From Friction to Flow: Orchestrating Customer Value at Scale

Good afternoon, and thank you for attending the session. And right before lunch, we promised to keep it under 45 minutes and give you plenty of time to ask questions. My name is Arun Parikh. I'm a senior director of customer success at Boomi, been with the company for eight years.

I lead a global team of around 50 CSMs across the geography. Joining me on stage today, we have Emily Mangan, who we love to say that she does all of our cohort operations, strategic projects at Boomi. If you want anyone to do anything, we go to Emily to get it done. We also have Matt Crisbach, who is director of revenue operations.

And I'd like to say that in Boomi, that hey, we have a lot of comp aligned metric, and Matt and Matt's team make sure that all of us get paid our commission checks on time. All right. So, oops. Okay.

Look, in the next 30 minutes, we'll keep it short. We'll talk about where we started 18 months ago, 18 to 24 months ago. The friction we faced, whenever we want to do a transformation in CS, we always would see a bit of a friction between different teams. We have account managers supporting our customers.

We have CSMs supporting customers. We have technical account management teams, et cetera. So we'll talk about some of the frictions we faced. We obviously started to build some playbooks and blueprints, and we'll talk about what blueprints we created in the light of AI at the center of all of it.

And most importantly, I was attending a session before, but most importantly, we started with, hey, why change? Why change? What are we changing, and how are we going to be measuring the change? So we'll talk about the metrics we measured.

A little bit about Boomi and the space we operate in. We are number one independent data activation company. We have about eight unique lines of platform components, and we help customers across all shapes and sizes automate their workflows. We have API management, data management, AI management, a full stack, a very comprehensive product stack.

We have been a leader in the Gartner Magic quadrant for 12 consecutive years, and we're the only independent one. Our biggest competitors are Microsoft, SAP, Salesforce, Oracle, et cetera. So what we do is not hobby. It is our passion in terms of what we do in our product team and our CSM team.

We have 30,000 plus customers globally. We sell in many different motions. We have direct customers. We have OEM customers.

We have resellers, referral, all sorts of customers. A pretty, pretty big book of customers across the globe. And we obviously are operating in the AI age. We have 75,000 plus customers with active AI agents running on our platform.

So I talked about where we started. I'm going to talk about roughly 24 months ago where we were. So we had-- so the team that we help is the direct customers, and there were like 5,000 of them. And I'm not going to talk about segmentation quite a bit, because I know a lot of the sessions in this team is about segmentation.

When your customer base keeps expanding, when you have customers paying anywhere between $10,000 to $10,000, you obviously have a relativistic segment. So we obviously-- segmentation also has ups and downs in terms of how do you tailor experience, how do you personalize content for them, how do you put people behind the right customer the right time. So we did all of that. That was a real problem for us.

At that point in time, our CSM to account ratio was 1 to 70. We traditionally had 1 to 1 CSM to account ratios. We obviously had a lot of smaller customers that were primarily what we call a scale team or a pool team. But for any customers who were paying anywhere more than $550,000, we had a CSM attached to them.

And that ratio was about 1 to 170. And our NRR was solid for a period of time, but then we saw a bit of a decline. And we had to close 5% of NRR gap. And for those of you operating in private equity, we own by a private equity company, we operate under strict rule of 50, which means we have to show 20% top line growth in terms of acquiring new customers, expanding new customers, as well as 20% in efficiency.

As we scale up, as we scale up, as we have more customers, we couldn't justify that, hey, for every 20 customers or 30 customers or 80 customers we will acquire, we have to have one headcount. That was a challenge from our CFO. The board wanted retention metric. They only cared about retention metric.

Everything else is vanity metric for them. And our CFO made sure that, hey, we have to make sure we segment in a way that we don't have to have human scale as we scale our customers. So again, like many other companies, many of you here, we started a transformation journey in terms of, hey, let's look at the metric that we are holding ourselves accountable to. And what do we have to transform?

I will have my colleagues, Emily and Matt, talk about the two pillars. But I want to talk about the first pillar. The human-led transformation is super important for me. And because I've been in the CS space for many years, I know how incredibly different is transformation.

Change is hard. People resist change, and change is hard. So what we wanted to do was we wanted to, as a CS team, bring a narrative around why churn is important, why we are changing. The ultimate metric that we're trying to measure and improve was NRR.

And a big component of the NRR improvement was, hey, saving bottom line churn. And churn is-- in most companies, churn is typically, hey, CSM owns churn. But churn is a company metric. Everyone is responsible for churn.

And everyone is responsible for customer success. Some people carry the title, but everyone is responsible for customer success. So we, first of all, did was we made the whole narrative around why churn is important and why everyone has to care about it. So $1 of churn is not just $1 of churn.

It is $3 spent in customer acquisition cost to bring that dollar back into the company. And for a PE company like ours, if you had to go to IPO or to be bought, the general valuation for a company is between $13 to $20. So $1 worth of churn is a 20x loss of valuation. So that narrative, we made sure everyone in our company understood from sales to services to support to everyone.

That, hey, churn is super important. Everyone has to care. And then the second problem we faced was we had a pretty-- not a very unique go-to-market motion or go-to-market team. We have account managers.

We have customer success managers. We have advisory services people. We have a lot of people supporting our customers. It's a good problem to have.

But the bad problem to have was all of their KPIs were misaligned in terms of caring about the right outcomes. So when a customer would say, I am checked out, stay with Boomi, or for whatever reason, we would have some folks say, hey, not my problem. Because it doesn't affect my comp. And comp drives behavior.

So we sat together and redesigned the whole comp in terms of, hey, everyone who's touching an account would have to care about churn. So we changed our comp matrix to be more retention-heavy and retention-driven matrix. And then we also acquired-- we also did some other changes in terms of, hey, every CSM have to forecast renewal six months out. And then it became 12 months out.

12 months was not enough. I did a session yesterday with another colleague of mine about a platform we acquired which helps us predict churn 18 months out. Because if we know about churn six months out, we are too late. If we know 12 months out or 18 months out, we at least can do something.

And I know a lot of CSMs and account managers go very reactive or very protective when we know there is a churn going to happen. But what we also did was we also changed. We fundamentally created a narrative around, hey, we will pay you for your heroics. So now every quarter, every quarter, and three or four quarters out, we are identified 100 accounts that have the biggest churn pressure.

And we provide SPF on those accounts. If we save those accounts, a SPF is paid out to everyone who's helping save those accounts. That has created a big impact in terms of how we have been able to save a lot of material churn which no one would probably care about. So with that, I want to let my colleague, Emily, talk about the other two motions.

Thanks, Arun. Hi, everybody. My name's Emily Mangan. So in this section, we're going to double click to the AI augmentation pillar that Arun kind of mentioned on the slides and really talk about how AI has transformed our customer success organization.

Now, there's three of us on stage here today. But as most evolutions, it takes a village. And this was really possible thanks to our organization, our leadership, really making this a priority. Really strong cross-functional partnership to really get our vision live.

And then numerous CSMs investing their time to provide feedback, test out these agents so we could really refine and make sure that we were releasing versions that made sense for the organization. So before we could really start rolling out solutions, we wanted to make sure we understood that we were tackling the right problems. Our goal here was to be able to leverage AI to improve both how we're working internally, but also, really importantly, how we're engaging with our customers. So these issues boiled down into three main areas.

First one's adoption. We wanted to be able to proactively inform our CSMs and also our customers about where they are on their journey with Boomi. We also needed a better way to catch an alert on accounts that were going dark silently, which was the big problem, and figure out a way to get them back on track. Second, we had engagement.

We wanted to be able to really tailor insight and next steps relevant to every single customer instead of using more of like one size fits all approach, where we have a template QBR deck that we're hoping gets augmented with the right information. And then last, we have productivity. So we needed to be able to reduce the manual lift for our CSMs to create impactful content for our customers, giving them time back in the day where they can actually be interfacing, interacting with their accounts. All right, so how did we do this?

So with those problems identified-- can everyone hear me? Is this good? OK. Thank you.

So with those problems identified, we dove into the work of actually building out our AI, our agent roster. Now, Matt will cover in a few slides that we didn't necessarily hit it out of the park immediately. We learned a lot in our initial attempts, thanks to a lot of our CSMs trying it out, giving us that feedback. We tweaked a ton.

We are continuing to tweak. This isn't just like, hey, here's our final version, and we're going to set and forget it. This takes constant refinement. But now we have 10 AI-powered solutions live today that are really leveling up our customer interactions and how we work internally.

So I'm going to dive into three of these AI-powered solutions and the impact that they're having on our business. So I'll start with our monthly adoption insights report. Now, on the next slide, I'll give you kind of a preview of what this looks like. But this has definitely been, I think, the most popular agent product that we've released for our customers.

So each month, our CSMs will prompt the adoption insights agent for their customer portfolio and get a polished, customer-ready email that provides their customers with an executive summary of their environment, a review of what has changed in the past month-- so what processes are running more or less frequently, failing more or less often, and also a spotlight on how they're realizing value in their organization based on those processes. So taking it that next step. We're sending around 2,700 of these reports monthly, so a scale that would have been absolutely impossible pre-AI. And that'll be, as you can see by these numbers, a very common theme on this slide.

AI is allowing us to do things at a scale that we were never able to achieve before. So next, we have our prescriptive growth roadmap. So this is a component of our QBR conversation. And this roadmap is completely tailored to each customer to provide them with a view for how they can enhance, extend, and transform their existing implementation.

And this is based on where they're at in their journey with Boomi. And also, like, peer examples of production use cases that light the way for them. So think about this as a white space report on steroids. It's not just highlighting the opportunity, but it's showing them a real world example of how a customer has implemented this and the benefit that they're seeing from it, which is pretty incredible.

And then lastly, we have our account transition agent. I probably don't need to explain this to anyone in the room. I think we all know the pain of there's a new logo that just signed, and now we have to get the seller to explain the history of that deal when they just want to move on to the next deal. Or having a CSM that's changing roles or is leaving the company, and now all of a sudden we have like 50 accounts that we have to transition to other members of the team or to a new hire.

And this was just costing us our most precious resource of time, right? Time spent on calls, doing an oral history of the account, trying to trace down documents. What's the latest contract? What professional services engagements do we have?

And now we have this agent that produces a very comprehensive brief that highlights key conversations over the last six to 12 months, depending on the range of interest. It'll attach key documents. So the account team, anyone on the account team that's changing has that up-to-date information and knows where to go. All right, so with that, I'll show you guys what our monthly adoption insights agent kind of looks like.

So we did kind of a before and after on this side. So on the left side is more of our before state. And our CSMs still use these dashboards. We really do have some great Power BI reporting available to our team.

But as you can see, it's like a ton of data to wade through and figure out what is noise, what is signal. Make sure that our CSMs are interpreting that data correctly and raising the right insights to their customers. Our adoption insights agent does the waiting for us, right? So it's surfacing these key findings in an easy to digest email for our customers.

And it's based on data our CSMs are already super familiar with. So if the customer does want to double click into anything surfaced in this email, they know exactly where to go to dive deeper with them. So it's been really incredible to see the positive response to this report from our customers. We had one of our customers just email us like a couple weeks ago saying that this is allowing them to validate development readiness and maturity and to identify risks before they surface in production due to these leading signals, which is exactly what we wanted to hear.

It was really that validation that this time, this effort we were investing was going to pay dividends across our customer base. So with that, I'll pass it over to Matt to explain how this all came to life. MATT BERKOWITZ-SCHERR... She trusts me with a clicker.

Maybe you shouldn't. So you've heard from Arun and Emily about what we set out to do. Our North Star goals, a little bit about our coverage model, the agents we built and the outcomes we delivered. All of that is true.

What's also true is the path to get there was not cleaned. So the video that you probably saw yesterday of Chuck cutting through with a machete of a forest to find a path was kind of what it felt like when we started. And we made decisions early on in the build that cost us months of time and in some ways lost confidence with our CSMs because it eroded trust with how we rolled certain things out. So the next section is going to be a really honest version of what happened and how we got to where we were.

And I'm not going to pretend like everything was perfect the first time. It wasn't. But we iterated and we figured out what was the gaps that we needed to solve for. So we'll get into the mistakes we made.

I'll touch on three that highlight it so you don't have to repeat them. And then I will get into a little bit of our tech stack and our approach to be able to solve for. So three mistakes. First, we tried to build a do everything agent.

It was kind of like an AI CSM almost. So it was one generic bot that tried to answer any question about the account. It was a really fast prototype. The demo of it looked fantastic.

So our execs love the demo. But the problem was the outputs couldn't be validated. So our CSMs had no way to know if what they're reading was accurate. And you can't build trust on a foundation that you can't check.

So that was the first kind of mistake and or problem that we had to overcome with ours. Second is we actually kind of kept it within a smaller group. So it was our CS ops. It was IT and CS leadership.

So we didn't engage marketing initially. We didn't engage support or sales. So the agents had a partial view on the customer. It didn't have a full picture on what the customers needed.

And third, and the one that really kind of set us back, was we actually merged and tried unifying our conversational data with our telemetry data in terms of the output. So conversational agents are interpreting human signals. So things like what did our champion say on the last call? Or what's the sentiment on this thread?

Or who's engaged? Whereas telemetry is going to be usage went up by 15%. They deployed three new connectors. Capacity's at 80%.

It's data. And so a CSM couldn't tell fact from inference. Which is a problem and eroded trust with our initial model for it. So we set out to kind of rebuild ours to be able to earn back that trust.

And the first thing I'm going to touch on is the architecture of how we went about it. So I'm going to go bottom up. At the bottom is our data foundation for it. So data warehouse, product analytics, CRM, revenue intelligence, marketing automation, intent data, contact intelligence.

All of that is aggregated into our data warehouse. So every source that has data is aggregated into one place for us to be able to pull. So that is our data layer. On top of that data layer is an intelligence layer.

So growth AI is using propensity signals and scoring for growth and signal detection. Cohort's AI is pure benchmarking for us. And sentiment AI for conversation signals and relationship signals. On top of that, this is where I feel like we have a little bit of a cheat code.

Being a company that builds with AI is we used our own product here. So we used our Boomi agents as an orchestration layer. So they are kind of domain specific agents that can synthesize our signals. They can route the work.

They can draft actions. So this really connects the intelligence layer with what CSMs actually see and do. So it's also kind of where-- before we talked about the signal influence of saying, hey, we have these telemetry signals as well as these conversational signals. That layer is also where we filtered those out and tagged them appropriately for it.

And at the top, we have our CSM facing opportunities. They should be familiar because Emily just touched on them. So I won't get into them. One thing I will say is everything about our approach to this was the signal flows up.

So in the top right corner, really small. Data stays in the data layer. Intelligence signal stays in that layer. And then we filter up into what ends up being our agent layer.

And our agent layer is what the CSM workflow should be for us. So that is designed around what is the ideal state and ideal workflow for our CSM top right. So we had to make pivots. So we talked about the problems.

I don't know if it's like OCD, but I do things in threes. So three problems, three pivots in terms of phases. And we didn't do things all at once. They were actual phases.

And the unique thing about that, because there's so much trust issues related to AI and rolling out things related to AI, you have to roll things out in phases to earn trust. For each phase, you earn the right to do something else once you get a trusted outcome for it. So our phase one was switching from that monolithic agent into microagents. So instead of having to do everything, bot agent, whatever you want it to be, we went small.

Each agent did exactly one thing. So it was like meeting prep. It was account triage, renewal timeline. And it drew from one or two verified sources.

So instead of saying, hey, we're going to pull from eight different sources to be able to understand that, what are the one or two sources we care about for that agent? And the beauty of it is our CSMs could fact check it in 60 seconds. So they understood where the data came from. They had links to the underlying data if they wanted to check it and verify it.

So once we got that phase one in, and our CSMs started relying on the microagents or Monday morning prep or whatever, they had permission or they gave us permission to move forward. They gave us permission to go into our phase two for it. And phase two was separating by signal type. So we have data agents that are producing numbers and trends.

And you have your conversation or contextual agents that are producing label inferences for us. A CSM now also always knows where the data came from and what type of data it is servicing for them. Is it data or is it a signal that's coming from revenue intelligence platform? And our phase three.

Phase three was interesting because it didn't happen because of something we did. It happened because of what other people were noticing. So phase three happened because sales saw what CS was getting in our pre-call prep. And they were like, can I get that?

Can I get that context right? The same thing with marketing, who was like, hey, we see in the engagement signals, can we use that for our targeting? And so we didn't really design phase three. It was kind of the pull from other groups that helped design it for us.

The main decision point we had to make in that phase was do we keep the CS agents that were built as CS tools? Or do we reframe them and start to think about them as like a shared go-to-market infrastructure? And we did the latter. So again, I kind of touched on-- it wasn't the plan for that last part about having shared go-to-market infrastructure, shared agents that are using across our entire go-to-market teams.

But when we talk about the problem that we laid out, 5,000 customers, 1 to 70 scale without adding people to it or burning budget. There's constraints for it. It's almost a math problem that we were trying to solve initially. But what we ended up with that we didn't design for was because of the work we did around our agents, CS almost became like the operating system for our full go-to-market engine.

They were kind of at the middle, at the center for it. So now sales reps are getting their adoption handoff briefs. And they have more context than maybe even our CS teams had 12 months ago. And then leadership and ops in terms of a data standpoint.

We have aggregated data for portfolio views, for renewal forecasts, for roadmap signals. And that all goes into our transformation scorecard that I'm going to have a room touch on here in a second. So again, if there's one thing to kind of take from this, it's like you can build trust from your process is kind of the start of it. So as you're working through building agents, trusted output for it of the data, where it comes from, what it means.

And that will help build the scale and will help kind of drive the flywheel. At least that's what it did for us. So I will turn it over-- I think Arun, you're covering this one right? Yes, thanks, Bant.

Look, every dime we get from our CFO has a question mark attached to it. Hey, what am I going to get into it on? So we obviously, before designing everything, we solved for what are we solving for and how will we measure it? So we obviously had a lot of measurements put in place.

But three key ones I want to talk about is experience. So our whole goal was how do we scale and how do we really not give generic content to our customers, super prescriptive, super contextual, super relevant context to our customers based on peer insights. Our platform has a lot of information about how customers have built solutions on our platform. So we take that solution, we anonymize it, we provide to other customers in terms of, hey, this is-- you're not the first person to do this.

Someone else has already done this. So we tried to kind of scale our playbooks in that way to drive adoption. So in-product platform adoption was a metric we tried to track. How do we change the rate of adoption for our customers?

And we changed the number of engagements from 50% to 80%. Now we are kind of covering every customer at least once. And we tracked a lot of things like email opens, click opens, links opens, all of that stuff. But overall, we were tracking dark accounts.

Customers were not responding to us or not talking to us. And that became from 50% to 80%. Efficiency was a big, big metric. We had a lot of challenges and a lot of questions about, hey, how are you measuring the spend that we are doing in these agents and whatnot?

What is it driving? So ultimately, what it did was we do calculate cost-to-serve. We have a baseline for cost-to-serve that we did in 2024, 2025, 2026, and 2027. We're doing it again.

And there are multiple ways we are measuring it. But overall, we would decrease our cost-to-serve by 20% point. And we use that money to fund hiring additional CSMs to scale our growth. And this is a verified metric.

We had a lot of challenges in terms of, hey, have the strategic finance team review this metric and whatnot. But ultimately, our CFO agreed. Yes, that is the cost-to-serve baseline decrease that we've been able to achieve through the rollout of all these AI agents in the last one year. And lastly-- and I put an X here because no one allowed me to share the number.

But I do say something. And for the genius in the room, we can do the reverse math. We saved $20 million in churn last year to all the stuff-- human-led transformation, aligning incentives, doing-- putting out space to do heroics and whatnot. But it's a big number.

It's a big number because it's just not a big number in terms of churn saving. But it's a big number in terms of, hey, number we have to-- the money we have to also burn to reacquire that dollar and the impact of valuation. So that's our narrative internally in Boomi. Like, hey, churn matters because it's just not the dollar that we lose.

But in terms of dollar, we have to spend to reacquire and the valuation loss. And when you say stuff like that, everyone cares. And I know this is a question I have got right now in terms of, how do you get sales to be part of this? Because we use our partners in revenue operations and have a CRO that, hey, this is everyone's problem.

And everyone has to care about it. Thank you. All right, so we-- [APPLAUSE] So we have questions. I'll read them aloud, and then people can jump in.

There's some I wanted to do too. So first one, what system are you using to send monthly adoption insight emails? What's the process to QA and respond? Yeah, so it's a great question.

I think, Matt, you answered this in your own way. So we use our own stack. We generate adoption email for every customer. It gets into a customer success person's inbox.

And as Matt said, we have 60-second rule for them to verify. So it's all built in our own stack. We use MCP for all the systems that provide MCP. We have a Snowflake.

Snowflake have an MCP. Our agent queries Snowflake MCP. It gets an email. We use a boomy process to send an email to our CSM, which they can verify and send it out to a customer.

But we do not automate this intentionally. We don't want-- because all these emails have numbers. And we don't want ever a customer to see a number and freak out. So we have CSMs who just stack check it and send it out.

So for the question around building these agents out in Clod, we're using Gong, Clod, MCP. So similar to what Arun and us, we touched on, the agents are built in our platform. We are not using Gong-- or we're not using Clod for anything for it right now. And we're not using Gong's MCP.

Gong's APIs are written into-- all Gong data is written into our data warehouse. So our data warehouse is the store for all of our data. So it is the one source of truth for all data from any system related to boomy, even including our platform usage data and what customers are doing in our platform. And can I draw down on that one?

That's architecture design choice. Because right now, every platform is saying they have an MCP. And yes, you can use an MCP. And we're not trying to build an agent which has context from 10 different systems and trying to do this.

It's very expensive. It costs a lot. So all of our data is stored in Snowflake. All of our conversation data, Snowflake Salesforce data, product telemetry data, all of it is in Snowflake.

So the only MCP we use is the Snowflake MCP. So it helps us govern the responses, govern the cost, and fact check all the information. And we obviously do build a lot of context layer in Cortex, which is Snowflake's MCP. How do you assess the efficacy of the agents, both in outcomes and CSM sentiment?

If you ask our CFO, she would say the only matrix she tracks is NRR improvement. Everything else is vanity. I don't agree with it. But we track productivity.

And we have a matrix in terms of click rates and how many customers are consuming our content. When they were not talking to us for about three months and whatnot, we tracked productivity. Productivity is a big metric for us because we, as a company-- I think the math was we acquire five customers every day. Not anymore, maybe.

But we have a lot of new customers that come into our business. And we're always trying to, as Matt said, we're trying to solve a mathematical problem. If we have 100 more customers who are spending this much, how many people do we have to staff in our business? That is the metric that we are tracking, like a cost to serve and productivity improvement.

Thanks, Dr. How do people provide feedback on the model if fact-checking services issues with recommendations? Yeah, so we have an AICOA team. We have an internal AICOA team.

And they report it to our CFO as well, which is great. So they're very, very gracious in terms of partnering with us. We do weekly meeting on Monday in terms of, hey, all the agents we have. Let's look at who is using it.

All the dashboards and all the feedback. And we're very gracious to have a full engineering team who's dedicated to not just building agents, but also managing existing agents in terms of feedback and whatnot. Yeah, and we have a Slack channel that has our CSMs in there, our AI strategy and AI development team. So it's very rapid triage of our CSMs will provide feedback.

And there's that real-time review of that. And based on whether or not it's like, hey, something urgent that we have to fix immediately, or something that is like an enhancement request, then on a weekly basis, we're reviewing and prioritizing what should go into the next version and iteration of the agent. Yeah. And also, one more thing.

A fundamental problem that we pivoted from was we were actually building an agent which was doing both conversational analytics and telemetry analytics. And we failed in that because the way agents work, they work differently. If you ask agent to look at a spreadsheet and verify all the data versus look at a call transcript and provide a recommendation, they're built differently. So when we were able to do this separate decision out, our agents that work on our data is up to 100% accurate.

And then it took some time to get to that accuracy. But our data in Snowflake is accurate because we run a medallion architecture in our Snowflake instance. And so we know this data is accurate. And the agents that work on the data are also accurate in that sense.

And in terms of how long the overall rollout took, including the mistakes in phase three. 18-ish months, right? Yeah. So we have been on this journey for the last 18 months.

A good part of this was figuring out tooling, tooling strategy, who's going to fund this initiatives, the metric we will track, and getting a business case build, getting the binds from all the leaders, et cetera, was took about three to four months. The rollout was about six or seven months. And the iteration took about a year. So cost to serve target.

Are you able to share your cost to serve target and what's included? Unfortunately, I cannot share my cost to serve target. But we have a lot of opinions in terms of what our cost to serve should be, in terms of support cost, staffing cost, TNE, expenses. Should we do it against ARR or ACV or TCV?

We have a lot of calculations. But we present a view which is across all these angles. But across all these angles, we show meaningful decline in terms of cost to serve. But unfortunately, I cannot share the numbers here.

But I can share the calculations if everyone is interested. And maybe for Emily, how did you decide what types of highlights to include in the adoption set in months? So this was from a lot of different sources. We talked to our CSMs about what are your customers asking about?

What are the insights that you want to be able to come proactively with your customers to the table with as well? I think we also looked at just common questions that were coming up across our call recordings. And then there's always that kind of question around, we wanted to be able to have a stronger point of view on ROI and be able to put that in front of our customers. So it was a mix.

It was like, what are we hearing from our customers? What are we hearing from our CSMs? And then we as leaders, what did we want to make sure that we were putting in front of our customers? And the question around commercial leadership members and conceptualizing, developing, bumping, and rolling out agents versus the roles of the engineering team.

I will say half the ideas come from his head. No, it is a partnership between all of our teams. So our leaders are involved with it as well as IT to talk about the approach for it and what we're trying to solve. Ultimately, we start with the outcome.

So we ignore the technology. And we're like, what outcome are we trying to drive from this? And what does good look like? How are we going to track it from a KPI?

And we don't have technology conversations around it or data conversations till we get that. And then we start layering in the data. Yeah, and one more thing I would add, we're also lucky to have a dedicated VP of AI initiatives in our company. And when we started, it was like, hey, throw us a problem and we'll try to solve it.

But now even that competency is matured now. Anyone can say, hey, I want an agent built for my org or my function. And they have to submit a business case justification. What are you trying to optimize for?

How are you going to track it? What is the measure of success? How pertinent is the problem? How pervasive is the problem, et cetera?

So we also matured in that front in terms of, hey, building a very solid ROI-driven use case for everything that we're asking our AI COE to kind of care about. And that goes across every department. Everyone has to go through the same process. I think we're on to the last one.

How do you actually drive change across the organization? Do you counter resistance from leadership? How do we drive change across-- I think change-- Think about incentives. Changing incentive drives change.

One of my biggest learning from the last two years in terms of, hey, if you drive a change, not incentivizing someone, the change is hard. I'll give an example. We do our president's club. We said the CSM who drives the best use of AI gets to go to the club.

And everyone started to use AI. Real story. And look-- Yeah. And we're doing it again this year.

We do so much incentives around using AI, being more productive, saving shown. We do a lot of this thing. And that's kind of an easy change in our business, in a way. And I think you can offset some of the resistance for leadership by leading with the problem and the ROI.

So if you say, here's what we expect to get in terms of time savings back by this as it scales throughout the business, when you go to them and they're like, hey, we're going to spend x amount of time, but this is going to be a 6x save on the business. That's a very easy math for them to agree to. Any other questions? Awesome.

Thank you to the Boomi team.