Name: Ember Agents: Live Demo & Q&A
Uploaded: 2026-03-17T18:16:32.603Z
Duration: 44 min 16 s
Description: Ember Agents: Live Demo & Q&A

Transcript for "Ember Agents: Live Demo & Q&A": Hello. Good morning for those on the West Coast. Good afternoon, for those on the East Coast and from from everyone around the world, thank you for joining us today. Really excited about today's product demo webinar and, got some really exciting things we're gonna show you. Gonna hand it over to my cohost in a second here, Wen Shu, but I'm the founder and CEO at Campfire. Previously, I was a finance exec and saw that there was no modern ERP that, that we were really excited about. And so as we scaled and we after we got acquired, went on to found Campfire to be the modern and AI native ERP for finance and accounting teams like myself. And so I hand over to to Wen Chew from AI engineering at Campfire. Yeah. Hello, everyone. Welcome. I'm a AI engineer here at Campfire. I joined about five months ago, and it's been a really exciting time. I've been working on Amber, our chat agent previously, and and more recently, I'm building out our agent platform, which we're really excited to share more about with you today. Alright. So I got a few slides, and then we're gonna hop into the product. We're gonna have when to kick it off on chat versus agents within the Campfire platform. Yes. Yes. So I think some of you might be familiar with our agent, Amber Chat Agent, where you basically have to initiate everything. You ask Amber, you know, what do you wanted, and then the responses stay stay in chat. So it was kind of limited surface, where you were presented with some of the data, and it was kinda read only, and and you couldn't really kinda work through your data. So Amber would advise, but you still had to, you know, kind of do the execution yourself. And some results weren't durable, so we didn't really have, like, a permanent sort of workspace, to build or or to, you know, sort of return to. And, one of the things that we saw that was challenging is that you had to, you know, sort of start every session from scratch. And, what we've done is we, you know, kind of evolved this, and now you manage, from this sort of agent workflow. You now are able to manage multiple agents that would work autonomously on your behalf. So the agents, they can do things such as, you know, auto code transactions. They can match your ARAP. They help you flag anomalies, and that is all done, you know, without you have to explicitly prompt with them. So it's a lot more preactive. And then from accrual and flux commentary perspective, you know, now we've added more permanent surfaces to them, and they are also built into the whole closing workflow. So you can configure the rules. You can configure your agents, and then it's all always working for you in the background. So they are, you know, working on on your task recurrently. And you can also from the configuration standpoint, you can configure these agents. You can configure the confidence thresholds. And high confidence actions, they kind of post themselves. You know, everything else, you know, they kind of go through your review queue. So the shift really is before, you know, you had to interact with Amber. Amber waited with for you, which is still available. But now we took some of the more specialized task and and created very specialized agents based on your need. Right. And so just talking a bit more about agents, and we'll go through a live demo of it. So we have an accounting intelligence, and this is running Campfire's own foundation model. Many folks ask, why did we build our own model? And it really comes down to three things. Performance of the model, it's, fully built for finance and accounting workloads. So it's not horizontal, but it is incredibly narrow for these tasks and incredibly deep at these tasks. Also, auditability. You know, so many of our customers are audited and publicly traded that our ability to open the hood of the model, provide full attribution, provide why an actual match was performed, write write accounting support, work papers on on an accrual that it generated. These are the things that folks look to us for. And then, of course, security. The data never has to leave the Campfire environment because the model is running in our own environment. And so this is powering our core agents for Ember. And so we'll talk through we're gonna talk through custom agents today, but, essentially, this is an area for all the long tail of tasks. Anything that can't be done by a default agent, we're gonna hand over to to custom agents. A little bit I'll kick this one off, and Wenshu, feel free to chime in. But we briefly talked about earlier that everything was prompt driven, and customers were saying, hey. We're looking for more of that continuous background. But, ultimately, the agents at Campfire, are all aligned around two activities. One is moving folks to a faster close and a continuous close for for those that wanna get to that point. And then the other one is financial insights. And so as a previous finance leader, we were spending so much time preparing the data, and there wasn't enough time left over to actually, like, execute on the analysis, that we've now switched that. And part of that is, of course, just Campfire's reporting and just the ability to grab any data you want easily and not have to go into spreadsheets to for custom reports. But more importantly, the long tail, whether it's, you know, LTV to CAC by product line or average days to pay by customer segment or some of the deeper analysis that maybe is not in a in a typical report. Folks are now saying we want to have this running, and we wanna have it, you know, asynchronously sending it to me via Slack, you know, 8AM Monday morning so I can kinda have a cup of coffee and look at this deep analysis that was performed at really any cadence of your desire. And so I think this was really where we have taken things at Campfire with this this kinda asynchronous background workloads that can be done. Alright. So let's hop into a live demo. Okay. So I'm gonna start with, the the new agents because that's the newer part, and then we'll talk through the, all the big changes we've made to the chat interface, which includes custom agents. So we're gonna start with the continuous agents. And looking at, we'll I'll start kinda left to right. So the audit agent here, this is going through all of the accounting tasks being performed in your account, and it's just gonna be constantly flagging. And so there's two setup on this one. And, you know, in the next two weeks, this this list is getting a lot longer as new task get added to it. But, ultimately, I'm getting a a daily Slack ping, four new items for review, and I can kinda click in from from Slack, and I'll show that in a little bit. But, ultimately, like, duplicates, high value. I'll show you some of the other monitoring, for subledgers as well in a second. And then what I can do is I can turn it on. We can look at the runs for it, so every time it's running, and then all of the actions coming out of it. So whether it's a high value transaction, maybe there's an anomaly we wanna take a look at what happens or a duplicate entry, and we wanna click in and take a look, you know, at this duplicate, that ultimately, it's just continuously going in the background and flagging things for your review. You can dive in and then make adjustments. Then if there was anything done in the past, we can always hop in and look at the audit trail of actions performed. On the continuous close, there's a lot of talk about zero day close. And at Campfire, we've moved beyond that legacy way of doing things into a continuous close state where for customers, if desired, the smallest unit of accounting no longer needs to be a month, and it's continuously running through kinda all activity in the account and automatically performing merges, matches, coding. You know, even if it's a custom dimension that's been set up just for your environment, our model is actually running individually for each customer, our foundation model. And so what that means is all custom dimensions, all configurations that are unique to you, you know, the complexities of your entity structure and chart of accounts, It picks up all of those news nuances, and it's being updated essentially in real time to ensure that every time you feed a new information, like, you no longer wanna use a specific chart of account for, you know, your anthropic bill, you wanna move it from one, you know, cost of goods item to another, that, ultimately, it's gonna, like, rapidly pick that up, and start to adjust how workloads are performed. So the old way of doing things is very rule based, and you set up lots of rules to manage everything. And then the long tail, we're largely manually manually updated. In Campfire, with the continuous close, the model is looking through kinda obviously, the rules will still run. If you set up rules in your account, we have that, and it's very powerful. But for everything that doesn't hit a rule, again, whether it's a bank transfer or transaction coding, everything will be, automatically coded if desired, and we put you in full control. And so folks are saying, I want to be able to really, like, decide what happens when. And so we have, you know, auto approve, threshold, or fully manual. And let's say I wanna set a threshold. You can actually see here the model is looking through, and it's saying that we actually think 70, we have a 100% confidence on, and there's 72 this is sitting at, like, the the eighty fifth percent percentile here. There's 72 that actually maybe a ton of customers pay me on the same day, the same dollar amount, and the the bank description all say JPMorgan. And so it actually has a decent understanding of which one's which based on historical data, but maybe it's not fully confident. So, actually, I want those to come to me for approval, and we can manually tie them out. But then if we actually wanna take a different approach and, historically, this yellow bar of 72 has always been accurate, then we can actually go ahead and adjust. And so maybe we wanna do an auto apply. And so now all 142 are gonna be auto applied, or we can set dollar thresholds as well. So if it's over a thousand and it needs a 100% or or 85 percentile and above, you can see everything is, Then, so lots of configurations in here, while allowing you to move to, hey. At this point, I fully trust the system or, hey. We're entirely manual. We're a public company or something in between. And then we can have a Slack ping, and so we can set that up here. We can get email alerts. So I wanna just have it kinda coming to me, and we'll go ahead and make these changes. Accounts receivable. So many of our customers came to us and said, hey. We're on a legacy system. We're manually coding a lot of invoices. And so this will handle partial matches. This will handle full overpayments. It's gonna allow you to really go in there and and perform work. And so, you know, in this dummy data, there's a lot of noise sorry. In this dummy, this sample account has a lot of dummy data. There's a lot of noise in here, so it's gonna it's gonna show lower. But we're typically seeing folks in the green. But, again, if desired, they can always change dollar thresholds, flag for review, look at all actions performed. And you can see here, we can, you know, we can view if we wanna be in full control. And, yeah, so that's that's accounts receivable. Accounts payable, very similar, but this goes beyond the matching. You know, it is also looking at, like, vendor and customer names and helping you merge them if there's a fuzzy match like Zoom and Zoom Inc. So many of us are, struggling with that, that it's very manual task to keep it clean. And then fixed asset capitalization. So we'll actually build this one from scratch together. This is something that's candidly brand new. We shipped yesterday. So I'm gonna go ahead and just grab I had, Claude write me a policy. Give me one second. Okay. One second. Let me refresh that. K. I put in the wrong one. So it's reviewing my fixed asset policy. It's, like, five pages, and, it's taking all of the rules, and it's gonna build a list of all of the rules. And then what we can do is go ahead and edit them if desired, and we can add rules as well. So, you know, one of the key ones for us would be, like, materiality, threshold. We, you know, we got a long tail of items. Do we control the asset? You know, furniture and fixture, we're using a five year useful life. And I can even you know, let's actually make some tweaks here. You know, I wanna actually do a three year, and we can delete items that came in. You know, unlikely you'd wanna do that, or we can add more rules. So we'll go ahead and and activate this now. And you can see here it actually built out the rules engine, for for a fixed asset monitoring. So now let's say we have an AP or PO solution. We're syncing items in, and someone forgot to capitalize a fixed asset in another system and send it to to fixed assets to be, auto sent to, the fixed asset subledger, this is actually gonna go ahead and flag it. And so none have come through yet on this one, but, essentially, it'll flag. So let's say there's, like, a you know, I think we had a, let's say there's a a $30,000 fixed asset that didn't get sent to fixed assets. It'll see it. It'll flag it for us. And so many accounting teams are running very lean, and the companies are growing fast, and they're saying it's hard to catch everything. And so this is just gonna help you continuously monitor. And in the next week, we're rolling this out for really all all subledgers and all accounting policies, whether IFRS compliance, any all the country specific items, even if it's for one of your subsidiaries that this is an example that we now in our sources here, we now have our fixed asset policy, and we could continue to add all of our policies. So that's one that, personally, I'm really excited about. We've got a couple offices at Campfire, and so it's keeping whether it's a a expensive laptop or furniture, it's helping us stay stay compliant for our audit. Moving on from that one, we're gonna talk through the on demand agents now. And so there's two that we've launched so far in the last week. The first one is on accruals. So many of our customers are saying accruals are very manual, and I hate doing accruals. And so I can go ahead and take a look at accruals going through all of the accounting. And at the end of the month, it's gonna be auto drafting suggested accruals. So here's one, just a sample recruiting firm, that it looks like we have, an item that should be accrued for. And if you're like, you know, why did that happen? So we can actually look at looks like we've accrued for three different three months, and, likely, we have all been this before been through this before. The bill comes on, like, the tenth of the month. We're trying to close on the fifth. So there was there was an accrual, and so Amber caught that. And so Amber added it. But I can also add my own accruals. Maybe, you know, Amber did not ingest or did not there's, like, a brand new vendor that we signed, but they haven't sent a bill yet. So, there's no way for Ember to know about it. We can go ahead and manually create accruals. And so this subledger is gonna run into perpetuity, and you can manage all of your accruals. And then we can go ahead and draft. And, again, we can move to a continuous state, or we can do kind of a this account set up in more of an end of month push, but essentially drafting all of the accruals for the month from the new accrual subledger. Okay. The other one I'll show for on demand is Flux. Now customers were telling us we love the Flux that we were getting out of Ember chat, but, ultimately, we need a permanent surface in Campfire to manage Flux. We wanted something that the team could work through together. We wanted everything, for, actually, the all of all of the workflows of performing the flux. And so what we shipped here was I'm just gonna grab, you know, one that I already built, is we've got a permanent surface now that it's still doing you know, it it's still writing all the commentary like before. We can now edit it before we kinda submit. It's linked to the closed checklist, so we can now push this to the close for approval. But a first pass of all of the all of the flux commentary and, by the way, when we're setting this up just kinda hop back in here. When we're setting this up, we can choose, hey. Actually, I wanna do everything, you know, by department or by GL code. How do we wanna think about materiality? Should this be on the closed checklist and maybe a recurring template? Am I just doing a one off because this is a year end, and I just wanna do a very custom flux? But then once it's set up, all of the commentary is written for you, editable for you, and really allowing you to streamline and automate and have full workflow of end to end flux. And suppose it's staring at a blank workbook for the commentary, we now have everything drafted on our behalf. The actions tab is where you can see all agent activity, and this could be things that it's already performed. Like, you can see here, we've approved some work here and all pending. So we built and this is little preview of the next section, which custom agents, We'll talk to you in a second as it looks like the custom agent ran and found something I should look at. But we can always filter on specific agents. We can dismiss all. We can manage. And so, again, you can be in an autonomous state, and you can just look at the audit logs of all agent activity, or we can move to a threshold base where we're in the middle, or we can do a full manual because we've got, you know, big four audit or we're publicly traded, and, you know, PCAOB is having us take a more conservative stance. I'm going to hop in now to custom agents. And this is one, that was told we're we're the very first ERP to have something of this nature. And the ability to take, what we've just built, which, like, all the default agents, but allow you to actually build anything you want. And so we're gonna work through an example here. I wanna have it, talk to many customers, and they say, I do a lot of, like, manual adjustments at the end of the month. And so we're gonna do an adjusting entry together. Alright. We'll just do, like, an accrued expense, and we want it at the end of the month. Yes. We want a reversal. And can you grab the Stripe? I'll just drop a bill and send. So let's say we have a free tier. And so part of our COGS, we're sending to marketing because people on the free tier, you know, our auditor allows us to push that below the line on gross margin, send 30% to marketing, and keep the 70 percent. And so whether it's, like, a department allocation, GL allocation, customer allocation, many of us that are AI natives are saying we're now doing contribution margin by customer. And so any sort of allocation and so with us, folks are saying, I'm setting up custom agents for really all manual tasks, including custom dimensions, and now I'm actually able to have much better visibility into the business. Now we have the capacity with the autonomy here to really do geo level contribution. We wanna do a p and l by business unit. There was a lot of maintenance on doing that, and we've now moved to a state where all of the maintenance is automated once it's set up. And so I went ahead and applied the split, and I'm gonna have it create a custom agent. You can see it's gonna reverse on the first. Okay. So we now have a custom agent that's gonna perform this on the first of every month at 9AM. And let's say I wanna make some adjustments. You know, I can easily make adjustments here. The full kinda agent can be managed, And I wanna you know, let's just go ahead and run it now. Okay. So we've got a new agent, and it's been fully set up. And now in our agent center and you can see I already created one of these. But, ultimately, we now have a custom agent that's been been been built out. So many folks are saying, my closed checklist is literally everything that is still manual for me to perform the work or review to ensure everything's accurate? How do we take everything that's still manual, the long tail of closed tasks, like a cost allocation and or an accrual, and we're now moving that into a custom agent or a default agent workload? I'm gonna that was one of the themes that we talked about earlier around continuous close and closing faster. Now we're gonna look at one that's more on the financial insights. And as a former finance leader, this is probably the one I'm most excited about that it can build reports, analyze data, can do research, it can do cash forecasting for you. So why don't we do a cash flow forecast? And the good thing about Campfire, have all your subledger data. We have very strong granular visibility of GL code. Let's say you're an s l you know, sales led business with heavy seasonality. Right? Campfire knows all of that in the historical dataset. So it can be very fine tuned for cash flow forecasting. Do I wanna look at all the subledgers and historical data? We wanna do both, and we're gonna do a monthly forecast. And whether it's complex margin analysis again, earlier, I mentioned, you know, LTV to CAC by product line, by customer segment. This, like, deeper long tail of analysis, we can now build an agent for, and it'll kinda autonomously be running and sending it to us, or we can have it on demand when we need it. So it's now building out our agent. You can see it's going through all subledger activity, GL code activity. And just to move it along here, actually, that we can essentially have you know, we've got a cash flow refresher, cash flow agent, you know, gross margin reports. So it's it's essentially picking up a lot of a a lot of the financial analysis now at Campfire. We can automate a workflow. So we can even pull in data starting next week from from email inboxes. So, like, for accruals, let's say your lawyer sends you estimates, now auto generating accruals off of off of emails for you. But I know we're coming up on the the thirty minute mark, So I'm gonna take a a quick pause as we're gonna start to get into into the q and a here, and I'm happy to keep demoing through the q and a if people have specific questions. But let's let's take a look. Someone asked about budgeting tool. Yeah. Campfire has a full budget versus actual functionality. So that's outside of the AI demo. But, ultimately, if you wanted to build a budget and then load it into Campfire's budget versus actual, then, yeah, it can be helpful there as well. Can you upload a revenue forecast? It's a great question. The as this group all knows, the further out a forecast is, the more third party context that the historical dataset needs to be accurate. And so the question here, you know, is really, like, how do we make it more accurate on the forward looking? And, yeah, uploading a revenue forecast, a headcount plan, that upload button we were looking at earlier, you can really have it, you know, put any third party context you want. I would still recommend going with an FP and A tool, an EPM tool, or Spreadsheets for long term planning. I don't think this is really built for long term planning. But for short term visibility where the granular historical data is incredibly valuable, that I think that would be a good place to start is we're not thinking about, like, your three year plan or your your one year plan, but, like, show me cash at the end of the month, estimate cash, you know, '2. I think that's where it can be can be helpful. Variable accruals month to month, I think that's prob or recurring JEs. I think that's probably the best use case is that's actually we've seen tremendous value. So at Campfire, we'll run some of the accruals like a three month trailing average, and Amber is just kinda auto performing the work. Or if there's some other variable trigger that you wanna run, like a human context, again, or off of some, you know, email estimate that you're getting, the custom agent will be able to pick up that workload. Does the flux allow for statistical info beyond the GL? Absolutely. Yeah. So, you know, Luke, we'd wanna go through offline more specifically of what info you're looking for, but there's a lot of non GL data that you can be flexing or dimensions you can flex. Does all agent activity require approval before posting? That was what I was showing around you can choose. So you can move to an autonomous kinda continuous state where there's no approval required on the agent activity. You can move to a hybrid where you set confidence and dollar thresholds for approval, or you can move to we need full control over the agent. We're a public company, and so then it can't do any activity without your approval through campfire through Campfire's full kind of approval routing, workflows. Is the Ember suggestion posting to the GL? Yes. It is posting to the GL. So all of these since Campfire is the general ledger and the subledgers as well with Ember, that ultimately, yes, it's posting to the GL. And, again, if you have approval threshold set up or if you have auto approve, then you'll either be in that continuous kind of posting state, or you'll be in the, you know, we need manual approval on everything. Wenshu, are there any that you wanna pick up that you see? I'm just trying to get through all the ones I'm seeing. Yes. I think, there's a question here by Ian, whether the commentary, seems to be in dollar despite, you know, only, you know, I think their currency, the book value is in pound. Right? So I think that's, that's something that we are addressing right now, and I think you should you know, it should reflect on book value, you know, probably tomorrow or the next day. Is extensive coding experience knowledge needed to run or create an agent? No, Samuel. We just created an agent together. It was prompt based. It'll actually prompt you to build the to build the prompt together. And so zero coding experience required. This was meant I'm nontechnical, and I build kinda been building custom agents in in Campfire since we went into private alpha without ever having seen a line of code. What LM are you using? For the ingestion of the actual kinda query because it's so broad and natural language, we're using Anthropic. But, ultimately, all of the the models are running or all of the agents are running on Campfire's foundation model that we build and maintain here in San Francisco. And as noted at the beginning around performance, auditability, and security, that's why, you know, we've taken the step to to have our own model and told we're the only ERP that is running our own foundation model. And so it allows us to run this vertically integrated stack of kind of our own infra. We got our own GPUs, our own model, our own agent layer, and then the general ledger layer. We're kinda owning all of it, and so it's all tightly coupled together to allow for folks to have, like, this very high performance in an audit environment and with the incredible, you know, control. As we looked at earlier, you can set confidence thresholds, dollar thresholds. There's a lot of fine tuning that customers are looking for from us, and so we've delivered that. I will say, like, we have our own MCP. We've got, you know, APIs on our website. So it doesn't need to be like AI is performed at Campfire. So I, you know, I build our board slides. Like, Campfire's AI doesn't build slides for you. So I build our board slides off of Campfire's MCP. For those that don't know, that's model context protocol. And, ultimately, like, I can go go into Claude Cowork, have it build slides off of the Campfire connector through just chatting with Claude. You know, here's the last board deck. Can you build this? Or here's three new slides I wanna create. This time, go into go into Campfire, create these slides, and write commentary on page one based on your observations. And then I, you know, fine tune. I check for accuracy on everything. You know, I you know, I'm still a cattle for all the work, and then I and then I ship. So I really you know, previously, I was editing x and y axes in in Google Sheets and Excel all day. Now that's being done for me. And so how we have performed financial analysis at Campfire has fundamentally changed. And and if I may add also to answer, Tina's question here, and the question is, you know, can you use can you use the FPA tool in conjunction with plan and sales goals for forecasting and headcount after the finance portion. I think, similarly, the answer is that you can, you know, kind of integrate, these tools with MCP server and the data layer that, you know, that we've allowed access via API as well as m m MCP servers and really sort of extend the capability of these FPA tools along with Campfire. We all know how fast Campfire iterates on new features. Thank you, Chris. Appreciate that. So I'm curious what other agents are currently in the works that you can tell us about. We're gonna be shipping new well, first, we're gonna see how much the custom agents can pick up. I think if you think about the custom agents, like, if if we can get them to a point where they can pick up all tasks that are not picked up by the default agents, then, Chris, that will be kinda so that I would say that's a big focus for us. The other one is pulling in as much context as possible. So we're connecting to all the other m c p you know, we're, of course, integrated through API, through hundreds, thousands of different piece of technology, whether it's, you know, accounts payable or or billing engines, like payment processors like Stripe. But now we're connecting to the MCPs of them all. So now our agent can autonomous performing tasks not just on our own data, but also connecting agent to agent into your entire tech stack, whether it's your CRM. And it's doing a reconciliation of all data in your CRM versus Campfire for closed won to tie everything out. Example, a deal closes after the quarter, but the sales leader decides to adjust it to an in quarter deal. And so, you know, there's a start date adjustment. It's gonna help with the reconciliation that, via API. That's a little harder to do. The MCP, we're able to you know, that what we're just looking at earlier, the audit, the monitoring agent is now picking up things like that in the next iteration that we're launching. But our customers set the agent road map, Chris. So I think, you know, we just love feedback, and we'll continue to ship if there are any gaps. I think I answered this one, but if not, I'm sorry. Can you reverse all agent activity postings if needed? Yeah. So you can absolutely do that. I mean, if you're nervous about the agents, of course, just treat it like anybody else on your team. You've got a new intern. Probably don't wanna let them just post, you know, on their own, have everything submitted to you for approval. You know, of course, if you're public or in Big four audit and you need kinda ITGC, you know, controls framework where everything needs to be submitted for approval and another human has to approve, then you can take a different approach. But, Alice, you can absolutely undo agent activity. You actually just prompt the agent extremely enough, and it'll kinda reverse everything out that it's done. Obviously, you need to be in a you need to open the accounting period, and there's, like, some steps to do that, but it is all reversible. But as I showed, it is all auditable as well. So all agent activity is kinda has full full audit support. Here's a question from Roberto, about, having to upload, Excel files to prepare for a month end close, for revenue recognition waterfall. And so the question is, you know, if there is, like, a custom file or a custom data source that he uses, you know, how that can be, used by Amber and our agents to help him accelerate close. And the answer to that is that, yes. I think, you can just treat the Excel files or your worksheet as additional data source. You know, you can add that along with all the other data sources that are needed and you can essentially, ask Amber and configure agents to, you know, sort of, pull from all the data sources and prepare for, you know, Yeah. this report, There there's described. that upload button, and then Campfire answer is yes, have a Google Sheet. and Excel integration. So we can also kinda pull it in as well via our live sync. Is there a way to train and teach the agents based on our own data? Yes. So one of the beauties of having our own foundation model is the agents are actually every customer inherently gets their own fine tuned model. And so, literally, every time a task is performed, even if you're telling the agent no, that was wrong or yes or just, you know, something synced in from a third party system, the agent is continuously learning. And so this is one that, you're in a very secure way, like, your data, you know, is actually training the model that is uniquely kinda run and fine tuned, within your own Campfire environment. So this is the step we've taken that that from an ask if I perform a a task today, you know, the next item that comes through will inherently be trained. The model will be trained for that next action that that that it takes with that feedback. I think there was one question around, like, you know, visuals. So let me just kinda rescreen share super quick. But just I know we're coming up on time, but I let's give this a second. Seems to be loading. I'm getting an error on the screen share. But, inherently, it can build all sorts of charts and visuals. They're interactive. They're downloadable. You can drop them into, you know, Excel, or you can screenshot. And then the agent, you know, again, let's say I wanna, like, very custom weekly cash flow or end of month cash flow forecast that I wanna get a Slack ping, but I wanna actually all the I want all the charts. I don't I don't wanna just see a number. I wanna see a visual of, like, the full walk, the bridge of how we get there from cash in, cash out uses and sources that ultimately, the visual that I'm trying to show is actually just that. But, in a follow-up, we can send some more info on that. Can we get the agent to ping law firms for monthly updates of their work in progress? Sounds like that one resonates with the group. I see five upvotes on it. Gosh. I think Kjell might have gone off stage. Yeah. It seems like we have some you know, I think that just to wrap up what John was saying, I think if that sounds like something that we'll be working on. I think being able to, you know, extend the ability of aside from, you know, kind of, intaking from from the invoice, from the attachments, you know, being able to kind of, do these out outbound reach outs or recurring reminders. That is a capability that we would also like to integrate with our, you know, highly capable Ember agents. So I think with that, we're gonna wrap this up. Thank you everyone for attending. It's been really exciting sharing with you. I hope you enjoyed the session, and you learned a lot more about Campfire's agent framework. Okay. I hope everyone have a good day.