Absolutely. I've heard that expression. But yeah, AI kind of change puts, turns that on its head. So

Isar Meitis

yeah. if people wanna follow you, learn from you work with your company, what are the best ways to, to do that?

Jim Spignardo

Sure. Um, I am very active in LinkedIn. Uh, I write about three articles a week. Oh, wow. topics in AI and technology. so if you just search for Jim Nardo, um, I noticed that my typed my name in wrong, my last name in wrong here. It's S-P-I-G-N-A-R-D-O. I'm the only Jim spi Bernardo on LinkedIn. Probably the only one in the world. Um, so you can find me there. my company is Pro Arc. Uh, it's, it looks like it's pronounced Pro Arch, but it's Pro Arc, P-R-O-A-R-C-H. and, uh, we have all of our services listed there. Um, we just did a webinar yesterday, similar exercise where we demonstrated building that hr, agent. so, um, there's lots of free content on there that, that you can devale yourself of. Um, I'm also coming out and I'm looking around my desk here, coming out with a, a book soon, an ebook that will be free to download, uh, on, it's called the AI Turning Point, which really discusses, how organizations have to, fix what's wrong in their businesses, to be able to take a best advantage of ai, uh, both from a business process perspective, but also from a business mindset as well.

Isar Meitis

awesome. Thank you so much. This was really well thought after and really well explained, so I really appreciate you taking the time and sharing with us and thanks everybody who joined us live. I appreciate you spending the time with us and, and learning with us, uh, while we're doing this. have a great rest of your day, everyone.

Head Shot

Isar Meitis

Host