Salesforce, Inc. (CRM) Earnings Call Transcript & Summary

All right. Good afternoon, everybody. Good morning for some. It's Keith Bachman here from BMO. We're sorry, we're a touch late. We've run over, I gather in terms of our virtual conference. But for me, this is my last one and thrilled to have Salesforce on with us. There's a few from IR, but we're just going to go to Susan and a way to start this is we're going to ask Susan to give her background before we launch into questions since as Alex has told me, this is Susan's one of her first engagements with the investor community. So Susan, why don't you tell us a little about yourself?

Yes. Thanks, and nice to meet you and happy to be here today. I'm an SVP on Salesforce's product -- in Salesforce's product organization on the Agentforce product team. And I've been in Salesforce for about 14, 15 years at this point. And along that pathway, I've enjoyed what I coined as the best job in Salesforce, which has been sitting at the edge of a lot of the innovation that we've been doing with AI and data. Prior to this, I had a heavy hand in a lot of our Einstein and machine learning products. But for the last 3 or so years, been part of the foundational team with all things generative AI and agentic Agentforce technology.

Okay. Perfect. I'm going to start a little bit differently in that a lot of investors ask us what the difference -- how do we get here? And what I mean by that is how did we get to this thing called Agentforce. We used to talk all about Einstein. How did we get here? How was the evolution? How has that unfolded?

It's a great question for me. Thanks for asking it. I mean, obviously, back in the 2016 era, about a decade ago, there was a convergence of data and processing power that made sort of a big step change in machine learning possible. And in those days, as you commented on the Einstein brand, we had a lot of both out-of-the-box capabilities for predictive things like lead scoring, opportunity scoring, classification, those type of more traditional machine learning things, which really defined our time about a decade ago. Now obviously, with moving to the current day and age, the capacity of the machine learning models crashed on the world very, very aggressively about 2.5 years ago in terms of not just the impact in the consumer marketplace for the ways we all enjoy it in our personal lives, but for in our personal lives, but for everyone managing a large enterprise in terms of how does generative AI impact not just their technical stacks and the user experiences they have for their employees and customers, but business models as well. And so the original working models with LLMs from that time frame was a lot around prompt engineering and leveraging generative technologies, to summarize things and to generate content. And the agentic shift takes us into a new category of things where we can permit and allow these applications to take on more autonomous experiences with the controls and the guardrails. And I would say all of the tooling that you need to, as an enterprise, which is much different from a consumer experience, bring to the foreground to put these things into place around workflow, around productivity, super cycles of your employees and new customer experiences externally. So for us, the step function change was releasing Agentforce at Dreamforce last year. And a lot of that step function was brought by our builder tools itself, resident in Agentforce.

How do you think -- and this is more of a market question, but you've had an interesting seat to observe this. In my simple brain, there's causal AI and GenAI, and those might not be the right nomenclature. But how do you think about those causal models and still the effectiveness and necessity of those versus a probabilistic model, which I think about is GenAI. How do these 2 worlds cooperate, exist, compete? How does it -- and again, this is not a Salesforce comment. This is more of a take a step.

We do not comment to the market. What I would say, like in the early days with generative AI, what I mean early days, like we're talking like 2.5 years ago, like we live [indiscernible] right now in this tech space. But the initial like -- and this is kind of winding time back for me a bit, like a lot of the questions were like, can we solve everything with generative AI and people would start like spitballing and use case ideation and you'd be, that is a great use case. You know what it is, it's predictive, like that's a regressionable. And so kind of the 2 observations is, one, it was a whole new level of creativity and permission to really think about bringing technology in because it was such an important technology moment. So that permission step like created all this ideation. But the way the 2 work together, like a classic example would be a machine learning model or something that's doing a predictive outcome will tell you the order of operation of what you need to focus on. The generative like use case will tell you how to do it and bring productivity to the foreground. I'll give a classic sort of Salesforce sales example. If I've got predictive signal about a customer that might buy or a customer that might churn, that's going to move them to the top of the list of engagement. But generative AI might bring additional capabilities in terms of creating customer briefs before a call or taking...

How to serve those customers?

Yes, taking signal from data and from all the experience information that's collected and crafting like using autonomous processes to create that engagement in a more powerful way. So it's like who do I call and why? And what do I say and I do when I get there, sort of like a really nice peanut butter and chocolate example.

Yes. And so both worlds will live on.

Oh, yes.

Obviously, we're still super early in GenAI, but the causal necessity and advances of causal AI will continue.

Yes.

Okay. Let's turn. Thank you for indulging me on sort of the how did we get here. But let's go on to the more pointed questions associated with Salesforce. Everybody's got AI. Why is sales -- why is it important to Salesforce? And really, the nature of the question is how is Agentforce differentiated?

Yes. I think it's a great question, and you're right in acknowledging it's such a big technology moment. We recognize there are a lot of options. And we have what we think a tremendously exciting set of capabilities. And the way I would describe them are a combination of things. Like we're gifted with this deep history with sales, service, marketing, commerce, analytics. So we have this amazing suite of applications that all have humans and all have processes and all have automation attached to them. And so it is a really nice opportunity to take those applications and modernize them with agentic capabilities. So this exploiting the app layer is super nice. The second thing I would say is that we've been, for a number of years, really accelerating our capabilities with data. Now obviously, with Salesforce data, we always have a special relationship with it because it's in the platform, it's permission aware, it's like workflowed and all that stuff. But the work we've been doing on the Data Cloud has really opened up the aperture and the technical ways that we can engage with data, which is so important for grounding these agentic experiences. So data, the application and then within the agent layer, we've been really busy doing 2 things. One, we are -- let's just write what I call it like the joint opportunity and obligation of we have all this deep knowledge of the personas that sit in Salesforce products all day long, like our researchers know that. Our PMs think about that morning, noon and night. And so we can create all these accelerated agentic experiences that, one, take the think time out of wondering what to do; and two, accelerate the time to value because you can configure the last mile versus starting with that clean sheet of paper. Now the second thing we've done in addition to these out-of-the-box applications is for the last 3 years, we've been building this deeply embedded enterprise-grade agentic AI platform in Salesforce because, as you know, Salesforce customers love our applications, but they also like making them their own, and that involves configuring and extending them. The same is true with our AI in terms of all of this tooling that we have for people to customize it, all this observability. So when you move it into production, you have real-time line of sight to what's going on in guardrails. -- and all of this enterprise-grade technology. So we would just call that like apps, data, metadata and agents as sort of a framework. And -- but maybe taking a more specific look because I'm usually discussing things at a deeper level. And what I would say just kind of giving the pace of AI, the answer to this question will probably change for you in another 6 months, just as it was different 6 months ago. But the things that make it very unique for our marketplace right now is sort of the following categories. And I'll call the first one surface area. And surface area, meaning -- it could be meaning a Salesforce user experience, someone who's logged into CRM, someone who's logged into Slack, someone that's on our Experience Cloud, but you have a human in Salesforce that is going to be superpowered and supercharged with these agentic experiences. That is a huge advantage. And while it is -- from my perspective, as one of the AI practitioners, a lot about AI, it also is about design and behavior. And so it is a really unique opportunity to revisit all those experiences and really next level them in all sorts of ways. So that, plus the fact that we've got this like super cool platform makes that great. I think the second thing is we have been very focused on some core principles around openness. And openness has come through our AI story in terms of openness to the ways we ground and work with data, openness in terms of selection of LLMs. We've incorporated that in our product over the last couple of years. And now with the latest open category of conversation, all these MCP and A2A frameworks, and so this openness provides a future-proofing state for our customers. And just given the rapid state of progression in this space, honestly, what people often think is unique and game-changing on day 1, by month 3, could already fast becoming a commodity. And so this openness allows us to really bring this future-proofing mindset to architecture choices that people are making in the enterprise. Number three, it's AI, and so you have to have great AI. And we have a number of things that I would say put us in that category. The trust models that we like really initiated in the marketplace in terms of things like your data isn't stored with these foundational models, we'll mask all that sensitive data, like all that sort of data safety. But trust is also about accuracy. And so the things that we pulled forward in our product around including citation. So you have, as a user, line of sight to what that source material that GenAI is using. The work that we've done in our reasoning engine and the work that we do in our RAG metadata pipelines, all of these things are around accuracy. So there's a whole bunch of things that make it very accurate and very trustful. And then when I look at sort of the big chapters across the last 3 years, 2023 was the year of like, does this change my business model in the march of the consultants in the boardroom? 2024 was the year of POCs moving out of the lab and into production in small bits. It was also the year where Anthropic and Gemini and others caught up with OpenAI. And the year of 2025 is around full-scale production, measurement and observability. And so we've been bringing a lot of our advanced research techniques into these observability models where not only are we using AI to generate the creation of these AI agents, we're using AI to create the test harnesses to evaluate them before they go into production. We're using AI to improve instructions because as we know, this is an emerging industry and people need help in those learnings. So we bake our learnings into the product. And we're using AI Eval models to understand if these agents are adhering to the policies and the instructions and the actions we're gifting them with. So this kind of production mindset has been very, very powerful for deployment. And then finally, a long-standing line I've had like since calling on financial institutions back in the '80s is everything is possible with time, money and code. And it's always spike for folks. And so the skill set that we bring to the foreground is very unique in terms of leveraging this sort of trailblazer mindset as well as having these command line interfaces for the community that enjoys that, all of this being a way to go fast with products that people have already made investments in a.k.a, this huge Salesforce suite with some of the best AI and techniques around enterprise suitability. Long answer, sorry.

Okay. A lot to chew on there. We could -- there's a lot to go on. But let me -- you said one thing that sort of piqued my interest. You said it completely -- the AI world could completely change in the next 6 months, so I'll say in the next year. But what do you feel like, a, you need to get right from where you are today from a technology platform perspective? And b, what's the most -- what's the greatest source of friction on customers not adopting that are Salesforce customers right now?

Yes. I think some of the things I just said around the pillars are the things we're working on right now, like observability is really important. These are generative capabilities and many organizations are still feeling their way through trust with LLMs. And so you put them in front of your employees, you put them in front of your customers, you sort of want this. So we've been focusing on that for quite some time. The second part of your question, it was about -- which is you call it like friction or barriers? Is that how you phrased it?

Yes, why are -- you have a huge installed base of customers. And candidly, a fraction have adopted or generating ARR for you guys. And so most customers have it. What's the -- what do you find as a common source of friction about why folks aren't adopting?

Well, I would put -- I'll just address it in terms of like where we see -- I wouldn't call it friction, but ways we can accelerate people's understanding of it because we -- like we are very, very busy servicing the needs of customers who want to engage with us, whether it's things like use case ideation sessions and workshops to train people and the trails that we put on Salesforce to help educate people at scale about both the possibilities and the actual tools. So there's a ton of activity there. And Alex can also reinforce some of the actual traction we're seeing with ARR and also repeat revenue. So there is massive momentum there. What I would say around like friction, and I think that's even too heavy a word, I'll give you 2 examples. Like when I think of categorizing use cases, I put it right now into buckets of the productivity super cycles for our employees, I would put the next category in terms of experiences that we put in the pathway of our customers, like these external autonomous customer experiences. Now if we take those 2 categories, like we'll start with the customer-facing use case. What we've seen at scale is -- and this is across like many different industries, retail, consumer goods, regulated industries and financial services. So there's been sort of no holdout. It's been very universal. It's very easy for people to conceive of use cases that face customers that do things like answer questions, deflect calls, if it's a call center that feels they can create a better user experience in a modern, adaptive, conversational way, like that's both reduction in cost to serve in terms of the technology to do that and a better customer channel. So like answering questions, and as a category, I would say, reading from a database, meaning where is my order or where is my [indiscernible] or where is my claim or where is my shipment? Those sort of like, tell me what the status of this is without me waiting in a long queue and fighting with an IVR system. Or the next category for that customer-facing experience might be, I need to initiate a process. I want to initiate a service request. I want to initiate a claim. I want to initiate a beneficiary on my account. So like those things come really naturally and easy because they know the processes that they're already serving on their call centers at scale. There, it's measured like crazy. So that is usually pretty -- like that can accelerate really quickly because the think time is compressed because they know where the friction already is in their business by servicing it with measurement. On the sales side of things, it -- people need more help in terms of where do I start and why? I have all these processes in my organization that may or may not be completely understood by me, especially if I have a large like sales team. And so helping people understand their business and their business processes and where AI automation and where the design of AI that is supportive of human can take friction out of the process might take some time for folks. So we've been responding by just getting in the trenches with our customers and helping identify this stuff. But that's where I wouldn't call it friction, but it's an opportunity to think deeply not just about jamming like some AI experience, but where do I have friction and how are my humans compensating about it? And how can I inject AI there? So I won't call it friction, but I would just call it, it's -- it might take a little bit more time to get that road map of everything you want to do and then put it in that 2x2 grid of high impact, like low risk kind of where do I start thing.

Right. Right, right, right.

Yes, I don't know if makes sense but that's sort of like the...

It does.

Yes.

It does. Let me ask about -- go back to something you said at the outset, and I'll use slightly different terms, but customers need to adopt the data cloud in order to be successful with your agents. And maybe help us a little bit with the why that is the case and I think -- data structures, but also as a technologist, a lot of customers already have Databricks or Snowflake, and you're sort of asking to stand up, for lack of a better word, another, Datalake, which is nobody really wants to do that. And so I just want to hear a little bit about the Data Cloud and how it's important to this process.

Sure. Yes. I mean there's -- I could talk for hours about this one, too. So what I would first say a little bit tongue in cheek, I wish we had named the Data Cloud, the activation data substrate. Like if I had invented sushi, I would call it like called cold dead fish on...

Cold dead fish.

Instead of like really kind of pragmatic name for it. So of course, people have made these investments in Snowflake and Databricks and all these lakes. And the answer is, thank you. like because what we're here to do is leverage and activate that data, not replicate it and rematerialize. That's one of the...

Exactly.

Of Data Cloud. And I think if there is friction, like to your question a moment ago, it might be truly people understanding that, that you don't have to copy data into our environment. We leverage it in a very modern way in terms of just treating it as if it was a Salesforce table. So there's -- so like that's sort of one thing I would say. It's not a separate data -- we call it Data Cloud. It doesn't mean bring your data to our cloud. It means let us help you activate your data in all sorts of creative ways across Salesforce. Now as a product exec at Salesforce, like I see Data Cloud in many ways. I see it as the original CDP in terms of a really modern [indiscernible] to get all this -- that's a category and people buy it, and it's awesome and it's leading. I also see it as a way to bring additional data into the Salesforce ecosystem in a very modern way, but not replicating and all that stuff. And that's terrific because you've got humans and processes all through Salesforce that can be leveraged by this in proactive and reactive ways. So new data types that historically were not deeply resonant in Salesforce. And all of our product teams build on Data Cloud as if it is a platform because it is a platform. So with the agentic capabilities, with the Agentforce things, with all of our GenAI like everyone has this little moniker of like AI needs data. Yes, it does, but not in the traditional sense of building models because we -- most people are using the pretrained models. So we're not using it to build models, but we're using it to ground and inform. And when you're interacting with an LLM, the better instructions you can give it, grounded with customer data, the more accurate it is. So like I was talking to a bank the other day, and he's like, I really now finally get Data Cloud. And I have these amazing user experiences from my advisers because it's not just LLM giving me a summation of notes, it is an LLM that fully understands my customer because I have FactSet data, I have transaction data, I have banker notes. I have position information. I can't get that without that. And so grounding with this data is really important. And because like this awesome modern platform, we also put all our log files there. That's where our e-mails go for observability. So like we're leveraging Data Cloud like in ways that it should be, but it's not to bring your data to our cloud. It's let us help you really next level the hard investments you already made in building out those Snowflake and Databricks environments.

And to add to Susan's point because this goes back to your prior question, Keith, around the level of Agentforce adoption, what we're seeing from a customer momentum standpoint has been unprecedented in terms of interest in Agentforce in terms of customers choosing Salesforce to start their adjunctive journey. Realistically, we know that takes time, and it goes back to your question around data and getting data in order. And so what's been encouraging for us is as we've seen customers choose Agentforce, and we mentioned 8,000 deals closed to date, they're realizing that it is a longer-data journey. And that's why when you look at some of the stats we've given in the last earnings call around surpassing $1 billion in ARR for Data Cloud and AI, a lot of that's still coming from Data Cloud. And it's with the lens of how do they harmonize the data on our platform? How do they bring in unstructured data, to Susan's point, where they previously weren't able to activate that data before, but now it becomes really key in the objected customer journey. And then how do they leverage tools like MuleSoft and eventually Informatica. So we are giving customers this unified data architecture. We're making it very simple for a customer to get all of their data in order with the lens of then you have your agents natively integrated on our platform, and you're easily able to tap into that data, get the value and activate that data within your customer journey. So that's what we're really excited about. And it's important to us as we go through this agentic journey with our customers to continue giving all of you key milestones in terms of what we're seeing from adoption but really giving you milestones into what we're seeing with customers when they eventually move from the POC experimentation phase that they're in now to a limited deployment to an eventual deployment at scale. And once we get that flywheel turning, that's when we really think you see Agentforce become more material in FY '27.

Okay.

Yes. I would really echo all those momentum stats. And one of the things like if I'm at a conference and people ask for guidance, like what do we do? I was like you got to start. Like that's the first thing, like start, commit and go because we are seeing like with the first use case, you have to figure out everything. You have to figure out your risk profile. You have to figure out how LLMs work. You have to figure out what data you have. You have to figure out user experience. You have to figure out all the boundaries. And then once you get that, you have this acceleration like platform. I'm working with so many customers right now that have launched like their 1, 2, 3, 4 first agents. And now they are creating what they call agent factories because either they're going use case by use case and functional area by functional area or they've got country 1 stood up and now they're going to go to countries like 2 through 56. So we are definitely seeing this scale, both in terms of variety of use case, acceleration of regions and things along those lines.

And so Susan and Alex, when you think about -- I use the word friction, which did not go over well, but let's say, discussion point. Yes, Alex would call me a source of friction. But when you think about your discussion with customers, is it understanding your economics, right? And so are customers still trying to understand how -- and the various scenarios underneath it or -- and/or how important are the discussions surrounding I got to pay more, how is this going to evolve?

Yes, it's both. Like it's understanding like what is this technology, what is your technology? How is your technology different because I got a ton of choice, like it's all that. And it's where do I start and why? And is that thing going to ROI and how is it going to impact my business? And I sort of categorize these different potential value points like it can be just categorically the productivity super cycle. The thing that used to take 9 minutes takes 4 minutes. The things that used to take an army of people is a smaller number of people that people are redeployed to hire like higher margin and higher profile activities, like so there's this whole productivity thing. For these customer-facing experiences, for many organizations, it's an opportunity to understand the cost to serve, but more importantly, better channels and better customer experiences that lead to increased loyalty, cross-sell and upsell, so they will bake those not just into a cost to serve return, but how is this impacting the growth of my business. Now we're working with some organizations that sort of take that even to the next level, like how does this digital labor change operating models for me in terms of ways that I just hadn't anticipated. Like -- and the classic example is like -- and I'll repeat it, I didn't invent it. You probably heard it a million times. When you first held your first iPhone in 2006, did you imagine Uber? Like so this whole thing is like people are now starting to imagine these new digital labor scenarios that just weren't possible before. So -- and I got some customer stories that I can tell there. And then as people move from these productivity super cycles for employees to customer experiences that are just next level, like their sort of -- the next sort of set of considerations is how do you have background agents doing the things that the humans are traditionally doing, which is sensing and responding to signal. The customer called. This thing happened. They incurred usage. They didn't incur usage. They opened a new account. They -- like whatever all these data signals are, the AI automation can start the whole process and pull humans in the loop in new ways possible. So that is -- it increased revenue, decreased cost and new ways of working are just category like what we're seeing everywhere.

Okay. Well, unfortunately, we have to move on a little bit, but we may come back to these. And the reason I say move on is we've had -- we had ServiceNow on yesterday, and they are talking a lot about moving into the front office. They refer to it as CRM. Part of it is their thesis is they have a horizontal layer for agents and agent orchestration. They have a little slice of what I'll call applications within the front office. But how do you think about your differentiation if we take AI Agentforce Data Cloud versus some of your competitors? And specifically, if there's anything you'd like to call out from a lack of a better word, a horizontal player like ServiceNow and how you think it gives you -- provides you with differentiation in the front office?

Yes. I come back to some of the things I said before, surface area where people live. context switching is terrible. Like humans are finite in our ability to concentrate. And so context switching is sort of the -- just not a great design technique, right? So having it in [indiscernible] work where you don't know you're using AI, like you're just doing your job and AI is supporting you it every way. The second thing I would say is this like in the dawn of generative AI, it was like write an e-mail for me and summarize this text. It is way beyond that now, right? And so the ability to have these things create an AI orchestrated plan, reason through what needs to do, be responsive to conditions changing, all of this like AI orchestration is just -- it sits on top of actions all day long. And Salesforce customers have deep investments in things like flow and workflow and actions. And then, of course, all their business processes that they've got armies of sales and service and marketing, both people and processes already there. So this kind of like surface area, actionability, the time and money and trouble takes to get there, the openness to data, the way we future-proof are all -- are really outstanding for customers in terms of things to think about for us.

Okay. Let me take a quick pause to see if investors want to jump in. I have a few more or Brad, if you want to jump in from my team also. I'm just going to take a 10-second pause. Okay. We will continue on then. I want to maybe, as we're heading down the home stretch, talk a little bit about how -- maybe not your direct area, but MuleSoft and Informatica, how this helps with -- because in my mind, Informatica is really an enabled nurture of the growth of Data Cloud. But why is -- let's take Informatica first. Why is it important to have Informatica to be part of Salesforce rather than just partner with them?

I'll start and then I'll pass to Alex. Like from the AI side, I think the excitement is palpable in terms of the ability to inject even more customer information into the ways that we embed these experiences. Clearly, that is going to unleash a whole lot of value. Not everyone has stuff nicely packed away in a modern Datalake. There is plenty of other applications that run organizations where Informatica is a big part of that mesh in that network. So that will be terrific. And then there are capabilities around lineage and governance that just kind of any good data stored or data pathway, whether it's just a straight data pathway or it's data pathways activated by AI will be very powerful. So I see it from my AI practice in terms of next leveling what we can do by grounding and knowing the sources of these data. But Alex, you probably see it from a larger M&A and Salesforce perspective as well.

Sure. So I agree with all the points you hit. We also think the element of being able to bring rich metadata from a number of different sources, whether that's on-prem, whether that's from the cloud at scale is an ability that Informatica brings to the table. And we think, as Susan mentioned, with Data Cloud and with a lot of customers cemented on our core applications, you have rich metadata tied directly to the customer. But as you think about deeper complexity in the agentic experience, you probably want to bring in metadata tied to product or tied to other assets. And Informatica has a very extensive data catalog with that understanding of different types of data sets where we are now creating almost an Asset 360, a Product 360 tied to a Customer 360 with the lens of you also bring in with Informatica very robust data governance policies. So agents have a permissioning set in place so they can read certain data sets, but they can write to other certain data sets. And so there's this element about data transparency, governance and understanding that we think is critical for why Informatica needs to come to the fold of our platform. We did have a successful partnership. But for us to be able to build out this unified data architecture and offer to our customers a complete integration offering, we felt like buying the asset was the right move. Our lens is this is going to unlock significant synergies, whether that's from the go-to-market side, the G&A side and the product side, where ultimately, we want to make it as simple as possible for customers to get their data in order on Data Cloud and on our platform. And now we think with MuleSoft, Informatica and Data Cloud and then you have course, Tableau and Slack from a visualization conversational layer element, we have this architecture that allows customers to do that data work and then have Agentforce, which is already natively integrated on our platform to be able to action and activate, as Susan mentioned, all of that data.

Okay. And Alex, just quickly because I want to ask one more of Susan. How much overlap is from a workflow perspective, not a customer perspective, is there between Informatica and Mule?

From a workflow perspective?

Like the common use cases, like how much is there common use case? Or do you know that number?

I don't know off the top of my head. There's likely some overlap, but we do think that there's differences between when a customer would leverage MuleSoft, let's say, app to app and when they're leveraging Informatica from an ETL, ELT standpoint and likely leveraging Informatica to bring in data at scale. So we do think that there are different use cases when you think about MuleSoft, when you think about Zero Copy and then when you think about Informatica.

Okay. Susan, we're going to end with you. And I want to just hear a little bit maybe about a customer example incorporating agentic AI and Data Cloud that you think is representative about where this is heading over the next 2 to 3 years? Any customer -- because you talk to a lot of customers who clearly have a sense, particularly on the technology side of the market opportunities. Anything you want to bring to life that should help investors understand where we're going?

Yes. Always happy to bring some customer examples to the foreground. I started speaking about one a bit ago, a wealth manager, where their journey with Agentforce definitely involved Data Cloud. And the -- as I mentioned earlier, what Data Cloud brings them in that is very specific customer information. Now when you're talking about GenAI and financial services, specifically, like especially when you get into areas of wealth and health, it's like you have to be careful about recommendations, right? But you want to show up like well informed, well prepared and to drive great conversations. So in their example, using the powers of our capabilities, they leverage all sorts of client data, whether it's generated by humans, whether it comes from their back-office transaction systems or third-party data. And all of that together allows them with the click of a button to get a really rich household summary of everything about that customer and all the ideas about the products and services that meet the risk profile, the stated financial needs and really sort of drives like the relationship in very positive ways. Now at the same time, they're also using the same sets of capabilities to help people do their jobs because there are all sorts of products and services to know. There are all sorts of escalation policies and procedures to know and things like that. Now I'll kind of pivot from that organization. They started with human in the loop with their employees. Now I'll poke in on another couple of sets of banks where their first agent was -- or their first generative experience was they would like to expand the markets that they're serving, but they don't have the human capital to do it at the moment. They want to go down market with white glove experience. And so the first thing they did was they set up a digital agent that explains everything about their commercial and -- their commercial and banking products, I will just say. And then from there, all sorts of ways to schedule and connect with a banker. Now they have this vision of being a sort of an AI-first bank in many ways. So at the same time, where they're standing up this user experience that faces a potential banking customer, they are also creating all these agents that take the friction out of the human population that are bank employees, things like agents that do sweeps, things like agents that do loan prepays. So they're like using first their human workforce to make sure they get everything right because then that goes as part of the digital and AI-first bank. So we have this idea of digital labor really at scale, not just call avoidance on a call center, but really activate a whole new category of work. Now also without using customer names, like I can give another example about digital labor. This is an organization that has, I'll just call it, a prescreening process. And of that prescreening process, potentially millions of people that might want to touch. So having a digital agent gives them the capacity to go at scale and truly operate 24/7, 365 and not really bound by 9 to 5 in human labor. Now what they're finding with this is that -- it's funny because we tell people digital labor and like Marc's on a podcast and says digital labor and then the customer says like, "Oh my God, it's like digital labor, like we were able to engage with customers outside the hours of operations and in language we choose. Now that was very cool for them, but the next thing they observed was that the fidelity that the digital agents have towards completing the process and executing was very, very high, much higher than they like anticipated like that was pretty good. And when they get to the end state of that process, like what their measurement of that process was is over -- like it used to be like 4:1, like 4 people on the top of the funnel, 1 at the bottom, a 2:1 ratio. So they've been receiving orders of magnitude improvement. And that's kind of a digital labor conversation. So taking people through these discussions about like humans empowered by AI, digital labor, new operating models, digital twins of what they do internally is just so exciting for organizations to imagine a future.

Okay. Perfect. I think we're going to have to leave it there because I feel we've run over by a couple of minutes, but we started a little bit late. Susan and team, thank you so much for joining us today. Super interesting. We could go on for much longer. We appreciate your time, and we wish you all the best. Many thanks again.

Thanks a lot, Keith.

Thanks, Keith.

Cheers.