What’s coming in VisibleThread

Hello and welcome to today’s webinar. My name is Micheál McGrath, I’m the VP of Marketing here at VisibleThread and I’m delighted to welcome you today. Wherever you’re coming from, thank you for taking your time. We know you’re all busy and we appreciate your time here and we hope you enjoy this webinar. Today’s webinar is a product webinar. We’re going to show you some live demos and demonstrations of the product and use. We’re going to talk about future VisibleThread, the market that we’re in today, and we’re going to talk about where we’re going. Just to start off, we have a attendee chat on screen, so if you’d like to network with other people on the webinar, we actually have a ton of registrants here today, so if you want to let us know where you’re coming from, where you’re tuning in from, I’m here in Cork, in Ireland, and we’d be delighted to see who’s with us. And if you have any questions at all throughout the webinar, there’s a Q and A box on screen, You can pull that out, and we’ll have time at the end for live q and a, but we’ll also tackle questions in the moment if it makes sense. Okay? So right. So in today’s agenda, we will talk about the market in general. We’re talking about the the software market, the technology market. We’ll be talking about the foundational platform workflow behind VisibleThread and the of the foundational blocks that makes our product tick. We’ll look at live demonstrations of the product. We’ll work through a BD and capture workflow, and we’ll also work through the proposal development workflow as well. We’re very excited to talk about our MCP server with our CTO, Andrew, so that’s a very exciting update coming in the future of VisibleThread, we’re excited to show a live demonstration of that here today. So without further ado, just to start off the webinar and set a little bit of context, we have a poll for you, the audience. So you can actually click the slide itself, you can click these and it’s a multi select, you can select which ones affect you the most. So which of these would you most want to improve with technology? So requirements management or compliance, opportunity assessment, intelligence, content creation, workflow. So which of these items are you most looking forward to kind of improving with technology or AI or or any factor that we might discuss today? If it is an other, feel free to drop it into the attendee chat. We’d love to hear your comments and what you think is missing from this list. While you’re filling that out and completing that survey, I can see about half of you have done so already, which is great. I’ll ask Fergal to introduce yourself first as one of our first speakers. Thanks, Micheál. Great to have everybody on the webinar, and hopefully you’ll find it engaging and and interesting. So my name is Fergal McGovern. I’m the founder and chief product officer at VisibleThread. I guide product direction, I work with strategic accounts to make sure that we’re addressing the challenges that our very strategic accounts and also our small business accounts, and mid sized accounts have, and making sure that we help them understand the value of technology and which technology makes sense in their context. Allison, you’re no stranger to a VisibleThread webinar, but we’ll definitely have you introduce yourself as well. Hello, back again for everybody who is joining us that’s been here before. My name is Allison Ritz. I am the director of product marketing at VisibleThread. So I have the unique opportunity to work with our customers, to work with other folks in the market and really understand the problems that we’re trying to solve and then work with the product team to try to help those become a reality. So I’m excited today to talk about what we have coming and what our roadmap looks like and what you can expect from us for the rest of this year and beyond. Fabulous and making your debut Andrew, your first VisibleThread webinar. Would you like to do an introduction? Yes, thanks Micheál. So my name is Andrew Locatelli Woodcock. I’m the CTO of VisibleThread and I’m thrilled to be here. I would run a software engineering and the IT function within VisibleThread. And so I’m very, very excited about the stuff that’s in the pipeline and gonna be released over the next month or so. And very much looking forward to sharing just a small snapshot of what we got coming. Very exciting stuff. Fabulous. And just to highlight once again, if you have any questions at all throughout the webinar for any of us here, feel free to drop it into the attendee chat or the Q box. And just commenting on the poll selection, so thank you for filling that out. And yeah, it’s leaning towards requirement management and requirements and compliance. That’s something that we’re going to cover off a lot today. It’s the foundation of a lot of what VisibleThread has done for a long, long time and for a big reason why people love using the platform and have invested in the platform for well over fifteen years. But what’s interesting, lower beneath is opportunity intelligence, which Fergal is going to do a lot on today, and content creation and workflow analysis, also stuff that we’re going to cover on that. But it’s great to see of the variance of the answers here. Before we get running, just to kick off the conversation, we want to talk about the market today. G2 Crowd is a software listing site, and within our space of RFP software, there’s over one hundred and seventy five vendors within that space. That has dramatically increased over the last three to five years from all different cohorts, from our advice software to CRMs, response software to pricing tools, all fitting in underneath this bracket, and we find the market is hectic and confusing, and what you’re noticing now is vendors almost quarterly appearing in the markets and maybe some quietly disappearing from the market, some consolidation and joining together, some that have raised venture capital funds, etcetera, a lot of noise, a lot of movement. And I suppose the main question that comes to us is who to trust and why trust VisibleThread over others. Our backing behind that is our fifteen years in business and the products that we’ve built over fifteen years and what we’re looking forward to building over the next fifteen years. Fergal is going to introduce the start of this session by sharing a little bit about what we’ve built to date, while also layering in on top of, you know, what’s kinda coming next and what’s kinda currently just live. So Fergal, I’ll pass it over to you. Thanks, Micheál. And and just kind of kinda compound what Micheál has just said, we we are noticing very clear trends in the market. Certainly since ChatGPT hit the scene late twenty twenty two, there has been a proliferation of AI first, quote unquote, products entering the fray. Most people in the space of bid, capture, proposal development, contracts analysis have kind of come become comfortable with the AI side of things, so that’s great. But there’s a ton of these products out there. And it’s not that they’re all bad. There’s some really impressive products out there. But for you, as somebody trying to make a decision as to which automation you will apply, it’s really kinda tricky right now. I have a lot of sympathy for anybody making that decision. We would just suggest the following. If you’re analyzing and looking at solutions, make sure that you trial the stuff for real. Make sure also that you actually use the latest and greatest of any of our solutions. So right now, over the last year and a half, when I’ve been leading the product side of VisibleThread, we have evolved massively. We are now five times faster in our product development side of things. Andrew is leading a very lean and very agile team. So we are delivering now on almost a monthly basis. So what you might have seen before is not what we have today. And part of the reason we’re doing this webinar is to help you understand that. So to that extent, let me just kind of click on. Big difference between kinda legacy VT, VisibleThread, the stuff that you might have seen two, three years ago, and some of you indeed may still be running these products. The big difference between VT Docs and our former product VT Writer, which is still out there, is that we have consolidated everything into one system of record, one solution. So it’s a system of record across the whole life cycle. For any of you who are existing customers, this may look strange to you. It’s different. We announced and delivered on this in March of this year. So this is VisibleThread seven. It’s a natural upgrade to VT Docs, And it allows you manage from end to end. It is not a stitched together environment. It is a very coherent, very usable kind of journey from the start opportunity identification, right the whole way through delivery. So to that end, we basically characterize it as following a lifecycle. So here is an example. And we’ll get into this in live in a few minutes. But this is the idea of actually identifying at the top of the funnel what opportunities might be of interest for us. In this case, we’re showing a screenshot of SAM dot gov. We also support eBuy if you’re on vehicles that are exposed to eBuy. Again, if you’re in the UK space, we have a direct integration with Find a Tender. All of this is new. We are also layering in more and more feeds. So by the end of this year, you’ll see state and local education feeds, and you’ll see additional geography feeds. So finding that opportunity that makes sense, understanding where it aligns with your business profile, and then pushing it forward into the bid no bid decisioning process. This is a view that we think is very important. This is one of the analysis reports that we ship out of the box. It is an alignment view. And what it’s doing very simply, and again, I will demo this in live terms, is it is taking the requirements from the actual solicitation, and it is mapping those to contracts that we may have worked before. So effectively, it’s a past performance matrix. You’ll see here in this case, the second column on the right hand side is showing a seventy eight percent alignment and lots of nice green stuff going on there. So this is important, and this is highly traceable. I’ll talk about traceability in a moment. Again, we see this expanding into multiple touch points within the solution. Proposal alignment is a really good example of where that will actually kick in as well, but more later on that. So bidding and making that decision, it’s very important that we actually spend our BMP dollars in the right place. Once we actually are satisfied that we are qualified to bid something, then we push it into a proposal lifecycle. Here, we’re seeing a stage gate approach. You see a Kanban board on the right. You can basically customize that, and you can start to flow through the various stage gates in your proposal development process. You might have a stage gate for helping you do you know, if you’re a Shipley shop or you’re a low fat consulting shop, whatever you might deem is a reasonable process, you may have things called red teams, pink teams, gold teams, all customizable, all configurable, and you manage that through the lifecycle. And the final thing overlaying all of this is the idea of the contract side of the fence. So for us, contracts is super important as a constituency. Contracts people need to be looking at proposals or possible solicitations that we may pursue early and often to coin the phrase or to steal the phrase, let’s say. So this type of view would be, for instance, something we will see early on in the cycle, we’re running shreds in a one hundred percent accurate and repeatable way to actually isolate areas that might potentially be risky when in the pursuit. So liquidated damages, indemnifications, that kind of stuff in this case. Again, we have evolved this massively if you’re used to some of our older versions of the products, and we’ll show this live in a moment. So for us, when we draw back, there are three things that we really kind of want to make clear in terms of our philosophy. So as we kind of build technology, we don’t want to build technology for technology’s sake, We want to make sure that when we build, it’s one system of record, and then there’s analytics layered on that system of record. So system record is a fancy way of saying, we want to support you on your workflow, we want to make sure that you have auditability, that you have traceability, and that you actually support the entire lifecycle progression stages that I just highlighted. The second element here is that we do not believe that we will throw AI at everything. We have been using AI in VisibleThread since twenty thirteen. Our discovery reports are powered by a type of AI called natural language processing. Generative AI is extremely powerful, but it’s one type of AI. And it’s really great when you have a need to find a good enough answer, summarize a document, give me an insight from a kind of win theme perspective as to potentially what the contractor or what the government is looking for, etcetera, etcetera. On the other hand, when you need one hundred percent accuracy, and you can’t afford to miss a requirement, that’s where you need what’s known as deterministic software. This is the stuff a bit like control f or find in Microsoft Word. If you want to find accurately certain occurrences of terminology, you should not use generative AI, because it is a probabilistic guessing game. So it is very important that we think about what we are trying to solve for and what technology we wanna use for that problem. The third and final part, and we’ve always been like this since day one, since I started the company well over fifteen years ago, we’ve always been built for sensitive work. We are FedRAMP ready. So if you want to use our GovCloud, that is now available. If you want to go completely air gapped, that is completely supported. Many of our biggest customers are completely air gapped and bring in their customer environments. So we support multiple different deployment times. And as you look at vendors, these are things that you build over time, you build a muscle memory for this over years. It’s not something that’s easy to do, but it’s something that you should definitely be considering. Brilliant, so what I’m going to do now is I’m going to pass it over to Allison, and Allison is going to take us through the next stage of the presentation. Excellent. And I am going to have Micheál advance the slides for me because I don’t have them on my end, but he’s going to go ahead and advance slides. So the next oh, are we good? No, that’s good. Okay, great. Thank you. So today what we’re going to talk about in this front end is really the engine that is underneath the platform. So understanding the components that make up the intelligence engine of VisibleThread. So those are going to fit into kind of three buckets. The first will be workspaces. So how do we segment our work and how do we segment our teams based on workflow collaborative projects, security, those types of things. So how do we create spaces that are most effective for the way that we work? And then the two kind of engines, our generative AI engine and our deterministic engine. So we’ve talked before, I know I have ad nauseam for some of you probably about and Fergal just mentioned our tech approach. So the combination workflow is the most effective way to leverage Gen AI in a proposal process. So making sure that we have guardrails to the Gen AI and applying things where they are most effective. So on the generative side of the house, we’re talking about our prompt library and our collections, how we can help standardize, how we can, make sure that we’re sourcing from our trusted data. And then on the flip side of that coin, we’re thinking about the deterministic engine. So our search dictionaries, our watchword list, how we can make sure that we are one hundred percent accurate where we need to be, and then also make sure that we’re ensuring quality in the back end of the content and work that we’re creating. So next. Okay. Are we Okay, sorry about that. So, thinking about, as I kind of get into each of the structures, I want you to think about this from an overall kind of look. So thinking about creating a foundation that mirrors your structure. So workspaces mapped to business units, programs, customers, looking at the way your organization is already shaped. So, we want to make sure that you’re being supported effectively and segmented in the way that you need to be. Collections are using the content that you already have. So, you have trusted content that is approved and in place to be used to source your new content. So when Gen AI is creating documentation, we’re making sure that that’s coming from a place that not only is trusted, but is referenceable and defensible. And then looking at search dictionaries, watchword lists, a lot of the information that we’re going to source here is institutional knowledge. So we have people working through these processes. They know what red flags look like. They know what they need to find, but that manual process is incredibly time consuming and it can introduce a level of risk. So how can we operationalize that in a way that is scalable? And then making sure that all of this runs inside your workflow. So the Kanban board to manage capture process and we’ll talk downstream about what that looks like in our release coming in August to manage specific types of reviews, collaboration, workflow management, all of those things happening within the platform to support the way that you already work and make sure that you’re optimizing technology effectively. Next. Excellent. So the five building blocks here, and I mentioned this on the top end, but that’s what this looks like in practice. So our workspace is our operational container. So our each workspace is unique from a organizational profile perspective, a customizations perspective with unique dictionaries, prompts, collections management. All of that is a completely unique workspace that users can move between. So if you have users on different projects, users in different spaces, users can move between, but you’re only seeing what you need to see to make sure that it supports the work that you are doing. So our workspace is our container and within we have our prompt library collection. So there we’re talking about repeatable AI workflows. So operationalizing our prompts, putting a level of standardization into the content that we’re creating, and then grounding all of that in collections in our trusted data to make sure that the results are repeatable and defensible. And then for our deterministic piece, our search dictionaries, like I said, and our watchword list, we’re making sure we’re being precise and making sure that we’re controlling terminology that we’re achieving clarity in the most effective way. Next. Okay. So first I want to talk about workspaces and I’ll do a brief kind of segment on each of these. So our workspace is going to be the secure self contained environment where you are going to do all of your work, where you’re going to start in the front end and assess your opportunities, we’re going to do analysis, we’re going to build our proposals, we’re going to manage our proposals throughout the process. The workspaces can be partitioned by team, program, security level, commercial spaces. So ultimately, folks are only seeing what they need to see. So of course, this is a big win from a security perspective, but also from a workflow and a collaboration perspective as well. So Workspaces, like I said, can be accessed by a single license. One person can access multiple Workspaces And this is the foundation that we’re creating to make sure that you’re protected from a security perspective, that we’re meeting you where you are from a workflow perspective and only showing you the things that are most relevant. And we are supporting you from a prompt dictionary, a business profile perspective for the things that are most relevant to your specific line of business or your group to optimize the level of content or optimize the product that you’re creating in the end. Next. So next I want to talk about the prompt library. So what does the prompt library do? The prompt library helps to operationalize prompting and standardize the application of prompts across your team or your working groups. Now, is incredibly important because when you are working with a team, if it’s a few contributors, if it’s ten, fifteen, twenty beyond contributors, and people are free prompting, they’re using different systems or they’re approaching the way that they’re creating content in a different way. What you’re going to get is a multi contributor response when you start to put this together that is incredibly disjointed. Additionally, depending on the way that you’re prompting and depending on the way that you’re structuring those prompts, the quality of what you’re creating can be very different. So a similar challenge that we had when we were creating things manually that you have a multitude of SMEs and contributors creating content and we need to make sure that we are that this looks like one single response and not chapters of a book written by different authors. That can be enhanced or a more sort of aggressive out putter version of that when we are inserting prompts into the process, depending on what that structure looks like. So, is a way to not only standardize across teams and create content that is structured in the appropriate ways, but it also helps to upscale people on your team that may not be your kind of top prompters or people that haven’t leaned into this as much. So it’s a knowledge sharing capacity. It’s a way to put regulation around the prompts and the structures that we’re using and overarchingly a way to improve the quality of the content that you’re creating. Next. So the next kind of other piece of that when we’re talking about the generative AI engine or the generative engine, the generative piece of it is collections. So I touched on this in the front end and there are a couple of things that I wanna point out about collections. So collections are really curated sets of approved reference material that the AI is gonna draw on to analyze and create content. Now, is really important because a lot of folks are sourcing from SharePoint. So that’s a pretty common thing right now that we have our system, our LLM, whatever we’re using to create content is connected to our SharePoint or connects to our repository in some capacity. The danger that you have there is that AI doesn’t have the ability to understand what is good content and what is bad content. If I pointed at a folder, it’s going to judge everything equally and it’s going to assume that the information that I have pointed it towards is accurate and can be used as source material. Well, when we think about the state of some of our SharePoint folders or some of our larger repositories, even if they’re in good shape, have a lot of information that then AI needs to wade through and make decisions, on its So ultimately collections acts as a bit of a buffer there that even if your SharePoint isn’t perfectly optimized today, we can create collections of content that are trustworthy, that are topical. So we’re not pointing it at a huge folder and trying to pull out little bits of information. So this information is topical. It is the most relevant content that we’re using and we know it’s trusted and we know that it’s the resource material we want to, reference. So ultimately collections gives you a grounded space that can be referenced, that can be traced and that you can work with to make sure that your content is accurate and defensible. Next. Okay. And so I want to kind of flip the other side of the coin here and talk about the deterministic piece. So I know that so much of the conversation right now is around GenAI and all the things it can do and how exciting that is, But it’s equally or even more important to have a deterministic side of that coin to create guardrails and to create safety nets when we are dealing with jobs that have to be a hundred percent accurate. Now, if I’m creating first draft content, I’m doing an analysis of an opportunity, I’m doing a market analysis or questioning my positioning. These are really great use cases for Jet AI. But when we’re thinking about risk, thinking about requirements identification, thinking about specific terminology that could disqualify or change the way that we would approach a bid, we have to be one hundred percent accurate. So we have to know that we’re going to be able to find these things every time. And this review process needs to be repeatable and scalable. So repeatability, scalability always hinge in accuracy. So we have to make sure that our process is accurate before we can appropriately scale it. With search dictionaries, these are curated terms of risk clauses, specific language, industry specific language, customer language that are going to indicate risk or indicate some sort of action. With approaching this in a deterministic way, we can make sure there’s no interpretation here. There is no hallucination from a technical perspective. It is not possible. Every match will show up the same way every time. So regardless of who runs this against a specific dictionary, they’re going to get the same results. So you can trust that the trust in the accuracy of the reports that are being generated. Next. And then finally the watchword list. So this is the last piece of this kind of deterministic side of the house. So watchwords are going to come in on the back end. So if we’re talking about the kind of sandwich or the bridge in this deterministic GenAI combination workflow, we’re going to use GenAI and I’ll show this a little bit later, or excuse me, we’re going to use deterministic technology to identify our requirements. We’re going to help that to inform our compliance matrix and then inform our outline. We’re going to use GenAI for outline content creation, but then we’re going to circle back. So once we’ve created this content, we have to make sure that it is at the clarity and readability level that we want it to be. We need to also look at the terminology. So whether it is sensitive terms, whether it’s industry terminology, customer terminology, we to make sure that we’re adhering to our style guides and to our quality standards with the generated content that we create. Now, that review can be even harder nowadays because in the past, if content was a little bit rough, we could tell content was a little bit rough, but a Gen AI is really good at making very plausible, reasonable quality wrong content. So this piece of it becomes more important than ever because we are doing those checks to make sure that this is meeting the quality and the standards that we need and everything is correct. And I like, again, I said, I’ll walk through that in the back end. But ultimately that is the overview of the engine within VisibleThread. So, workspaces to make sure that from a security perspective, from a focus perspective, from a business unit perspective, that we are in the space that we need to and we’re following those workflows appropriately and then incorporating the Gen AI support with the deterministic engines to create an accurate work accurate, foundation and then allow GenAI to do the things that it does best by applying it in the appropriate spots. So what we’re going to do now is we’re going to jump into a couple workflows to show you how we leverage some of the engine, leverage some of the foundation of VisibleThread and in specific workflows and we’ll show that combination deterministic and GenAI approach. So I’m gonna hand it over to Fergal to walk through our BD Capture workflow. Great. Thanks so much, Allison. Let me attempt to share my screen. Hopefully, this will come off. Just give me one moment. And I just want you to confirm maybe you might confirm me all that you can see my screen. I see it. Great. Thanks so much. Fantastic. So to Allison’s point, she talked about workspaces. So much, Micheál. She talked about workspaces just so that you see my environment here. These I’m operating in a workspace that happens to be labeled US government workspace. This is the set of workspaces I can get into. So I could go into a different workspace. And that’s very important because perhaps I’m allowed to get into certain workspaces that my colleagues should not. So it’s really a way of kind of segmenting your data. Workspaces tend to be line of business oriented, they tend to be for instance, in the US federal context, I might have a CUI workspace, I might have a more general workspace for non CUI stuff, etc. So that’s workspaces. And just to kind of put flesh in the bones of what Allison was talking about, she talked about the prompt library. So this config area on the left, that’s where we set up our prompts, you can configure that. She talked about collections, which is the idea of guardrailing the actual AI to make sure you’re looking at the best quality content. She talked about search dictionaries. So these are kind of fundamental building blocks of the engine. That’s how we we think about it as the engine that we’ll see in the life cycle in a moment. And she also mentioned watchwords, which is around the writing point and putting guardrails around the written word. So for instance, I do not wanna use the word expert because that could be legally risky. So it’s very important that we actually don’t use that, irrespective of whether it’s AI generated or whether it’s me just writing. So that’s the idea of those core engine components all living in this one workspace. So I’m gonna focus on BD capture, and I start with the idea of actually sourcing opportunities that could be of interest to me. You will see on the right hand side, we’ve got a number of tabs. The first one is SAM dot gov. The second is ebuy. Should I be on certain vehicles? So for instance, in this case, I’m on a vehicle, Oasis Plus, MaaS. One of our great customers gave us permission to use their ebuy feed, which is fantastic. And then for any of our colleagues in the UK side, this is an automatic feed from find a tender. As we build out towards the second half of the year, you will see more and more tabs coming in here, including SET, etcetera. Let me just focus on SAM dot gov because the principle is the same. You’ll see there’s a column here called business alignment. Business alignment is useful because it helps identify those opportunities that are aligned with what our business requires. So if I sort on that, you’ll see I’ve got highly aligned opportunities here. This defense optical fabrication system is aligned because and I’m now going into the detail of it, because we’re finding that the DOD is one of our customers. This is our org profile on the right. So we’re finding a high degree of alignment. That’s a very important configuration, and each of these feeds will have its own alignment characteristics. Super. What’s also interesting here is that you might have certain searches that you wanna run repeatedly. In my case, maybe I want all DOD and IT services contracts. You’ll see now I filtered down the three hundred and seventy eight thousand odd opportunities down to ninety seven that match these NAICS codes in the US context, and this agency. And again, you can add more filters to this. Once you do this, you will start to track an opportunity. So this pushes things on from the large potential pool of opportunities into specific kind of push them into a life cycle. So here’s my tracked opportunities area. In my case, I’ve got thirty two tracked opportunities. As we look at this, we will be introducing also very soon the idea of gates from an opportunity, life cycle standpoint. So you’ll see that coming in the next, month. And here, we can see that there’s a number of icons around these opportunities. So I’m just gonna do a quick search on an opportunity that I know I’ve been working, which is the health care IT support for Extract Lab Management System. You’ll see quite a few icons here. It tells me how much activity is going on. You can also see where the source of this was. I can see a mix of Sound dot gov, some manual stuff, some find a tender stuff at play here as well. When I drill into this opportunity, now we have a an ability to see the documents associated with the opportunity, which is up here. And if this was ingested from SAM, we would automatically pull in these, documents. And you can also see that there’s a side menu bar here allowing us to do certain things. So again, we’re at the pre go midpoint in the cycle, we’re in the evaluation cycle. We’re evaluating this for, you know, certain characteristics. So one of the first things that we will look at is technical alignment. And technical alignment is where we’re basically saying, look, what is in the solicitation that we’re evaluating. So we are automatically using an ability to distill the core concepts at play here. We use a mix of technologies here. We use Gen AI to allow us to categorize certain things. And then we also double check everything that we’re finding to make sure it exists and that the AI is not hallucinating. So you’ll see software development is a big theme here. Why are we seeing software development as a theme? Because, basically, things like PWR, ELMS, FTE, dot net core is being pulled out. In this case, we’re seeing alignment between these requirements that we distilled automatically, no work involved here, and the second contract over here, which is seventy eight percent aligned. So that’s super important. And then if we wish, we can go right into the detail and see, well, why is it aligned? So I’m gonna click on this, and you’ll start to see exactly what makes this aligned. So here’s the exact mention of dot net Core. No guesswork involved here. No kind of hoping or, you know, trusting that Gen AI know, give us an approximate likeness. This is reality. Dot net Core appears ten times in this document, and it appears in these areas. So that’s really important as you look through this that you have a good sense of that. And you also also see there’s a high fidelity view on the right hand side. So it gives you this full end to end traceability. So it’s no longer this idea of black box GenAI. It’s using GenAI in conjunction with this idea of determinism, making sure that what we get back is reliable and trustworthy. Okay. If we switch and we look at this so just to kinda say, you know, future direction, where we’re taking this view is that we see a very close correlation between this type of visual. And for instance, if I want to see alignment between a proposal sections and the proposal requirements, we will start to weave in these types of views in that context as well to make sure that we’re answering the mail later on in the proposal lifecycle. The next thing that we typically see our customers using is research. And research is really where we kind of start to ask questions of the requirements. So in my case, I’m going to ask some risk questions. Or in fact, I might just kind of do a shortcut prompt. So these are predefined shortcuts. These are what Allison was talking about in terms of the prompt library. They’re coming from here. You can configure this with your best prompts. So I happen to have a prompt called exec summary here. So it’s basically saying, develop an executive summary that still is the core strategy for pursuing this opportunity. Highlight the most critical problem requirements, driving priorities, operational stuff. I’m gonna add one final thing to this prompt, and I’m say base it on the requirements in the document attached. So that’s cool. And I will add in now the solicitation. So what we’re doing is we’re taking this solicitation, and we’re grounding the AI with that. I’ll just kinda fix that typo a quick second. I’ll let that run. Now what’s happening now is really important. You may have heard the term agentic software. So this is a an agentic workflow that’s going on right now. About a year or two years ago, people were doing what’s called one shot prompting, which basically meant you send off the prompt, you get a result. Now most modern day systems, at least the best, are actually doing a full analysis, and they’re actually going through these kind of questions internally. So it’s a multipass approach. And what you get back is basically a much better quality answer. You’ll also see that since I actually ran the prompt against the document, there’s citations coming back. So you’ll see ones and twos and threes. And as this keeps generating, you’ll start to see the citations. So performance from fifteenth of February twenty twenty six through fourteenth of February twenty thirty, that’s coming because we found this in the document. Now I should have restricted this per prompt to be a little bit shorter for the purpose of this. It’s still generating. And I should also say that while we’re doing this generation, you can configure a VisibleThread to point at any number of of large language models. So we’re not restricted to a certain large language model. It’s doing its thing. It’s still definitely should have made a shorter version of this. And once it does that, you can see there’s quite a few citations. Every last requirement is actually cited directly to the particular particulars in the document that we use to ground this AI. So now we show exactly that. So citation thirty four around FAM a and FAM d, it’s coming from here. You can see it here on the right hand side. So that’s super important as you actually look at AI generally. You want this citation stuff, you wanted that paragraph level. Some products will stick those citations in, but the problem is they put them at the very end. There’s very little point in me seeing fifty two citations all listed nicely at the end, but I can’t have the explicit traceability right back exactly to the paragraph where this thing is derived. This builds trust. It allows you then start getting a degree of confidence, and you insert it here. As we actually evolve our approach to AI, we’re starting to learn more and more smarts, in particular on the agentic side, to make this behavior of AI even better. And you’ll see that coming out a lot. Here, I’ll save this, and I’ll call it a general summary solicitation, and I’ll save that. And you now notice that basically that’s here. It’s my archive. It’s stored on the system so all my colleagues can see this. This is really a great use case for generative AI. It’s one of the best use cases. You’ll see this all over the place. We need to mix this with standard determinism as well. And again, we’ve used that word a lot. It’s coming into the lexicon more and more. Deterministic software is pattern based software or software that actually is repeatable and works the same way every time. For that, and particularly for shredding a doc, and let’s assume that we’re into the shredding space, what we are going to do is here, start to run a really a kind of a distillation of the documents to do some risk analysis. So I’m going to run a risk shred here, call it risk. I’m going to choose the same document. Again, this is the idea of actually using determinism and AI in concert. And I’m going to choose a search dictionary. Allison talked about search dictionaries as the core config down here. So this is now where we take advantage of that. It’s reusable search effectively. I’ve got a search dictionary categorized as risk compliance commercial. I’d like to run it now before we pull the trigger on this in case there is any tricky clauses that we need to redline and get rid of. So I’ll run that. And now what you see when we actually run that is the ability to get a complete breakdown of terms and phrases that we would ordinarily have to do by hand and not do in a repeatable way. So if you’ve got junior colleagues who would normally do this, this is a great way of empowering those colleagues to actually do almost a consultant in a box type approach. So we’re looking for liability clauses. I see one of twenty four. You can see as we shred it out, it’s one hundred percent accurate. You can see as we select in the middle pane, this is a full shred, we see the high fidelity viewer on the right showing us this. We can then start to do reviews on this. Again, that’s important. So now I’m in review mode. If this liability clause is of some concern for me, I can start to trigger high, medium, low on this. In our next release, we’re going to allow you configure all this stuff so you can put your own markings on this. And I’m also going to add a comment on this, and I’m gonna say red line this clause is of some concern. And again, as we build this out, what you’ll see here is threaded comments coming soon. You’ll also see name checking where I can add people, and so on and so forth. And all of this forms the basis of a audit trail that you want in the system. So I’m gonna save this for later. When I do that, the great thing about this is that this is held in this workspace. I’m using this shred to actually expose that to my colleagues. There’s my wrist shred two. There is the one that I just literally ran. You can see it’s in progress. And, again, any colleagues can also go in here and start to make their markup. Most important point, of course, is that you’re not restricted to work in here. You can take these outputs, and you can shred them out to Excel as we have always done over the years. So you can take that and bang it up to Excel if you need. We will be extending this to allow you customize the columns here, allow you put in attributes, name check people, and so forth. And you’ll see that in the next release. Okay, at this moment, what I’m going to do in the interest of time, I’m gonna hand it back to you, Allison. I will stop sharing, and you might just take the baton from here. Sounds good. Let me share my screen, and then let me know if you can see it. Can everybody see that? Yes. Yes. Okay, great. Thank you. All right. So, Fergal talked through the front end of potential BD workflow. What I want to do is talk through a proposal workflow. So transitioning to an opportunity, we have our Kanban board. I mentioned this in the front end. So I have all my opportunities that I’m managing. I’m going to do a bit of work in one specific opportunity. So I’m going to select that opportunity and then here I am in my summary page. I have all the information that I need. I have all the documents, whether these are uploaded documents, documents that we’ve created, everything that we’re working with on this document. Because the excellent thing is that with the structure of VisibleThread and supporting the workflow, we’re not losing anything in the handoff. So we can see the work and the analysis that folks did on the front end, that executive summary that Fergal created any SWOT analysis, any opportunity analysis, differentiator and approach analysis, I can see that as well. So I have visibility into all of that. So we’re making much more informed decisions. But when we’re looking here at our page, you see that we have some options just like we did in the front end. So there’s several things that we can do and we want to focus here on that combination workflow that I talked about in the front end. So we’re going to start with requirements. We are going to then move to outline content creation and then come back to look at Watchwords and some of those deterministic guardrails. So the first thing that I’m gonna do is I’m going to click into shredding. Now, important thing to call out here is that you’ll see this looks very similar to the shred that Fergal just ran on the front end and that’s because it is. So a really great thing about the platform is there are not multiple modules with different types of functionality that you’re going to have to learn and there’s not a huge learning curve there because things like creating a shred and content are going to look the same, but you can access them in multiple places in order to support your workflows in the most meaningful way. So in this instance, I am going to be looking at requirements. So I’m going to look at requirements. I’m going to choose my document here, all the documents that I have to choose from associated with this opportunity. I’m going to select my solicitation and then I’m going to choose a dictionary. So here I want to choose this requirements review dictionary. You see I’ve tagged this. These tags are all customizable to make it easy if you have a multitude of dictionaries for folks to come in and find the things that they’re looking for as well as upscale folks in knowledge share. So same process, I’m just going to run the shred here, but we’ll see that based on the terminology of my results, this is going to look a little bit different. Sure, it’s my size of my screen. Okay, so here, instead of some of the contractual language that Fergal had, I have a lot of indicating language broken down into categories. So different types of requirements. We have our mandatory requirements, our will, shall, must, ensure required. We have our prohibitions, will not, shall not, may not, is not authorized. Our conditional language, deliverable language, performance standards, what’s acceptable, what’s unacceptable. So the things that we need to see when we’re really digesting an opportunity and trying to find all of those requirements. So the same way that we saw previously, I will click here and then I’m immediately jumped into the language that I’m seeing. So I can click one item, I can click a whole section and we really quickly can start to break down this opportunity. Again, if I select one, it’s going to jump me right to that spot in our source document. So this is an amazing way to perform a guided analysis along with doing a review for these types of terms all at the same time. So this is obviously helpful from an efficiency perspective, but also from an accuracy and visibility perspective. Again, the dictionaries are completely customizable and can be applied at a multitude of places in the process. So just like you saw with Fergal, we can start a review here. And like he mentioned, the important thing to call out is what the development will look like. So right now what we see are categories for a paragraph content, we can do a review, but the future state and the next release, our August release and then beyond will be additional categories, categories that are customizable, categories that help you build the compliance matrix and tie that to your outline. So a lot of different work that we can do there or that we have planned to do there to help build the level of customization and the way that this is accessed. We’ve worked with customers for a long time over the years to help them build macros to use within the VisibleThread system. So we have a multitude of say a great piece of source material to understand what a real compliance matrix looks like and what type of customizations that you need. And you will see those in practice over the development, in the coming months. So we have created our I’m going to exit this review. We have created our shred. And just like we saw in the front end, I can come back and it’ll be here. So anybody who’s working in the system can collaborate. We can leave notes. We can work on review. And then those workflow pieces will be enhanced as well with additional ways to call out folks, tasks, work through reviews and those types of things. So we’ve created our requirements and what we want to talk about now is creating an outline. So we’ve identified our requirements. We wanna talk about the outline for our document. So I’ve gone ahead and run this just because we don’t wanna sit here and stare at each other for a few minutes while it generates, but I’ve gone ahead and run this. So this is a combination approach background for classifying the type of document, examining the structure, then using JN AI to create an outline. Then you have a lot of different options here when you can edit, you can change levels, change structures, add, delete. So you have, a lot of flexibility here around the outline that’s created. But the thing that’s really important is we can use the outline as a source of truth to begin to drive action. So what I wanna do is I wanna start to, as an example, to write to a specific So here in my volume one, my technical capabilities section, I have this ad hoc support and responsive services. So my transition in transition out plan, training certification, documentation transfer, etcetera, etcetera. So I say I want to start writing to this. So I can click here, start writing, and I’m immediately in my proposal writing section, and it’s pulled in this requirement that I can start writing to. So this is where collections and our prompt library are going to come in. So here we have our requirement. This is what we want to respond to. So I’m just going to highlight and then I’m going to select from my proposal requirements response section this address requirement. This is a selection based prompt, so it automatically knows that I want to use this piece of text that I’ve highlighted as my source and I’m asked it to address the requirement and create sections for each unique requirement. As you saw Fergal do, these aren’t static. I could add to this and do different sections. But the important thing that I want to do is I want to use a collection. So ultimately, the LLM could come up with an answer for this, but it’s not going to be incredibly accurate. It’s not going to be referenceable and certainly doesn’t have proof points that we can stand on. So I want to use a capability narrative. So I’m going to point this at my capability narrative folder and ask the system to address this requirement based on this source material. So, let’s go ahead and send that off. And like Fergal said, so we have the workflow running in the background and here, it’s going to ask me a couple clarifying questions, sometimes based on the information you may need to. So I say I’ll say brief just for this instance, I want to hold everyone forever. I would like a formal document and we are writing for, an evaluator. So, it will factor that in and then create my content. So, depending on the structure that you have in your prompt, depending on what the system determines it needs in order to give you the highest quality content or the most relevant content possible. At times, Vizi will ask clarifying questions. So, again, working through that agentic workflow in the background to assess the document, assess my source material, and the requirement that we are responding to. As this starts, you’ll start to see those same, reference bubbles or the same kind of reference tags that we saw in the front end. But in this instance, it won’t be referencing back to parts of the solicitation, it will be referencing our source documentation. So that will allow us as we are working through our review to number one, identify if this information is accurate. So if we’re talking here about my transition in plan and training and certification, so is the information that it’s given here accurate when it’s talking about my process and, our credentialing and all of the different things it’s addressing here, is that accurate? Is that traceable? Is that defensible? And that’s what the referencing does for you. And it’ll let me show you as soon as it’s finished, but that will, allow us to pull up the areas from our source documentation that are directly tied to the content that we’ve created here. Give it just a second. And then I’ll show you what that looks like quickly. So here it will pull straight from the capability area. It will tell me what document straight to the section so I can check this, reference it and make sure that the content is created. I can then insert this content, review, start to work, work with expand prompts, and start to refine my results. I can also save this and the system will automatically insert my headings into my outline. So the two are staying in lockstep. So what I want to do before we leave here is I want to talk about readability. So here I want to save an analog. Okay. Save. So I saved the document because I want to do a readability analysis. So this is where the watchwords came in that I talked about on the front end. So here, the important thing is we want to understand where are we at from a readability perspective? Where are we at from a grade level perspective? What level of passive voice do we have? And what are some of the issues? So here I can click down and this watchword, so we have a couple of watchword hits. That’s a good example of what that looks like. But we’re calling out long words, we’re calling out long sentences, we’re calling out watchwords and it’s telling me that I have an approved term to change. I can click to change so I can do some editing there. But ultimately, this is our guardrail from a quality perspective. So we can assess the quality of our content and make sure that we’re meeting terminology right within the system. Now, of today, this is happening in the web based browser, the browser based, approach. But this is, if anybody has used VT Writer before, you’re familiar with the word add in. So the word add in, with full access to the LLM and full, integration, into the VT seven platform will be available in our August release. So at that point, you can work where you’re at. So VisibleThread will meet you where you’re at and you can work within Microsoft Word with the ad and doing all the same thing and accessing all of the same items. So I’m going to stop there, but that’s the general highlight of the deterministic checks. So kind of the sandwiching them on both ends. So we use the deterministic approach to identify requirements, identify risk in the front end. Then we leverage GenAI for creating outlines, creating content, working through some of our drafts and analysis. But then we always come back to our guardrails from a readability perspective and make sure that we’re aware of watchwords, we’re aware of quality and that we’re structuring the content in an appropriate way. So I’m going to transition, over, I believe to Andrew, but that’s where I will stop. Thank you, Allison, Micheál, and Fergal. So the upcoming versions of v t seven going to introduce a number of very exciting new technologies, including an MCP server and a secure API, which I’m delighted to demo here. The MCP server and API will are designed to give our users much greater flexibility in how they interact with VisibleThread and give them the ability to integrate it with their own tool chains. You can think of an API as a headless version of the existing product. It’s a way to interact with it outside of the browser. The API also underpins the RMCP server to ensure that that service behavior is deterministic and predictable even when integrated with AI agents. The API itself is secured with authentication and authorization. You must have a valid account to use it, and you can only do in the API what you can do as a user in the main product. And we have an audit trail that will tie the user who is using the API or MCP server to the actions taken using their SSO sign up. Both the API and MCP server are aware of workspaces. So as a user, you will only be able to access documents, dictionaries, etcetera, and perform actions from within the workspaces you have access to. The API and MCP server will allow our clients to integrate VisibleThread functionality securely and safely directly within their own tool chains, giving much greater range of interaction. For those who aren’t aware, MCP is a plug and play AI technology that allows AI agents not developed by, VisibleThread to expand their tool sets to include our functionality. We’ve developed a secure MCP server that I will demo today that interacts with the secure API and exposes those features flexibly to any compatible AI agent. For the demo today, I will show us using, integrating with a Microsoft Teams client and integrating it with our MCP server, and I’ll perform a brief shred of a document hosted in v t seven. So I’m just about to share my screen. And if someone can shout when they can see it. It looks perfectly normal or you’re in the what you’re in your whiteboard. Yes. That’s correct. Okay. So what you can see on screen now is a standard, Microsoft teams. I selected the whiteboard because there’s nothing controversial on it. So I wanted to show you where the VisibleThread app will appear. So it appears in the more apps on teams. We click on this, you’ll see that my previous conversations that I’ve been having with it today are saved. I’m just gonna jump straight in and say, please list the documents available in my workspace. And as a tip, it’s always good practice to be polite to AI agents because we never know when the great robotic revolution is going to occur. And I, for one, don’t intend to be enslaved. So always say please. Now we now have two documents that are listed in my workspace, which is ID seven. And you can see they’re fairly complicated names combined synopsis and TXZ Corp, whatever that is. So one nice feature here is I’m going to say well, first off, I’ll say, please list my dictionaries. And it should give me a list of all the dictionaries I have available in my workspace. There we go. Okay. So again, we can see now we’ve got, roughly ten or eleven, dictionaries available to us. Now one of the strengths of using an AI agent is I’ll now show show you the English language of how we’re going to shred this. So please shred the combined synopsis solicitation document using the contract kickoff dictionary. This one, I go and run a shred in the background. And notice to date, everything has been tied to my workspace. I’m not able to access anything outside. Okay. Okay. Yeah. Please use the name. So I have to give it a different name. So I was playing around with this earlier. See if that works. Yes, okay. Now, what we have now is it’s performed the shred, it’s given, it’s used the name I asked to, it’s reporting success and it’s given me a URL. So the URL will take me to the VisibleThread document. So product and it will load up our shredding document in the background. You can see there that we performed the shred and we’ve integrated this with our existing, if I go into miscellaneous analysis, we can see that it was integrated with the miscellaneous with with the current product. So we can actually go in and then see the shred. And that’s really what I wanted to show today. So just to show you that and you can put spot over on the right as a final point, our ongoing conversation, because we were created a general copilot AI MCP agent, and we’ve now brought that in automatically into the, product with us when we clicked on the link. Right. I’m gonna end the demo there, and I’ll hand back over Micheál to whoever’s rounding up. Yeah, Andrew. Sorry, Micheál. Just, Andrew, on on that one. Again, our customers will obviously be, you know, concerned about security boundaries and making sure that they have access to the right stuff. What what what kind of guarantees do we give in terms of MCP exposing, you know, internal data on workspaces? Maybe you could you could speak towards that a little bit, please. Yes. So the MCP server can only access our product via the secure API I mentioned. That API is tied down so that it uses whichever person is using the product, it uses their single sign on to ensure that you have exactly the same access via the MCP or the API as you do when you log on to the web server. So we we have it completely locked down. You can’t actually go up, Jesse. We might demo this if you want. For example, there is no ability at the moment to delete a shred. So if you ask it to delete a shred, it’ll you know, the MTP clients to delete a shred, it’ll tell you it doesn’t have permission to do so. And if you try and access something, let’s say I knew your workspace was workspace ID six, for example, I don’t have access to that. So no matter how many times I give it the ID number, use workspace six, it won’t because it’s checking what I have, what I’m not only am I authorized, but am I not only am I authenticated, but am I authorized to do what I’m trying to do? So we’re leveraging all of the high level security we have built into the existing product, to make certain that we’re not intrude that we’re not widening the attack surface for any bad actors. Great. Thanks. That’s fabulous. Andrew, thank you so much. That’s so exciting. I know we’ve gone a little bit overtime, so I thank everybody who stayed on. The recording and slides will be available after. And if you do have any specific questions, feel free to add it into the chat box. We’ll get back to you afterwards or email us or contact us in any way and we’ll be able to round this out for you. But just to summarize and maybe final is the webinar before we go, Fergal, if you’d be happy just to kinda round it all off together, that would be fabulous. Sure. Thanks, Micheál. So this slide is is kinda showing you how we think about kind of evolving the platform. We break things out into themes. We have six working themes that we address or kind of feature extensions around. The first theme here is team workflow and collaboration. So this would include things like notifications, alerts, making sure we can name check people, make sure we can allocate tasks, and so forth. So you’ll see a lot of evolution around the platform and grant what, both Allison and myself done with earlier, in that regard. Insights and analysis, we see a huge opportunity to provide better insights and appropriate points during the lifecycle that will be very clearly delineated at the right time. So for instance, as you get into a proposal workflow, and I mentioned this, earlier in my session or in my segment, we really are keen to make sure that when your different people are writing different sections, which is very typical and more complex, pursuits, that actually all requirements are completely addressed in the various sections in the various docs owned by different people. So that would be better insights and analysis. Reporting and visibility, you’ll see a lot more dashboards coming into play, you’ll see a lot more insights and better visibility into some of the data we sit on. For us, market intelligence is all about giving you, the customer, better market insights. This, you know, many of you already use competitive intelligence platforms. We’re not in any way suggesting that you stop using those, but this gives you a little bit better head start, particularly if you kinda just wanna make sure that you kick start the process earlier than you’re doing at the moment. Workflow integrations, what does that include? It includes things like Salesforce integration on our opportunity track opportunity side. It includes Microsoft Word support, which is coming in August, so it’s literally three or four weeks away. And then we get into usability enhancements. We have always been a company that prides ourselves on building product that is easy to use. One of the biggest challenges that a lot of vendors have is that they’re building a huge amount of features, but not thoughtfully integrating them in a way that does not mandate you to go off on a one week kind of course to understand how to how to use them. We spend a ton of time using usability studies, making sure that what we build is understandable, accessible to you, where you don’t have that much time to understand it. And we, of course, have a fully empowered customer success team who really can take the heavy lifting of any configuration you may need. So usability, that’s a big part of our mantra. It sounds kind of squishy. But every time you use the product, we hope that you’ll have an experience that you don’t have to relearn the product. So that’s kind of the six core pillars, and we hope that’s summarizing our direction generally. Brilliant. Fergal, thank you so much for that. Just some final notes. Thank you to everyone who attended today. I know we went on a little bit long, but I’m delighted to see so many people still there and all the well wishes and thank yous from the audience are very well appreciated. Our next webinar is interesting. It’s with WPS. It’s with a BD manager and a proposal manager, and they’re going to speak about the goals in between the two departments. So you saw two workflows today. You want to see them in real life on the fifth of August, you can see WPS Healthcare go through their actual workflow, talking about the as well. So thank you to the audience and especially thank you to Fergal, Allison and Andrew for their time today and the preparation. The recording and the slides will be available after. And I hope you all have a great day.