Name: November GovDash Live
Uploaded: 2025-11-05T17:16:10.097Z
Duration: 28 min 26 s
Description: November GovDash Live

Transcript for "November GovDash Live": Alright. We're on we're at 11:00, so I'm gonna get started. Welcome, everybody. Happy Wednesday. Really excited to, continue this Govdash live and have some more of our engineering team join us. Today, we're gonna talk about how GovDash organizes, retrieves, and then generates proposal content, using your data. So really, really excited to get into this one. So those of you don't who don't know me, this is, our fifth, I believe, GovDash live, and we have our engineering team join us, monthly with different parts of the product. So I'm Lauren. I'm our product marketing manager, and I work with our, sales team, our customer success team, and our engineering team to communicate some of the, great, innovations that go into GovDash. And so I'll let our engineers introduce themselves. Hello, everyone. My name is Jaden. I'm a software engineer here at Govdash. Been working here about nine months, and I'm primarily working on the contract cloud and the data library. And I'm, Kyle Karunas. I've been working at Govdash now for a year and a half, primarily working on the proposal cloud, on the content generation system. Great. Thanks, guys. So before we get into the technical piece of today's webinar, I just want to go over, like, a high level overview of some of the really pain points of users, before a lot of innovative technologies have have hit the market. So, you know, using old processes and siloed systems creates a lot of data chaos, and that comp compounds throughout not only a team's process, but their all of their teams. So things like outdated CRMs that aren't really serving today's market, manual trackers like Excel where data is just, really everywhere. And, again, these processes can lead to not only burnt out teams, but also, like, in compliance, things that get get missed. So at Govdash, we've really taken all of these pain points from people in the industry, and tried to structure our technology to solve a lot of these, these problems. So, leading from that, AI can get rid of that data chaos. And that is one thing I would like everybody to take away from today's, webinar. We're gonna get into a lot of technical pieces, but, AI can really organize and automate a lot of that chaos that comes with having so many documents in government contracting. So, that's kind of the high level overview of really what we're gonna try to show you today. And now I'm gonna hand it off to to Jayden to kind of go into that, from the technical aspect. Okay. Thanks, Lauren. So, yeah, as she mentioned, I'm gonna be talking primarily about the data library, and I'm gonna get into some details about how the GovDash system processes documents. And kind of the the problem here is, like, how do we take these unstructured documents with lots of content and there's lots of different details? They may be very large. Some documentation is not always very clean. It might be kind of messy with difficult to read things and lots of, different data. And how do we take all of that and turn it into actual, like, structured intelligence and something that you can use, to get, like, better insights. So the four steps that we try to use here in document processing, we'd like to start just by analyzing the layout of the document, and we use optical character recognition for this. The primary goal is just to get every word on the page. So we're going to extract all of the text and all of its location within the document, but we're also going to look a little bit deeper into the document so that the GovDash system can see the document in the same way that a human could. So we're going to analyze, like, the structural layout of the document. So where are the paragraphs? What are the page numbers? Where are the headings and the subheadings? What are the titles? All of this information is going to be very useful as we continue on down the pipeline. The second step is something that I'm gonna show quite a few examples of as we keep talking today, and that's entity recognition. So part of this is going to be looking through all of these documents and finding, references to people and references to companies, maybe contract awards or just, like, structured details that we can actually grab from those documents. The third step is getting more information about those. So if we find a reference to a person in a document, maybe we can find more details like their email address, their phone number. If we are detecting a company, maybe we can get their office address or their website. And then we can also find relationships between these, maybe an employment record between a person and a company. And the last step is just to organize all of that data into a nice usable interface, and that is where the what you can actually see inside of GovDash. As you're clicking around, you can see all of the things that we've learned from your document, and you have full control over that information. So you can edit it, you can add to it, you can delete and view everything, and you can view all the citations of where we found that information. So just to get into an example, I wanted to show a very standard document that everyone will be familiar with and go through some examples of what the entities might look like. So this is just a copy of my resume, and I've highlighted a few of the things that GovDash might pick up on here. For example, it's gonna start with my name, and that's going to it's gonna identify me as a person, and then it's going to look at my skills, those different technologies that I've listed. It's going to attribute those to my profile within GovDash. It's also going to spot those companies, so Ensemble, Tap Sports, PenLink, and it's going to create structured records for those. It's also going to find employment records. That's where you can see it's gonna look at the job title, the start and end date. It's also gonna look at all those bullet points. If this were a different kind of document, maybe like a contract document, it'll also pick up on these things like government agencies, contract award details, maybe who is the company that is performing the work in this contract or who is the agency that is doing the awarding or the funding here. Those are just some examples of details that we might pull from the document. Here is an example of one of those contract documents and what it might look like within GovDash. So if you have a standard form, you could have people that are identified here, like Diane, who is, working as the contracting officer from Department of Homeland Security in this example document, and you also have Sarah, who's working as the project manager for Carmichael Industries. Gowdash is gonna be able to identify all of those and the relationships between them. And the final step of this is going to be the retrieval. So once we have all of this nice clean structured data, where the real value comes in is being able to use that data later on in your workflows. So one example is our, GovDash chat assistant, which is named Dash. And if you are chatting with Dash and you have questions about your data, Dash is able to tap into a tool called the data library retrieval agent. And this agent is able to intelligently search through all of this data structured in the way that I've been discussing. So the documents, it's going to be able to search through all of the text and understand maybe where certain things were referenced within the documents. It's also going to be able to search through those entities, people, awards, companies, things like that, that might be referenced in the document. The three modes that this agent uses while doing its search, it starts with just a basic keyword search. Super straightforward. If you search for an exact term, it's going to be able to find that. So maybe you look up a person by their name or their email address or a contract award by, it's like, task order number. It's also going to be able to use semantic search, which is maybe if you have a query that is a little bit less straightforward. So, for example, if you're looking through your the people in your data library and you're going to ask for who has experience with programming. And maybe my resume doesn't have the word programming on it, but it lists different programming languages or it says I have experience with software development. I'm we're going to be able to make that connection based on their semantic similarity between those terms. And then the third phase here is the agent can use the logical and agentic mode of search. So if you have a more complicated question that even the keyword or semantic might not pick up on, maybe you're asking it to identify who is the best fit for a certain role. It's going to be able to evaluate what are your needs that you're looking for and compare them exactly against the information to make sure that you get the most relevant response. And with that, I am actually going to jump into a demo here and just show what some of this looks like in the GovDash system. So let me pull this up. Here we have the data library. And if you look into the documents here, I've uploaded a few of these sample documents that I've been showing in the slides. So, for example, here's my resume, and you can see that it has picked up on, like, my name here. It's pulled me out of this document, and it's also found some of the companies that I worked for. And it's going to show you exactly where it found that information, which is something I think is very useful because some of these documents, obviously my resume is only one page, but some of your documents might be hundreds of pages. So being able to find exactly where something is referenced is super valuable. Another example here would be, like, this contract document, which is a little bit longer but still not nearly as long as some of the, like, real documents that we work with. And so if I wanna see everywhere that Diane is referenced, I can kind of just scroll through these different references here. It's going to pick up every sentence where that person is mentioned. It's going to find the different companies that are referenced and all of those details. And if we wanna get a little bit more information, we can always go view that person's page. So if we want to see what information do we have about Diane, we can open that up and we can see the system identified Diane's email address, phone number. It created an employment record mentioning that she's at the Department of Homeland Security. And we can see a little bit more information if you upload a resume. So I'm gonna jump over to the people tab, and maybe we'll filter down just to people with resumes. We'll take a look at Bailey, who is one of my coworkers. And we can see from Bailey's resume, we got his email address and phone number, but we also got all of these details from his employment. So we can see it's different employers, and these are kind of those line item bullet points that we would have grabbed from the resume itself along with start and end date, position, all of those details. And this is something else that I referenced that all of this is, like, completely editable. So you have control over all this data if maybe this is out of date. For example, my resume says that I still work in Ensemble because that's the last resume that I used when I applied to GovDash. So if you wanted to come in here and make those types of corrections the system would be blind to, you can always do that with GovDash. And then the last thing I wanted to show here is actually showing that retrieval portion. So if we open up GovDash, which is the chatbot here, we wanted to ask it a question like, find, personnel with experience in low level software development. It's going to be able to search through the data library and use those tools that I referenced, like the data library retrieval agent and searching through people. And it's very quickly able to find Bailey, who has plenty of experience working with these low level software, like, languages, and it's also going to find whoever else it thinks is relevant in this use case. And you can ask it a lot more complicated questions than this, but this is just, an example that I wanted to showcase here. One of the other main use cases I'm gonna stop my share here really quickly. But one of the other main use cases of this data library data is to use it in our content generation system, which is something that Kyle works on heavily. And he's been working on some really great advancements here, so I am going to pass it off to him to talk more about that. Thanks, Jaden. So now I wanna talk about how the content system in the proposal cloud, uses information from across the system, like the data library and capture cloud, etcetera. So the system starts by pulling in key details from the the solicitation package that you upload. Right? Things like full preparation instructions, task areas, evaluation criteria. And then what we allow you to do through the section editor, which you see a screenshot of here, we value to edit the writing plan. You can edit the citations. You can add documents you want this section to focus on. And then when we actually click generate proposal, what happens is we start to retrieve other information, right, from data library, contract cloud, and the capture question information that has been filled out earlier. And all this has come together to culminate the most best relevant content for the section at hand. This from there, the system runs across the entire proposal to make sure certain, compliance things are are accurate, right, like acronyms are consistent and certain quality standards, things that you couldn't reliably do with ChatGPT by itself. And the way we ensure these quality standards is we generate content at the sentence level. Right? So instead of getting a block of text that is chat g p t when you respond to it that is relatively opaque and it doesn't really say a lot besides that, you know, it gave back to you something that is at least somewhat relevant, what we do is we write the sentences and then have the metadata or we call metadata attached to each of those sentences. So that includes what proposal preparation instructions were used or addressed, what task areas are being addressed, what, examples from the data library and contract cloud have actually been, you know, sourced to write this sentence. And this allows us to make measurable improvements to content, And it also allows us to measure content in concrete ways. Like, for instance, we can see how many task areas how many unique task areas were addressed, which proposal preparation instructions were addressed. We can run evaluation, criteria to simulate how an evaluator would view a section, and then we can average all these things out to create a composite score of the proposal. And, right now, I'm focused on make I'm building a way we can expose this to customers. So I wanna share a quick little, demo, to be relatively short, of of what what this could potentially potentially look like on, production. So this is in my development environment here, and this is like an outline. Right? I'm sure people have seen this. You have a section. Right? And there's basically a little score up here. This one's only 70% mostly because you can see I have a lot of task areas to cover. So sometimes bumping the page count would help here, say, allow it to write more. These are sentences that would potentially need more supporting evidence. In this case, there isn't any. This is a technical section, so it's a little bit harsh on the evaluation criteria, and it doesn't actually say that, there's any implementation details, so it needs more supporting evidence there. Other sections like these or whatnot, you can see that, we hit all the task areas. We hit all the coverage, but it might be a little weak and how an evaluator would view it. So things like that are what we wanna bring to the product. So and that's, sort of what's up next at least for the proposal cloud. That's gonna be coming very, very soon. I wanna build it more interactively as well so that we can interact with the sentences directly and see what was used on them and what was cited. And I'm gonna flip it back over to Jaden so we can talk about what's coming in Contract Cloud. Alright. Thanks, Kyle. So a couple of the items up on the road map for contract cloud. I know I spent most of today talking about data library, but I also primarily work on contract cloud. And this is an area that we are really going, like, full speed ahead trying to develop within GovDash. We have lots lined up on the road map. And one of the things that is, I've been working on for the past several weeks and I'm hoping to release here in the next two weeks is, contract modification tracking. So as we are dealing with these contract records and the government continue continues to put out modifications that may be changing the obligated funding or different details about the award, maybe changing the periods of performance or maybe exercising another option. All of those details are important to track within GovDash. And they're also, like, very useful to see broken down into a nice table so that you don't have to maintain some sort of spreadsheet to track all this information on your end. We can just keep it all in the system, and you get all these nice benefits like Dash can understand this information, and the system gets a broader understanding of, like, how far into this contract are you, what parts of the PWS have you actually completed. So that contract modification tracking is coming up soon. Definitely something to look out for. Next up after that is going to be task area extraction. So we'd like to be able to look at, like, the PWS document within that contract package and break it out into all of its task areas, show those in a nice, like, hierarchical table. Again, trying to avoid something that you might have to do in a spreadsheet manually outside of GovDash. We'd like to bring that in house and give a few more, like, tagging features onto those task areas. For example, we have, some people who would like to mark certain task areas as not yet completed or in progress, or you may if you're a subcontractor and you want to mark, like, hey. As a subcontractor, I only completed a few of these task areas. You can actually do that and you can say these other ones, we didn't do that work. You shouldn't write to it assuming that we did that work. And the final step in the road map here it shouldn't say the final step, but the next thing up on the road map is deliverable management, which would be looking through these contract documents and breaking out all of the deliverables and dates and requirements that, are needed from that contract and making sure we can track those and help you meet those deliverables. So if you have certain tasks that are due, we can provide reminders and show them in a calendar, and then we can start to work with Dash and make dedicated workflows to help you achieve those deliver deliverables. So just a quick overview of a few things that are upcoming. Definitely stay on the lookout for that, and I think we're hoping to have that as the subject of a future webinar. Great. Thanks, Jaden. So in order to find all of these, updates that the engineers just talked about, there's a few different spots. So if you're a current user, we have a change log release that goes out weekly. So if you go to your settings tab in GovDash and then navigate to what's new, there you'll find a full list of all the different features, enhancements, or even bug fixes that go into the platform. Because we really wanna make sure that everybody knows all the new features since we're releasing so much, so many innovative features. As well, you'll get a email when big features come out. There should have been one to your inbox this morning about, SLED being integrated into our bid match. Really exciting release there. So there's a ton of information, about that on our website as well. And then webinars, we host pretty much weekly webinars at GovDash, from a range of topics, not only product driven, but industry, leaders and also, pro team members, which, they talk about a lot of great things on how to utilize the platform, and automate some of the the, workflows that government contracting teams are, wanting to automate. And then if you wanna see more road map items, further than what Jaden and Kyle just just mentioned, you can go to gov-.com/roadmap, and see all of the great things that are gonna be, expanding in every part of our product. Alright. So now I'm gonna open it up to q and a. So if anyone has questions, you can drop them in the q and a spot on your left hand side of your screen. So I'll wait and see if anybody has some questions for the engineering team here. Alright, Jaden. I have a couple questions for you to get started. Okay. So what are the main differences between the data library and the contract cloud? Yes. That's an excellent question. And since I work so closely on both, it's definitely there are some overlapping, like, pieces of functionality here. And I think the best way to think about it is that the contract cloud is very, like it's a very dedicated place just for contract awards. So I mentioned the different entities that GovDash can kind of extract and deal with from documents, like treating your key personnel or companies. And awards is one that's big enough that it has its own dedicated cloud, and that's where contract cloud comes in. The main thing that you do there is you manage your active contracts. So that's where a lot of those features we talked about on the road map come in with contract modification tracking and task areas and deliverables. These should all be, like, living things that may change and need adjustments, throughout the life cycle of that contract. So there's a lot of dedicated features there versus the data library is more intended as a a general knowledge base that can feed into other parts of the product. So if you upload people's resumes, you can use them to identify the best key personnel to use while writing your proposal. Or if you upload certain, like, company records, you might be able to link those as, like, teaming partners if you're working on a capture opportunity. Awesome. Thanks, Jaden. Kyle, I have a question for you, from Shannon. When when will the content quality, scoring system within the outline section be going live? Very soon, hopefully. Thank you. And then Bill asked, did I hear that GovDash is building its own CRM? I can actually take this one. So our capture cloud is already functioning as a CRM for a lot of teams already using GovDash, and a lot of them are actually, leaning off of their their old CRMs and using GovDash. So, yes, we are building that out as we speak. Let's see. I have a couple more here. How do you verify that the information from AI is correct? So, Jayden, I don't know if you wanna talk a little bit about that. Yeah. Absolutely. That's definitely, it's definitely a tough problem to solve because as I'm sure a lot of people have noticed, like, using AI tools like ChatGPT, it's kind of a black box sometimes. It may cite its sources. It may not. And sometimes you don't always know where it's getting its information. And the way that we like to view this in GovDash, I think there are, like, two main ways that we tackle this problem. And the first is, like, always trying to cite our sources. So as I was showing in the data library when I was kind of clicking through if we're looking at a document details page and that document, it says it is attached to different people, different companies, you can click on those and view exactly what sentences from the document we found that referenced that person. So you can always go in and check our work. Kind of the it's like the best analogy I probably have here is if you had, like, a difficult math problem that you're working on and you have two friends that are gonna help you with the problem, and one of them just gives you a number back. And they're like, trust me. This is the answer. The other one shows all their work exactly how they got to that number. GovDash is trying to be that friend that shows all of the work and make sure that you have full confidence to verify that result. That way, you are always, like, in the driver's seat. And the other main thing that we try to do is what Kyle has been talking about, where after the fact, after we generate a response or some sort of content, we take another look at it and we implement some sort of scoring. So we evaluate it against various criteria and make sure that we're actually checking those boxes. And And if for some reason it's not checking those boxes, we make it very clear so that you can take action and step in to make sure that your information is always compliant. Great. Thanks, Jaden. Kyle, this one will be relevant for you. So before the review feature to the proposal cloud comes out, what's the best way to determine which areas need more deeper focus content to add before proposal generation? Right now, I believe we do highlight certain things in Word if, we detect that there should be more there. But also utilizing the Word Assistant is a great way to, interact with the proposal content once it's been generated. Great. And then, Jaden, I have a question from Lee. Does GovDash currently help users find contracts that match the SOW and opportunities that they're looking for? It does. Yeah. So if you, if you have a certain thing that you are looking for, you have certain requirements, there are, like, various ways you can try to find the best contract in your contract cloud that meets those criteria. And that kind of comes into what I was talking about where we use those three different modes of, like, similarity while we're searching. So we can look at keywords that match. So if you have certain requirements in your PWS, we can match those by keyword. We can match them with semantic similarity, and then we can also use, like, agentic or AI knowledge to make sure that what you are looking for is met by what's in the contract club. Awesome. These are great questions. I think we don't have any more, but feel free to drop some in the next couple minutes. We're wrapping up a bit early. Alright. I think that's it. But, yeah, thanks everybody for attending. Really great stuff today, talking about more of the technical aspects. I hope you took something away from this webinar, and, look look forward to the next ones, coming up in November. Alright. Thanks, guys.