Speaker 1: 0:22

Welcome to this new episode of TechSign Talks on Tour. We're at AgentForce World Tour in Amsterdam and we're talking to Susan Emerson, who is Senior Vice President of Product AI Data and something I believe.

Speaker 2: 0:38

Just AI, ai, just AI nowadays, yeah Well, I have a history at Salesforce, where I had also been part of our AI and analytics and machine learning products over the last decade, but now just generative AI and agent force Okay nice to meet you.

Speaker 1: 0:53

We're talking a lot today about AI agents and how you can implement them with agent force on the Salesforce platform, and it's a lot about the potential of what it can do for you. But I was wondering, since you're from the US, I I guess you've probably seen some customers that are a bit ahead already implemented agents and what it has brought to them. What are the?

Speaker 2: 1:18

learnings that are already there. I would say, yes, I'm from the US, but that geography doesn't define innovation pace at all. But what I can say about agent force and specifically your question around that geography doesn't define innovation pace at all, but what I can say about AgentForce and specifically your question around how are people using it and what are the advantages they're seeing, prior to making this product generally available, we had a pretty large set of customers who were testing it. So these are all like large, you know, fortune 500 and Fortune 1000s who had early access to it and worked with us as we were, you know, putting the final touches on it. And I'll stay within the examples that are publicly available or have been, you know, cleared for by those companies' press teams.

Speaker 2: 2:01

So one would be Wiley and if you're not familiar with Wiley, they are trade press and professional publications and they do universities as well, and their idea was that they have periods of intense activity around things like requests for text and return of texts, given universities all coming into session and a lot of these things are not high-value conversations from the perspective of maybe that's a high value as a judgment statement, but it's a lot of repeat like where's my stuff, where's my order, and these are things that can be automated.

Speaker 2: 2:45

And so they put Agent Force into play just right at the beginning of the academic season and what they saw was a pattern of customers being able to self-service those questions in terms of where's my stuff, where's my book, without having to go through calling a call center, going through an IBR tree and burning down someone's time, including their own. It's all self-service now with go through calling a call center and going through an IBR tree and, you know, burning down someone's time, including their own. It's all self-service now with agent force. And they saw something on the order of 40% of all of these types of routine things are now solved by these agent force applications, so that you know and if you kind of break that down to like, why does that matter? Reduce cost to serve, freeing up the human agents that have more complicated requests that they really need to throw them into, and then, from a customer perspective, a much more delightful experience because they get it resolved really quickly.

Speaker 1: 3:43

But that's like an alternative to a track and trace. You see it like that.

Speaker 2: 3:50

Yeah, and when you call up a help center and they have to look up your order and then look up another system and see where it is, and all of that just takes time, and now they can just do it autonomously and agentically with these AI applications.

Speaker 1: 4:05

So the biggest improvements you're seeing at the service side of things.

Speaker 2: 4:08

Well, that's where we first started. Our first sets of GI products are on the service side, and so the types of like maybe kind of taking a step back and categorizing the types of things people are doing. One would be FAQ answering. You know, even for some people, scrolling down to someone's website and finding the FAQ document is like painful, right. So what if you can use these autonomous experiences to ask and answer all those FAQ types questions in a really good experience where they get answers fast? So that's stuff that if you have an FAQ like we love pointing these agentic technologies at it.

Speaker 2: 4:47

Another one would be the where's my stuff, where's my order, and that can be everything from where's my textbook to where's my restaurant reservation, to where's the shipping container on the ocean, to where's the hazardous material in my supply chain, like all those type of where's my order, because, again, those types of questions to require, you know, access to that data in real time to answer those questions and then sort of maybe the next category of maybe stepping up complexity and value is when you're not just reaching into systems and retrieving an answer but you actually are doing things like updating things, and I'll give some examples of use cases that we see a lot of interest in in the financial services sector of doing things like updating beneficiary records and other kind of account maintenance things, or even requesting things like can you send me my tax documentation because I'm getting ready to file? So FAQs, where's my stuff? Reading from systems and then, with the right kind of controls, guardrails and protocol, being able to update those systems as well.

Speaker 1: 5:58

You also gave some demos where people can call into an agent by doing a voice call and it can talk back to you. Are there customers that already implemented that?

Speaker 2: 6:09

So the demo that we are showing, the Saks Fifth Avenue Sophie demo we acquired a little technology company right before Dreamforce and it's that technology that will be powering that voice digital experience. So we have that in pilot right now with the goal of having it in the market early next year. We work very widely already with telephony systems. We've got partners in that ecosystem space. A lot of large customers have investments in CCAS systems. So we're already doing that kind of work where we connect to CCaaS systems and we're using those digital transcripts in interesting ways already. So this is yet another digital channel that we add to the ecosystem early next year.

Speaker 1: 6:57

Yeah, I believe with Service Cloud you have a connection with Amazon Connect, for example. Amazon Connect.

Speaker 2: 7:01

Genesis, like all those types of things, and you know where we're doing. Things like getting those digital transcripts and then using them to help proactively answer questions for agents, or using them for things like wrapping up cases, those types of Einstein for service types use cases that we've had in market for almost two years now.

Speaker 1: 7:20

And that new technology will be a major upgrade to that. When it can talk back, I would call it a major new feature.

Speaker 2: 7:27

Upgrade to me has like a lot of loaded implication behind it, but it would be a new channel that people can activate, if they like that digital channel for the customers in their marketplace.

Speaker 1: 7:38

And can you say something about the pilot you're running now with it?

Speaker 2: 7:42

Know that it's just beginning, yeah.

Speaker 1: 7:45

Okay, so we talked a bit about service, about the chatbot experience. It's becoming a lot better. I think a lot of listeners might think a chatbot, please give me a human agent now, because that's the experience, how it was. How are you changing that behavior? Because I know a lot of people when they get presented a chatbot, they just want to circumvent it as fast as possible.

Speaker 2: 8:12

And the reason I mean, yeah, right, and one of the things that and this is like a personal story like now that we have this technology and I know how it works and I see how quickly it is to build anytime as a consumer, I have an old school chatbot experience Like I can barely contain myself, right, because you know it can be so much better. It doesn't have to be limited to those QA pairs and it can be conversational and multi-turn and all over the place. So personally, I get a little like I see the potential for innovation and I'm like man, this company has to go faster.

Speaker 1: 8:51

But we need to change the behavior of people interacting with chatbots because they had this bad experience over the last couple of years. So the first reaction is to get rid of it or get a human agent, or get someone else or a chat or whatever.

Speaker 2: 9:07

Maybe organizations will use that as an opportunity to brand it in some way as a new experience, but it is radically different and it's so much better.

Speaker 1: 9:17

Is that your advice To brand it differently or make it look differently, or do you have any experience in that?

Speaker 2: 9:24

I wouldn't personally have any experience in that, but what you're suggesting is the old way of working is you go to the same corner of the application and that same little button is there. You're suggesting that you'd want the window like, meaning not the actual technical window, but you want the window dressing to suggest it's a new experience what you're suggesting? Is we're playing off of that Okay.

Speaker 1: 9:50

And besides the service industry, do you see other big industries that can benefit from AI agents in the near future?

Speaker 2: 9:59

Well, I mean you said the service industry, and the examples we've been using so far have been more customer service, but in multiple industries right.

Speaker 2: 10:07

Like where's my textbook, where's my shipping container, where's my order, where's my stuff, all that kind of stuff.

Speaker 2: 10:13

But, if we like, move to other functional areas, for example the sales side, some of the things that we're doing there with the autonomous agents.

Speaker 2: 10:23

There's two things that are just coming out of their last phases of pilot before they go officially full GA, and one is something we call the agent force, sdr or sales development rep.

Speaker 2: 10:41

The agent force, sdr or sales development rep and what this is is the ability to take content about a product or service or policy or process or whatever that company's product is, and expose the AI agent to it so it learns everything about said product and service. And then, via an email channel, this agent can engage a consumer with, like a QA, dynamic experience as there might be, you know, moving towards a purchase and that's all autonomous, with all the guardrails and things that you choose to put around it. And then, as it gets to the end of that journey, when they're ready to speak to the human and the sales rep, to then package up everything like these are the series of conversations we've had. This seems to be the stated need for the product. I've set up a meeting for you and pass it off to the sales rep. So they have, like you know, a really good um experience for the customer, who's getting you know, uh, acclimated to a product and learning about it, like straight through to the sales rep that will prosecute it.

Speaker 1: 11:40

You get a way warmer lead, doesn't it A?

Speaker 2: 11:42

way, warmer, lead, more productive, and it's been a really long time where consumers educate themselves before they talk to a customer. So this is just a different way of doing it. I mean, you can still scour a website and read PDFs and read articles, but to have the human having this personal agent that's answering the question. They have not the content that they are served in a website, so kind of a nice experience from that perspective as well. So that's one thing, and then the other one is we've been calling it agent force coach, and the way I describe this is a little bit practice.

Speaker 2: 12:21

Your pitch, you know you've got new products and services. You have to train your salespeople to be able to represent them, to defend them, to price them, whatever all those new skills that they need to learn on. So this is an AI agent that allows a rep to practice that and you create the content and the persona of this AI agent that responds to someone practicing the pitch and gives them coaching and feedback. And this is I mean I can tell you from our perspective at Salesforce we are just going so fast with new products and services. We are putting that into action for our sellers, because it's just more time and day where we can actually give the coaching that people want and need, without it being gated by but there are so many organizations with so many different sales strategies.

Speaker 2: 13:17

Yeah.

Speaker 1: 13:18

Is the coaching based on previous deals you have closed yourself, or is it based on certain Salesforce data?

Speaker 2: 13:26

Well, you get to choose what the content and what the focus of that agent is.

Speaker 1: 13:30

If you saw the keynote.

Speaker 2: 13:32

We have this perspective of there's always topics like our jobs to be done or work to be done, and then there's always data or content or actions that go along with that, and so you can imagine a world with, like, say, sales reps need to be chained on a new product Well, that's a product specialist. Maybe that new product has a new pricing model Well, that's a whole other skill in terms of defending and discussing a pricing model and negotiating it. So that might be a second little AI agent. And maybe, with this new product launch, you have a new set of competitors and so you want your personal deal companion competitive intelligence agent that help you practice these things.

Speaker 2: 14:14

So you can imagine a very like an ecosystem of these little agents that all are supporting these different jobs to be done.

Speaker 1: 14:21

And this coach does it. Do you only train with the coach, or does the coach also look at your deals that you're currently working on?

Speaker 2: 14:31

So the coach that we're working with right now is I'll put it in the category of the practice your pitch, learn new content, experiment it with a safe place and get the feedback that is structured around what best in class looks like. We do see a world forward, though, where you use the collection of capabilities we have at Salesforce and you use the collection of capabilities we have at Salesforce, and the way I put it is I call the sense and respond part of the equation. That is normally what the human has to do, and I'll take a sales example. A salesperson has a number. They have activities that lead to opportunities, that lead to progressed pipeline and closure. Right, that's a math model, right?

Speaker 2: 15:13

So what if you had an AI agent that knew your goal and saw your activity, could do the math for you, tell you that there's like great job, you're going to hit it, or you've got a lot of daylight between your expected commitment and what we're tracking for you, because we're seeing your activity and predicting it.

Speaker 2: 15:35

Now, that's the kind of thing that sales managers and sales reps do all the time. They pull out the back of the envelope, they stare at a dashboard, they do all that sense and respond the human way? And what if the AI agent did that for you and then took it to the next step and said and therefore, susan, I have gone through your, your territory, I've seen where you've got cold leads or deals that be progressed or customers that are lonely. I think you should call these three customers today. I've created the pre-call brief for you, I've determined that these are the products that would be the best conversation to have and, by the way, I emailed them and they accepted your invitation. So that kind of like where we move into that world where these autonomous agents do the sense and respond in conjunction with that, in that case, a sales rep, to be even more productive.

Speaker 1: 16:28

But I can also imagine if you're a sales manager and you have like a dozen sales reps under you and I don't know, the average deal closing is at 60% and you have one sales rep that's only at 25. If the system can tell the sales manager like, hey, this guy is not performing well because I don't know Same kind of thing, the old way of doing it.

Speaker 2: 16:50

You'd look at a dashboard right. And then you'd ask a couple more questions. You'd be like well, is it only this month? Is it every month? Is it a new person?

Speaker 1: 17:00

Can you also analyze their emails or their phone calls?

Speaker 2: 17:04

Well, we do have capabilities with a feature that we call Einstein Conversation Insights call like Einstein conversation insights and for organizations that do reported calls, you know part of that reported call capability is a big assist because then people don't have to take notes, because you can use generative AI to summarize them and you can use the generative AI techniques to extract intent and then do process like sentiment or even create the order and do other types of actions.

Speaker 2: 17:32

so like if people recorded lines, there's already a really good baseline. They are just for doing your job. But it also can extract things out like keywords and mentions of competitors and the way we've built that is also not just for the productivity of the sales reps but for the managers to see things at scale like, oh, we're seeing all these mentions of a certain request for information or competitive pushback or whatever it is, and then they can use those aggregate insights to take greater action.

Speaker 2: 18:02

That's already available so you could use it to coach an employee. Like I understand, like listen to the call, caught this keyword when they said competitor X. Like this should have been response Go, take this little coaching thing, this agent force coach thing, Learn these things, you'll be all good. You know that kind of thing. And how about the AI agent doing that for you?

Speaker 1: 18:20

So you can definitely improve your sales team with it.

Speaker 2: 18:22

Yes, well, and the improvement on the sales team. There's all sorts of ways you can approach that and in general, you know anything that is productivity gains usually are expected to be paid back in better time spent with customers and sales pursuits, so there's always those types of things. But then there's this perspective of repeatable, predictable process. So let's say, you've got a sales organization of 10,000 people and they all behave like humans, so they do everything a little bit differently. Like humans, so they do everything a little bit differently. What if you had AI all over the system that was doing things like?

Speaker 2: 18:57

Here's your perfect pre-call brief we've read everything about this customer, we've done a search externally, we've read the 10k, this is what you should go talk about. And with that you just get this organizational alignment about what showing up good looks like. So there's also that sort of impact in terms of like raising the bar of an entire workforce because they're being prepared in a greater way. So that goes beyond productivity. That's like showing up great consistently and being a great representative of your company and product. And then there's things that are more from the predictive side of the house, where you're using predictive signal to focus on this lead, but not that lead or this opportunity or this customer, because the predictive signal is giving you insight around a propensity to buy or to churn or things like that.

Speaker 1: 19:49

But that sounds like something only an existing Salesforce customer can use. That there's a lot of data already in your platform.

Speaker 2: 19:56

Well, I mean, we are talking about Salesforce customers and people who sit in this example.

Speaker 1: 20:00

But if someone's listening who's not a Salesforce customer. How hard is it to get that?

Speaker 2: 20:05

Well, you asked two questions you said what if they're not a Salesforce customer? And then you also said what if their data's not in Salesforce? So if they're not a Salesforce customer, we want them to be a Salesforce customer. So I'll just leave it at that. Because, we have like marketing CRM.

Speaker 1: 20:19

The question basically is to do what you just said, you need a lot of data from the company to give that advice. So I'm wondering how long will it take, if you become a Salesforce customer, to get enough data to make use of those kind of features?

Speaker 2: 20:35

So it's a really great question and you know, if you pop open a new Salesforce org, yes, it's new right.

Speaker 2: 20:46

And so the good news is with us with data is we've been really open to outside data like forever, forever as part of our journey at Salesforce and over the last couple of years we've really been maturing out a product line called Data Cloud, and what Data Cloud has is one of the most popular features is its ability to work with data that you don't pull into Salesforce.

Speaker 2: 21:10

We can use data in Snowflake, azure, gcp and AWS and incorporate that data into processes and user experience in Salesforce without moving it into Salesforce and without a really long technology process to gain access to it, so that the data is everywhere. We embrace that, either because we have the tools to bring it together and harmonize it, or because we have these zero data copy relationships with all the popular data lakes to be able to bring it into our ecosystem, whether it's visual, predictive or agentic or, you know, human in the loop, ai so if someone is not a sales customer but does have a lot of data on this previous sales cycles there is a big opportunity to import it, or to scan it, or to use it to make it even better.

Speaker 2: 22:02

Well, you'd want to be like sales service, marketing, commerce, like that's why you'd be working with us. You're either a Salesforce customer or you want to be a Salesforce customer.

Speaker 1: 22:13

That's up for the listener to decide, of course, but you have a lot of AI projects that are in pilot. Is the one we should look out for? Is there one you're personally very excited about? That's gonna change even more in this in industry, because a lot, a lot, a lot, a lot of it's already changed the last two years. It's going so fast, yeah.

Speaker 2: 22:36

Well, we talked about a couple that are just coming out of pilot right now on the sales side, you know, with the coach and with the SDR, the sales development rep.

Speaker 2: 22:45

One of the things that we have just on the pure platform side is, you can imagine, when you create these autonomous experiences, even though they're grounded with data, they're grounded with guardrails, they're sorted out with topics and instructions, you still want to test the heck out of it before you go and launch these things. So we have a feature that we're working on right now I don't know if it's going to be called the AI testing center or the workbench. We started to showcase it at Dreamforce, and the idea here is that we can use LLMs to go ahead and create example conversations where you know the utterances of the words are pushed into the tool and you can see how your agent responds. So, as you're, you know, working up that curve to go live, you can see how your agent is performing, and so the testing center is, like, I think, going to be a really sort of core feature of going live fast with the Gentic capabilities. And then there's other things that we're doing beyond that in terms of AI work benches, but that one's already out there from Dreamforce.

Speaker 1: 23:49

We'll keep an eye on that.

Speaker 2: 23:50

Yeah.

Speaker 1: 23:51

Thank you for this conversation.

Speaker 2: 23:53

All right, Thank you for the conversation Okay thank you for listening.