Name: Static to Strategic: Transforming Contracting with Purpose-Built AI
Uploaded: 2025-09-15T22:16:07.057Z
Duration: 26 min 46 s
Description: Static to Strategic: Transforming Contracting with Purpose-Built AI

Transcript for "Static to Strategic: Transforming Contracting with Purpose-Built AI": Thank you so much, Jennifer. Alright. Hi, everybody. Today, we are going to be jumping into, a pretty interesting topic within the Agiloft platform. As we said, transforming contracting with purpose built AI. So when we think about AI, you know, Agiloft has a variety of out of the box AI functionality that helps you do a lot with your contracts from asking them questions to automating redlining. And you may have seen some of those, capabilities in recent Agiloft webinars. But today, we're gonna, we're gonna show how Agiloft can get a lot more specific to your exact needs when it comes to working with your contracts. And we're gonna see that using the AgileV GenAI Prompt Lab. So with that, we'll jump in and take a quick look at our agenda, that we're gonna be walking through today. And this is going to start by taking a look at and answering the question, what is the generative AI prompt lab? Then we're gonna see how you can actually use it and set it up. This is gonna be a know, slightly more technical demo. This is a very cool feature that gives you a lot of power, and it's on Agiloft's no code configurable platform. So we'll see how that works. And we're gonna see some example use cases of how the Agiloft generative AI prompt lab can be used. But, ultimately, the great thing about this capability is that it's up to your it's only limited by your imagination. So we'll see kinda why that's the case and how it works as we dive in. So first, let's answer the question, what is Agiloft's generative AI prompt lab? I wanna kind of preface this with a few examples that kind of, you know, dance around it before we hit it specifically. The generative AI prompt lab is a capability that enables you to use AI to ask any question or write any prompt about any of your contract data in Agiloft, including the documents themselves, and get the answer to that question or prompt put into a field in Agiloft. So to give you some, you know, analogies or some comparisons, imagine you were to take a contract document and drop it into ChatGPT and start asking a question about it. Right? That would be somewhat similar to what generative AI comp lab does. Alternatively, imagine you're using and I'm gonna jump into Agiloft here and open up an example contract, one that I prepared for our demo. Imagine you're opening up a document in Agiloft and you're previewing that document and using the Ask AI feature in Agiloft. This allows you to ask any questions of the document and get an answer right here within the system. This is also somewhat similar to what the generative AI prompt lab can do. But in both of those examples that I gave, you are kind of interacting with that document as a individual person asking an individual question and getting an answer. Where generative AI prompt lab kind of takes things to the next level is it allows you to do these types of prompts or questions in an automated fashion in the Agiloft platform. So let's see an actual example of the prompt lab in action. Here, I have it built in to an action button within Agiloft. Because Agiloft is a no code configurable platform, you absolutely have the capability to create metadata fields, create action buttons that trigger workflows, do updates, or in this case, generate an AI summary and description of our contract using the generative AI prompt lab. So to see that in action, I'll just go ahead and click that button here, and it is going to review the document, generate our description and summary, and populate that information into our corresponding fields within Agiloft. So, again, that still kind of, like, took an action from me as an individual user where I had to come in and click the button. But I think where the true power of the prompt lab lies is in the ability to automatically trigger things to happen based on a rule or an automation. So, for example, you might have a rule that says when a contract is first submitted into draft status, then automatically generate the description summary. Right? No button clicking needed. I like to just have the button click so it's kinda clear what's happening. But that could be a totally automated process kicked off by the prompt lab. Alternatively, once a contract moves into signed status, you know, maybe you want the AI to analyze how this compares against some other contract or analyze some other element of the contract. Again, that could be an automated process in Agiloft. And today, we are going to see how the generative AI prompt lab is set up, how it works, and how it gives you that flexibility to work in any way you need to. So just to jump back and kind of level set to ensure we're all on the same page, the answer to that first question, what is the generative AI prompt lab? Is it is an action within Agiloft that can either be used in a button or in an automated workflow to take any data in Agiloft, run that data against some AI prompt of your crafting, and output the AI answer into any field in Agiloft. So that's what the generative AI prompt lab fundamentally is. Now we're gonna see how you actually use it to start creating the prompts that you like. And, we're gonna actually start with an example prompt that is already in existence. We're gonna tweak this AI generated summary prompt that you saw me use just a moment ago, and we're gonna tweak it to contain some extra information as a great way for you to see kind of how the generative AI prompt lab works. So I'll go ahead and close this contract, and I'm gonna go into the setup of my Agiloft system. So we're going to the setup of my contracts here and open up my actions tab. And this is where that generative AI action, this is the prompt lab essentially. This is where that lives. And as you can see, it's one of many different types of actions like sending out emails, updating fields, you know, doing exports, etcetera. One of many actions, but the generative AI action is a newer action in Agiloft, and we call that the prompt lab. So I'm gonna scroll down to one of my existing actions, our generate AI summary action here. I'll go ahead and edit it. And we can see that it's made up of an AI template. So in this AI template, I can define what it is, how it works. I can choose which AI model it's using. You can see here out of the box today, Agiloft provides GPT 3.5 and Claude Haiku. In a future release, we will also be adding additional more, up to date models in this picker, but you can also use your own existing enterprise account from Amazon Bedrock, OpenAI, or Azure OpenAI by providing your own API keys. So this is a bring your own AI style feature. I'll select Claude Haiku here, and we'll move on to our next page. And this is where all the magic really happens. So it's pretty much a two step process here, and you can see it's, you know, laid out for me clearly on my page. I can design my prompt, and I can test the output. So I'll walk through this prompt, and we'll kinda break down everything that we're seeing here. So up at the top, we're giving kind of the preamble. You are a legal expert, and then what we wanted to do. Populate the below contract summary from the contract. The instructions in the square brackets below are to help you. Please remove them in your response. And then we provide the format for that contract summary, which is essentially telling the AI, you know, here's what we wanna get out of that document. Finally, we feed in the document itself. So here we're saying, here's the contract, and it says dollar sign latest attachment. This is in Agiloft, what we call a, field variable. And the cool thing about this is that you can actually easily if I were to delete this, you could easily insert those field variables with this formula help button right here. You can go to fields. And in this case, we do want to use our latest attachment field, and I would insert that in. But as you can see, this is how I could bring any data from the Agiloft system into my prompt. So just to recap our instructions, the, overview of the summary we want the AI to provide and feeding in the data, in this case, the contract document. So pretty easy to write a prompt here in Agile. And then step two, I'm gonna select the record that I'm going to test my prompt with. So I'll select the one we were just using and click generate. From here, the system is going to read the document, it's going to read the prompt, and it's going to write the answer. The cool thing about this is that you can always kind of review and make sure everything was correct. Here, when I click show process, I can see the entire input to the AI, including where the entire actual contract document. Scroll down and see a lot of information here. The entire document was fed into our AI prompt. But as you can see, I quickly got the AI answer I was looking for, including kind of every line in my summary. Now when a lot of people think about AI, they always think and kinda ask this question, well, how do you train the AI to do that? Right? And that's the cool thing about these modern foundational large language models is that they don't really require training in the same way that the old AI used to. This language that we're seeing in most legal documents is pretty common, and the large language model essentially just understands more or less what most of that means without additional specific training. So, for example, if we wanted to add something else to our summary, I could just, you know, hit a new line here and just type in, you know, limitation of liability with a colon and a space. And again, the AI, just by me adding that line here, you know, I'm not doing anything else behind the scenes. I'm not tweaking anything else. But just by adding to my example, summary output, telling the AI that I also want it to give me an overview of the limitation liability, it understands what I mean. And I'm gonna go ahead and click generate. And we'll see in my response here, I'm gonna get an overview of the limitation of liability. So that is the power of the prompt lab, which is effectively that I can ask any question, quickly tweak and test these prompts, and then, as I said, output that to any field in the system. So now if we go ahead and, you know, click finish here, I I just updated that prompt. We can jump back to our example contract that we were working on, and we'll go ahead and, you know, delete what the AI extracted before, and we'll have it do a another extraction. And this time, we'll see that limitation of liability that we just added as part of our prompt being extracted into the field. So, ultimately, again, I can easily create a prompt. I can feed any data in, and I can output that data to any field in Agiloft. It's a really quick and easy process that allows for a ton of flexibility and brings AI into the automation. Now as I said, Agiloft does have other fantastic out of the box AI features like Ask AI, like screens for AI, automated contract review. But if you wanna do something really specific and, you know, really geared towards your process, I think PropLab is an extremely powerful feature in order to kind of achieve whatever you would like using your contract data and AI. So with that, we are going to actually kind of create a new AI prompt and see what that process looks like, getting it set up on our contract. First, I wanna preview this contract and take a look at it and try to find something else that I might want to extract. So as I'm scrolling through here, you know, I might notice that I have, an SLA paragraph. Right? So my service level agreement, have all of this detail about, you know, uptimes and minimums and notices and penalties and stuff like that. And I really want an easy way to get that out of my contract so that if I wanna be able to search that information, I can easily find it in a a data field rather than having to kinda dig through the document to find that information. So we've kinda decided on our target. Right? We wanna get the SLA information out of this contract. I'm gonna go ahead and start creating a, in this case, a button. I'll I'll do a button again to pull that information out of our contract. So I'm going to first create a new destination field for that information to land. So I'll create a new text field and we'll call this SLA details. I'll make sure the field has enough space to fit all of my data. And just like that, I have created a new field to, put that information in. Next, I'm going to create a new action button, and this is going to be extract SLA details. There we go. And again, we'll take our button title, put it over here, and we'll now create our generative AI action. Pretty much, you know, creating the output field and creating the button are like the prerequisites, but now we actually wanna create the generative AI action itself. So we'll go ahead and add an action here and we'll choose our generative AI action. And again, we'll say extract SLA details and we'll create a new AI template here for this prompt. I like to name it related to the specific table that I'm doing it on for clarity, and we'll use our Cloud Haiku. Between GPT 3.5 and Cloud Haiku, Cloud Haiku has a larger context window, which means it can intake more information in the AI request, which means it handles larger documents better. And now we'll start designing our prompt. So we'll say, you are a legal expert. You will be given a contract to review. Please provide a short three to four sentence summary of the SLA and penalty details that you find in the contract. So we'll start kinda simply there, you know, without getting too specific. But, of course, as you saw before, if I wanted to provide a, format that I wanted it to give me the answer in, I could do that as well. And then finally, we'll provide the contract. Again, we'll go to formula help, and we'll use our latest attachment field. And voila, we've created our prompt, quick and easy. For our output field, we'll find our SLA details field that we just created, add that in, and then we will use a record to test and generate our AI output. Alright. So we're getting our answer. The service provider guarantees 99.9% uptime per calendar month, etcetera. And it's just giving me an overview of all of that. Now for the sake of brevity on our demo, I am having the AI kind of provide this to me in paragraph format. But if you wanted to get this data into multiple other metadata fields, you know, if you just wanted, you know, uptime requirement, 99.9 in one field, you know, some of these other numbers in individual fields, there are additional kind of configuration steps that we don't have time to cover on this webinar that you can do to get that information into the proper fields. The other thing that I like to do, and this is this is funny, you know, this is just kind of the nature of AI where it is today, is it always kinda frustrates me when it gives me this preamble. Here's a three to four sentence summary. Like, okay. I don't really want I don't want the preamble. So I like to add in, you know, please only provide the answer. Do not provide a preamble or any other commentary. And we'll regenerate. And now we have, you know, actually slightly different formatting this time. Again, that's the benefit of, if I wanted to provide some sort of, you know, format for the AI to aim for. The the benefit of that is that you'll get a more consistent output whereas, you know, I got a slightly different output. And that's kind of the nature of this AI is that it is probabilistic to some degree. So we're always gonna get a slightly different answer in the way it's formatted, but we are getting the content from within the document accurately. But as we can see now, we don't have the preamble where it's saying, you know, here's the answer that I got for you, which we we didn't really want. So we've created our output field. We saw how quick and easy that was. We've created an action button, and we've created our generative AI prompt to extract that information. We'll go ahead and click finish here. And now we have that action in our button. And the last thing we need to do is just add our button and our field to our contract layout. So it appears on the contract records. We'll just drop it right here and there and click finish. And now we have a new action button on our contract where we can hit that and have the AI, again, analyze our document to extract the information we are looking for from that contract. Now that this is here, it can be included in reports, it can be part of, search in a much more simple way, and you can use it in a variety of other ways in the system. So with that, we'll go ahead and just jump back to our agenda. We've covered kind of steps one and two, you know, what is the GenAI prompt lab? A great way for me to use prompts of my own creation on any data in Agiloft and output that to a field. And we saw how to use it, how to create a destination field. In this case, create an action button, although you could add this to a rule in Agile. And then create our prompt, feed data in, and flow that right into, the destination field. Finally, we're gonna see just a few other example use cases of GenAI PromptLab in action. I've got a few cool examples here. One of them is kind of a variation on that SLA details example. I wanted to come over to my finance tab. Maybe I want to extract some pricing details from my documents. So I'll go ahead and use my, prompt lab action button here to extract pricing details. And you can see that from that contract, it's extracting pricing details. And I even, asked it to give me an overview of the total cost of ownership over a three year period. So not only is it taking the pricing details that are in the document, it's doing kind of like an additional extrapolation that I asked it to do to give me on these different plans that were part of the contract. And I'll, I'll let you see kind of what it's pulling from here. There's a little pricing sheet down below. It's reviewing all of this and doing a total cost of ownership kind of summarization for me. So pretty cool capabilities there around, pricing summarizing. One last one that I really like and and we hear people request things like this, you know, pretty frequently is how can the AI help me actually generate contract documents? So Agiloft, of course, has very detailed and thorough template generation capabilities. And what we can do within this is we can use the output of the AI to feed into a template. So for a common example we see on this one, I have this MSA that is currently in, active status, and I can come to my renewal and related contracts tab and generate an amendment. And you'll notice here I've got AI amendment instructions on how I wanna generate this amendment. And I'm saying amend this contract to change the governing law to New York. So what this is actually gonna do is read the parent contract with AI and then generate a block of text to amend the governing law to New York, and then that text is going to flow through Agiloft's normal template generation capabilities to generate a simple amendment document for us. And it's gonna do that in one click. So I'll just say, you know, I typed in these instructions here, and then I'll just click create child contract. The system will read that parent document, generate the, text, and then generate a new amendment record, which was generated right here. And in this amendment record, we'll be able to see that we have a very simple template, you know, contract amendment, some of the details. But this text right here in the middle of the document was generated by AI. And then we just flowed that right into a normal template. But we're saying, you know, here's my original clause text, here's the amended clause text, and we're giving our overview of the purpose of this amendment. So, again, another great example of how you could use AI to actually help you generate a contract document. And I've seen people, you know, do this in a in a variety of similar use cases like generating termination letters or things like that. But the ability to really take that parent document, feed it into the AI prompt with some instructions and say, I wanna generate the text of a related document, I think is very cool. And it's certainly, you know, my simple example is just, you know, just the beginning of what I think is a pretty big untapped use case. So a lot of cool approaches on how you could use generative AI, PropLab, in Agiloft. Ultimately, as you can see, the rest of it is up to your imagination because you can ultimately feed in any data, run it through any prompt, and output it to anywhere in the system. There's a a lot a lot you can do here. So I hope that, this has really gotten your, you know, gears turning on how Agiloft's generative AI prompt lab, can help you out in automating some of your contract, processes with AI. So that covers it for my demo today. I do see we have a question here in the q and a from Emma. Emma asks, for the AI amendment, can this auto populate directly into the company template? Yes. In my demo example, the template that you just saw me use there is a really simple basic template that I created, but you can, set it up to populate into your own template. So it doesn't have to be a, you know, not not so pretty template like that. I I don't claim to be the, you know, template master, but you could have that output populate into any of the, any template that you have in your business. And Agiloft, does have the capability to work ultimately with any type of word document to make a template. Alright. If there are any more questions, now is your chance to put them in the q and a. I think we'll give it, thirty seconds. And if we don't see any questions, we will wrap up today. Thank you so much, everyone, for, attending..