Name: Pigment AI: The Next Evolution
Uploaded: 2026-03-12T17:01:06.365Z
Duration: 44 min 12 s
Description: Pigment AI: The Next Evolution

Transcript for "Pigment AI: The Next Evolution": Hello, everyone, and welcome to this Pigment AI webinar. My name is Michael, and I'm part of the product marketing team here at Pigment. And we are super excited to have you all, with us today. As you certainly know, last week was a very big week for Pigment. We launched something that we believe is going to fundamentally change how you build and extend planning models with the concept of intent modeling, thanks to our new model agent. But that was not all. We had new custom agents, new version of the analyst agent, a new UI experience. There was a lot, and I'm sure there was a lot for you to digest. And if you're like, oh, maybe I missed some of that. Oh, I'm not even sure what this guy is talking about. Well, don't worry. That's why we are here today. So we have gathered our top AI experts who are ready to walk you through, what's new with Pigment AI, what this means for you, and they are going to show you everything live. So before we start, a few things. This session is being recorded, and you will receive the link to the recording about twenty four hours after the event. The next thing is that you can submit questions at any time using the q and a panel that is on the top right. We have our amazing product managers that are here for you. We have Andrea, Andres, Lucas, and Tongi. So please feel free to drop your question in there, and they will answer direct in the chat, or we will have a live q and a at the end. So what do we have at the agenda for you today? Well, we start with the presentation of the next evolution of Pigment AI, and Andrea will come on stage to introduce all the AI agents that we launched last week. Then we'll have Anna that will show you everything live with a live demo, and I'm sure you'll find it super cool. And finally, we will finish with a ten minute live q and a session with our product managers. They will join on stage and answer all your questions. So with that, let's get started. And, Andrea, over to you. Okay. Hi, everyone. I'm Andrea, product manager at Pikmin working on AI. And as Mikael said, like, we have something super really exciting things to share today. So I will try to make, the next ten minutes worth your time by giving you a quick overview of the latest launches, but please, raise more questions in the q and a panel. I think that we have a lots of questions, so take that opportunity. So let's start. So let me paint you a picture here. So let's imagine it's q three. Your CFO walks in and say, hey. We need to add a new sales capacity model, and we need it in two weeks. And you think, okay. Who do I call? How long actually this will take? And even where where do I even start? Right? So that has been, like, the reality of financial planning for decades. Leading a model meant designing the structure, writing the formulas, checking the logic. You really needed, like, the right people, and on top of that, you needed time, sometimes even months. And you really had to speak the technical language of the tool. Honestly, it feels a little bit like you need a PhD just to build a budget. So that's exactly what we set out to fix. So we are introducing something called intent modeling, and the idea behind that, it's pretty simple. It means that instead of building your model yourself, you do describe what you want. Something like, I want to forecast our top line, or I want to create a sales capacity plan, and pigment takes it from there. So now it used to take months. That happens in days and even sometimes in hours. And it's not just faster. This is not the only thing that it is. It's more than that because I feel now building a planning application is accessible to anyone who knows their business and not just to technical experts. So now let's see how does it work in the next slide. Okay. This is where the modeler agent comes in, and the good news is that it's already available to all our customers. So if you haven't requested access yet, please, you just need to open a support ticket, and you will get access. So let's back to the story I was telling. Imagine you are that person who just got the two week deadline from the CFO. You will open the Pigment app. You will describe what you need, and the modeler is gonna build it from scratch based on just what you said. And if your business actually or requirements changes, because they always do, it can extend and adapt existing model too as well. And what it's really like, the most important thing I think about the modeler agent is that at very step, he could explain what it built and why it built it in plain language. So you need to think about the modeler agent as a very experienced planning consultant that, first, it's always available, and it's never too busy. So the modeler agent works across blocks, formulas, views, and boards. And coming soon, you will be able to use it directly from tools like Cursor or Cloud Code via our MCP connector. So this is really exciting, especially for the more technical users amongst you. So let's introduce the next one, which is the analyst agent. Some of you have already used it before, but we have made some big improvements recently. So let me take you through them. Picture this. You are looking at a board and you want to dig deeper into the numbers, but you are not a power users. You are not familiar with the UI, so you do what feels safe. You export to Excel or Google Sheet. But export is never quite complete. You're missing dimensions. Your data isn't fully there. And by the time you figure it out, you already need to export again more data. So in the end, that kind of analysis flow feels a little bit, like, slow, partial, and it's never quite right. So now with the analyst agent, you just need to start a conversation directly in Pigment from a board with the full dataset so the agent knows which board you are on, which data you are looking at. So the only thing that you need to do is to ask follow-up questions, go deeper, and explore data progressively step by step. It really feels like having a back and forth conversation where we with a real analyst. And the analytics agent is also a lot more powerful now because it now generates, like, Python code behind the scenes that allows you to handle more advanced calculations. Things like, for example, time series analysis, correlations, anomaly detection, the kind of thing that used to take hours before now just happens directly in the chat. And last thing, and this one is really handy. In Pigment, you can set up what we call missions, and missions is a safe set of instructions that tell the agent what analysis needs to run automatically on a schedule. And until now, you had to write those instructions by yourself, which, let's be honest, not everyone loves doing, and it was a little bit complex. So now you just have to start that conversation, ask the agent to create a report, and just by clicking one button, which is called save as mission, the agent is gonna write the mission instructions for you. So that way, the conversation becomes a repeatable and automated workflow just like that. And last but not least, actually, this one is for the ones that who's ever thought, like, I wish this tool work exactly the way my team works. So this is for you. And here's the story. You have business users coming into Pingman every day, looking at dashboard, trying to understand the numbers, and every week, the same kind of questions comes come up. What does this metric mean? Should I compare to budget, or should I compare to forecast? And you end up answering the same things over and over again, and that's what custom agents are built for. They let you create an agent that is tailored to a specific use case in your organization. Think of it like your own version of ChatGPT or Cloud, but one that actually knows your business, your data, and the way your team works. In order to configure the custom agents, you need to give it instructions so that actually you set the tone it should it should use, how it should behave, what it should do, what it shouldn't. You give it knowledge as well about your business. For now, you can paste it as a raw text, but file uploads with will be coming soon, and you need to choose what it can do. For example, it could run the analysis. It could navigate through boards. It could answer questions about pigment. So you are the one actually, like, choosing the the capabilities of that agent. So all that to say that instead of writing that internal Wiki page that nobody nobody reads, we all have one. You build an agent that shares knowledge directly in context when people actually need it, a consistent, always available expert for your team. Custom agents are available now. So, honestly, once I think once you you're gonna start thinking about the use cases, it will be hard to stop. So, yeah, that's a little bit, like, the the story about this week. I had the modeler agent, a much more improved version of the analyst agent, and the custom agents. So not not a bad week. And now let's move, to the best part. So I will be handing over to Anna who's going to show you all of this live. So the stage is yours, Anna. Great. Thank you so much, Andrea. Go ahead and get my screen shared here. Thank you all for joining today. So I'm Anna, solution consultant at Pigment here based in San Francisco. But for the next twenty or so minutes, I'll be taking on the persona of a financial analyst at a company called Lumina. Now before we get into things here, what we're seeing right now is an ARR forecast model in pigment. In the chat, go ahead and let me know how long you think it would take you to build a ARR model end to end or how long it has taken you in the past. Just wanna understand, you know, what that current state is without the air without the model or agent. Because you wouldn't believe the day that I've had. What you are seeing before you hear this beautiful, interactive, cohort driven ARR model did not exist just a few hours ago, so I built this within one day. What would have taken at least a week took just a couple hours with pigment agents. Let me take you back to just a few hours ago and talk about how pigment brought this to life And pay attention to a few things as I go through. First, how this started from a simple business conversation. Next, how it was like building with having a pigment expert by my side. And lastly, how quickly we went from having raw data to being board ready board ready. Now to set the scene a bit more, yesterday, my company, Lumina, we closed our acquisition of another company, Vivid three sixty, a subscription based health tech company. Historically, my company, Lumina, has really only forecasted hardware revenue, so we didn't have an ARR model available in Pigment yet. And here's the kicker. Our board meeting is tomorrow, and we've been asked to present an ARR forecast through 2027 for Vivid three sixty. So that's a pretty big ask that I need to get done quickly. To help us with this endeavor, Vivid three sixty sent us some different different information here. First, a dataset that covers their historical cohorts, including pricing and subscribers, as well as a business requirements document that they detailed out their approach to forecasting. Now the normal step would be for me to have to scrub through all this information, set up some scoping calls, and have days or even weeks of building ahead. But I needed to make sure that this cohort based model was available today, So I went to the pigment modeler. Now as I mentioned, I'm a financial analyst here at Lumina, but I've actually helped build some of our hardware revenue forecasts. I kind of play multiple roles. So I'm familiar with building, but I wouldn't say I'm a pigment expert by any means. Now after receiving that information, I didn't even open one spreadsheet. I went straight to pigment agents. And from there, I went directly to the modeler agent. Now with the modeler, I can describe what I want and attach different files like CSVs, copy and paste in BRD information. So today, I said that I need to build an ARR forecast model today, and I can send that off to the modeler agent. So I'm not having to start with configuration or thinking about all, you know, the different blocks I need. I can start just with the intent first. And this isn't just a blind automation. It's a conversation. The modeler agent is embedded with a wealth of knowledge around best practices, the planning, but also pigment best practices. So it has questions, ask me questions, and I can even say that it's show see that it's recommending specific actions. So I can have this back and forth conversation, even type in my own answer if I want there. So it's not again, it's not just all happening in the back end. It's really happening from this con conversational perspective. And what used to take hours of decision making, the modeler is helping me do live just within a few minutes. It's reading the structure of my data, understanding the requirements, and in the end, crafting a plan for us to implement together. Now it's gonna take a few minutes for it to run through this information, so I'm gonna jump forward to a completed chat where I had a couple more questions that were asked to me from the modeler. Questions around retention and churn, around how far into, the future we need to forecast this. So we had this back and forth conversation, and that resulted in a plan. Now this plan includes all of the steps of even going as far as creating a folder to organize everything to make sure that everything, you know, follows the, how things should be organized, creating metrics, as well as creating different views and dashboards through different implementation phases and a to do list that the modeler will actually check off live as I go through the build. Now I was really able to go from business intent to a full implementation plan before I even finished my morning here. Now this is truly a fundamental shift in how we approach building and problem solving. You don't have to start with configuration or complex scripts. I can describe what I want. Pigment translates that into a plan, and the gap between idea and execution has collapsed. Now we're at our next hurdle here, which is then taking all these great requirements and turning it into a real model. Normally, we have a project kickoff call, have, you know, days of configuration ahead. But once I had my plan here in pigment, I was able to approve that plan, send it off, and start kicking off that build. Now next, I'm going to move us forward and take we're gonna look back at a conversation that I had over the course of a couple hours to get this model built out. So, again, I did this all in one day here. Scrolling down to the bottom of my plan, I can see that as I went through the build here, the modeler has actually checked off different steps live. And one of those first steps was creating dimensions. So here we have cohort, cohort age, and, I'm sure all of you on the on the call are familiar with dimensions, but those are those lists that organize, the structure of our business and how we want to forecast, report, and our different granularities. Now for anyone not familiar with cohort modeling, it requires analyzing when subscribers join and then how they attribute to ARR over time. And it's something that, you know, teams have normally struggled with in the past because of having two different time series as well as a large volume of data. But because of Pigment's multidimensionality and because of our scalable engine, Pigment takes something that is normally complicated like that, something that commonly happens in spreadsheets, and brings it into a more intuitive experience and faster experience in Pigment. Now that CSV that I've included of just flipping back to here is that raw data, we can now see that it exists in a transaction list in pigment. We see all those different colorful cells coming to life that are our beautiful dimensions here. What's really key is that I did not have to do any training. The modeler, with its embedded knowledge as well as the data set and the requirements that you give it, that alone allows the modeler and then, you know, the questions that it's asking me and the answers I'm giving allows the modeler to create that whole plan and build out this core structure of our model and not having to do any coding. Now moving on to a the more complex part of the process here past the dimensions is really all of the metrics and and the logic, so really the math of where everything is living, really bringing the true model to life. So the modeler has taken pages and pages of that information. And in hours, it did what a couple a team of people normally takes days to complete. And it's not just building like crazy and then letting me know what happened after. We're having a conversation, and it's showing me each piece of the puzzle that it's creating, whether it's creating a metric, creating a dimension, updating a formula. And at every step, we have that conversational back and forth, but I can also expand the thought and thought and work process information. So I can see that it's understanding that based off a change I had, I need to update the view. So I can go through and actually learn from the model as well as how it's thinking about this. I kind of like to think of this as, you know, if we're working from home and we start talking to ourselves, it's like having that train of thought there that the modeler has taken. So there's true transparency. Now about an hour into my build this morning, I realized that I was missing a bit of detail in our ARR output calculations here. I needed to make sure that we had both product and channel. We had some city details originally, but I wanna focus just on product and channel. And we just had month and cohort month currently to to report on those final ARR metrics. So something that I just hadn't caught earlier, probably because I hadn't finished my matcha. And so I just simply shared that with the modeler that, hey. I see there's an issue. Can we go back and add some dimensionality? So within each of these metrics, the modeler is updating the dimensionality. It is going through and making sure that any formula changes that are necessary are happening as well as doing some validation of before and after and then letting me know, you know, exactly what what happened. Now in the chat, I guess, go ahead and give me, like, a hand emoji or a heck yes if, you know, you've updated one formula and then maybe it breaks other things. So in pigment, thankfully, it's easy to figure out where that where those things happen. If you're in a spreadsheet, you know, that's even more difficult. But understanding that we're all our pig all pigment users here, when we don't have agents, we still lose time trying to piece things together. So the modeler is just making that pigment experience of understanding where things are broken happen faster. So this is really that moment where I felt like I had a pigment expert by my side while I was building this model. And, again, the modeler is embedded with best practices. It shows me exactly what's happening every step of the way. This isn't a black box AI model. It's transparent, and it's audible, and that is what can give you the confidence that the modeler is truly that pigment expert by your side. Next, let's go ahead and zoom into one piece of the puzzle here. We're zooming in and looking at one of our metrics. So this is new cohort ARR. And, again, we have cohort month as well as calendar month. So we have those two time dimensions in one view. And with the dependency diagram, we can see everything flowing together. So, because this is just directly embedded in pigment, the model is creating all of these different metrics that have millions of cells. I'll just get back here. Metrics that have millions of cells. So, again, we're making things explicit and visual. And the metrics weren't built off of just aggregated dataset because of that core elastic engine that you have with pigment. Again, everything can hold millions of cells. You can get the most granular details and decide what you want to do where. So the modeler built it, but the pigment on its own is creating this visual web and supported by pigment's, you know, unmatched scale of its core engine. Now as we're all familiar, I can come in and pivot any view in pigment. Now I can bring my cohort months down to my rows, filter by my channels, and I can really see that waterfall starting to come to life. So Pigment gives me that flexibility of being able to have the way I prefer to see it as an analyst, get into those granular details or have the aggregated view aggregated view that my CFO prefers when we need to go ahead and report our out to the board. Now this true flexibility and scale is only possible because of that enterprise grade foundation that Pigment has. And everything you're seeing here in your list build is not a side experience or something experiment or something we just five coded for a one off use case. Because pigment agents are directly embedded into the platform, you get the power of AgenTik AI along with scalable enterprise grade EPM platform. Now as one of the last steps that the modeler and myself identified here was to actually create a dashboard. So I'm gonna search down farther into our conversation here, and see where when we were coming to the end of our conversation, we identified that we wanted to create a dashboard for this because the working model and all the math, that's not the finish line. We wanna make it usable for leadership. Right? And so this is where we come to another hurdle. Sometimes this feels like decision after decision, you know, how do we wanna present the data, what visuals to use, what story and pigment because it's so flexible. Sometimes there's just so much you can do. And as someone who can suffer from analysis paralysis and decision fatigue concurrently, sometimes this process can take a long time because there's just so much you can do. But what the modeler is able to do with its design best practices is take that puzzle that we built together and help build a visual story. So I've clicked this button here. It's taken me to the dashboard that it created. So it's not just numbers, inputs, charts. It's presentation ready and collaboration ready. So the modeler gave me you know, that's a a good starting point to work off of in just minutes and probably would have taken me, you know, a couple hours of back and forth, or just seeing all the cool things I can do and trying to make that decision. The modeler really helps me get past that blank page fear of now I have everything. How do I make it look the way I want it to? Now this main board here works for finance and the CFO, but I want to tailor this. We've I've been able to do this so quickly here that now I wanna create another board. I wanna tailor something for our product marketing managers because this new company we've acquired, that's gonna affect how they're thinking about channel performance along with our different, you know, hardware products here. Because with the modeler, I cannot only iterate on formulas, I can also iterate on boards, extending out my model to different teams and customizing for different views based on the stakeholder. So I have simply just had a quick conversation with the modeler. We can see I with Pigment agents, you can multithread within the same agent, like the modeler, analyst, and be working on all different conversations at the same time. And, you know, even after you log off, payment agents are working for you. So I had another conversation here where I wanted to start thinking about creating a board for this product marketing manager's channel performance. Just gonna scroll us up to the top here. I essentially gave it, you know, all the information, information about the stakeholder, really describing what I want to see, and then having that back and forth conversation. We had some different phases built out, so I wanted the, modeler to just go through phase one. So those are for implementation phases can be broken out into how, you are actually taking those steps. Now scrolling down a bit here, it's telling me exactly what it did, giving me that detailed information, and then it created that dashboard. So directly, again, from the modeler chat, I can click and go to that dashboard. So within just that couple minutes back and forth conversation, making sure that the modeler understood exactly what I wanted, it built this for me. So I didn't have to, you know, export this data to create a custom view outside of pigment. We're building everything in pigment because we're able to simply just extend that governed foundation. The modeler is generating all of this, but, again, Pigment at its core is what's keeping it all connected in a central place. The modeler didn't just build formulas and metrics, but it built that full application based off my request, raw data, and requirements. It then applied UX best practices to the boards to provide a clear layout and focused storytelling. The modeler helps us extend our current pigment models out to different processes, use cases, and stakeholders faster than we've ever been able to before in Pingtn. Now at this point, I have a trusted model, multiple customized dashboards, but I thought, what if I can do more? What if I can take a monthly report that my finance team sends to marketing every month and automate that in Pigment? To do that, I can simply just start a conversation with the analyst agent. And to show that there was a question earlier about accessing these chats. Anywhere where you're working in Pigment, you'll see this AI chats circle here essentially, and it allows you to come into those chats but also see, coming back to all my agents, all of the different agents you were able to interact with. So I'm gonna go ahead and start a conversation with an analyst agent. As a modeler, you can even be building something at the same time. So just like I would brain dump to a colleague about a report that I need their help with, I can just speak to the analyst agent and say, I want to create an analysis report for growth and momentum from last month to this month of ARR by cohort and channel. You know, brain dump like I wanted to the person sitting next to me or to myself if I'm working from home. Now, again, the analyst agent knows the model that the modeler just built. You do not have to train the analyst agent on the structure of your model, that business context. You know, it already has all of that information. So based on that request, the analyst agent is going through and it's thinking, just gonna jump us forward a couple minutes here to look at the finalized report. Now I had a back and forth conversation here, including some visuals of, okay. I wanna add in a bar chart or let's add in, you know, the specific summary. And the analyst agent asked me, would you like me to create this final report? I wanted it to update this report with a specific visual here, and I can see now that looks like a mission output. Right? So what's really key here is that by having a conversation with the analyst agent, and once you get to that report that you want, you can simply just click this button and save it as a mission. Now what that does is the analyst agent writes a prompt for itself to then run this on a recurring basis. This is very different than the previous experience, right, of having to write out the mission, testing it. Again, you don't have to know how to prompt correctly. You can just use you know, have a simple have a back and forth conversation, just speaking to the analyst agent, and then it will write that prompt for itself to do that ongoing analysis. So this really is changing the game in how you interact with Pigment AI as everything is moving towards this agentic view. So prompting is off your plate now. We can just chitchat with the agents. Now I know it's going to come next. I have this forecast, but, you know, the board and my CFO aren't just gonna say, okay. That's a great forecast. Let's move on to something else. They're gonna say, how are we going to actually achieve that? I already was thinking that that's going to be the question here, so I've actually started a chat with the modeler to think about how we want to think about implementing sales capacity into this model. So I'm again, I've multithreaded working with the modeler in different capacities within the same application, and I have now a full build plan for sales capacity. And these agents, like I mentioned, you know, whether it's building or running that analysis, they can will continue to work even after I log off here. So my true AI teammates. So here's a recap of my day with pigment agents. I built an ARR model this morning. I'm going to be building sales capacity this afternoon, And tomorrow, I can build whatever comes next. Appreciate all of your time. Now I'll flip back to being, you know, the solutions consultant, and I'll go ahead and pass it back to Michael on my team. Thank you all for joining. Thanks so much, Anna. That was an amazing demo. It's always so cool to see that you just have to start with the conversation. You build the model. You build the board. And then another conversation, and you just run the analysis. So everything in one place. Very, very cool demo. Okay. So with that, now I'm gonna ask all our product managers, Andrea, Luca, Francoise, Tongi. Please come on stage for the live q and a. We'll have around seven minutes, and I saw a lot of questions in the chat, so we are gonna pick a few. So maybe let's start with a question from for. So there was we had a few people asking, how do we enable the attach file option that was Yes. that at the beginning of the demo, is it available yet? So attach file option is not available to to everyone yet. It's being under kind of beta release for now. It should be available next week. We have a stable version that can enter CSV to upload data, which is that was been shown in the demo. And you can learn about format and template. So you don't need to use any templates. The model agent should be able to understand the CSV as is, especially if you have clear headers, clear separators, and that is if the CSV is not broken or corrupt, meaning that there is no miss, it can handle empty cells. But, yeah, if the CSV is is is not corrupt, it it should work fine. Sounds great. Another question for you was are the modeler agent changes associated with the modeler user Yeah. or with the the agent itself? So the way it works is that the modeler is propose you is proposing you some modeling actions in the form of a card in the chat, and you accept or reject this action based on what you see on the card. And the action will be so the modeler will do the action on your behalf. And in the history, you will see your name in front of this action. I wanted to add something. Yes. We are planning to add a little tag to make sure it's it was it has been proposed by an AI so we can differentiate actions taken by human and taken action taken by modular agent on behalf of human. And finally, if you do not have the right to write in a metric, for example, or you want to input data in the metric, other agent could suggest you a card. And when you will click accept, it will fail because you do not have the right to to do that action. So it will respect access rights and permissions. Sounds great. Thanks, Francois. Okay. We're gonna switch gears here. And one for Lucas, please. So, Lucas, do you still reckon granting Pigment AI access to specific metrics? Or now that you can query agent referencing a specific board, should you grant Pigment AI access to every metric? Yes. Great question, Michael. Thank you. You should now, you should now enable this for all metrics. Okay? This is our strong recommendation. Okay? If you have some very, very sensitive blocks that you want to make sure that nobody in interacts with in the chat in the AI chat or VMCP, you can still keep those set to false so they don't so they're not accessed by by AI. But, otherwise, your experience and your business users' experiences will be the best when all of those, all of their blocks are accessible by AI. Great. I think another one for for you, Luca, that was also in the q and a was about the challenge with some of the instruction in the past to to create mission with the analyst agent. So can you please confirm to to the audience how it works now to to turn a chat into a into a mission? Yes. Absolutely. So today, you can use the analyst agent to perform an analysis, a one shot analysis, ask a couple questions about the data that you're seeing on your board, couple follow-up questions, etcetera, etcetera. Maybe you get to a point where you've collected a lot of interesting information and you want to generate a report, something that you can distribute amongst your colleagues. At this point, you can simply ask the analyst agent to generate a report or any similar phrase, and it will create a report just like you did with the missions previously. Similarly well, previously, we had some trouble with the with the missions in writing instructions. It was quite challenging to be able to create, very precise instructions that gave you the mission output that you were looking for. Now since we have the opportunity to create a report with the chat, we've included a button that's available on those generated reports that says save as mission. When you click this, basically, it takes that report that's been generated, creates a mission, and autofills the set of instructions so that you can continuously redo the same analysis. Amazing. Thanks so much. Okay. We have some time for a few extra ones. So, question for you on the modeler. There are some question about, like, is the modeler only to build net new apps, or can I use it to extend an existing app? So can you please clarify that? So you can create net two apps, of course, and you can also extend current app or maintain your app, and you can even ask question to your model. For example, if you ramp up on a new app and you wanna see you wanna know how it's built, how is this I mean, how how is this mystery calculated, what does this formula means, you can ask this kind of question and you can ask. But Anna's pictured it during the demo. She added, for example, a dimension to an existing model, and it's typically the kind of task that is quite tedious to do manually. And here's the model agent. Does it in one click, adapting all the formulas, dimensioning of the dependent blocks. So it's very powerful to maintain. And we have a dedicated tool made to understand current model and formulas. Okay. Thanks so much. Okay. I saw a couple also for Tongi on MCP. So Tongi, I think one was about the access right for MCP. How does it work? And, also, can you please just clarify what is accessible via MCP Because I think Andrea talked about the modular, but, of course, this is not the only thing that will be accessible and is already accessible in CP today. Okay. Thanks, Michael. So for the access rights, what, both the agents and the MCP do is that they impersonate the user. So when a user access connects to the pigment MCP server, they need to connect with their credentials. And so we can, basically, we apply their access rights, their permission. This means that by design, someone who is using our MCP server cannot do more than what they can already do in Pigment. They cannot see more. That's not possible. That's the first thing. Back to what Luca said, for MCP, we put a filter, and so only the metrics where AI is enabled are visible via MCP. If there are metrics that you don't want, you know, to enable the MCP, you can just deactivate the setting. Then about the capabilities that are available via MCP. So today, we expose basically the tools of the analyst. So you can use the MCP server to query and analyze data. We are currently working on also exposing the tools of the modeler, everything, to create blocks, to write formulas. They will be available very soon. And in particular, we are working on Clothecode and Cursor plug ins to allow you to install and use this in a very efficient way. This is an option that will not be enabled by by default to all users. You know, it would be an advanced feature so that only the users who wants it can activate it. Great. Well, thanks so much for all the answers today. Thanks so much for for all your help, Andrea, Anna, Lucas, Pangui, and. So we will continue to keep track of, like, all the question in the chat. We'll make sure that they are answered in in some ways. So just wanted to, yes, say a huge thank you for everyone for being here today. And before we close, if we can just go back to the slide, we have a few resources and next step for overview. So first of all, make sure that if you want to get access to the modeler agent, to the analyst agent, to the customer agent, everything you saw today, please reach out to support. So you had the link on the chat. I think we can put it again. Then if you are excited about what's ahead and you really want to hear not only about AI feature, but all the new pigments feature that are live, Join us next week for our quarterly product release webinar on March 19 at 4PM GMT. So we will share, as I said, all the new capabilities that are live now in Pigment, but you will also have the opportunity to connect with other pigment users in our virtual user group. And, of course, you don't want to miss the announcement of our pigment award winners, and maybe some of you are here with us today. I don't know. Finally, a quick note. If you are joining the next session with our PSP team and, should have received a different Zoom link. So don't stay here. Make sure that you switch to the webinar to join the conversation. And, again, thank you all for being here with us today. I really hope that you enjoyed the presentation, that you enjoyed the demo. From our side, I can tell you that we are super excited to shape what comes next with all of you. We really appreciate all your feedback, and thanks so much again for being here with us today, and see you next time.