How One Company Cut Support Tickets With Agentforce AI

Many organizations experiment with AI tools but struggle to turn them into real operational value. In this episode, Eric Housh and Zack Terry walk through a real case study showing how siloed AI systems can create inefficiencies instead of reducing work.

They explore how Agentforce and Data Cloud connected AI directly to Salesforce CRM data, knowledge articles, and customer records. The discussion highlights how contextual AI reduces support ticket volume, improves response speed, and allows customers to self-serve answers through an Experience Cloud portal.

Prefer to watch? Find the full episode on our website.

Transcript

SPEAKER_01 0:09

AI in Action is brought to you by Fast Slow Motion. Our team helps growing businesses put AI to work with practical, scalable solutions. To learn more about how we can help you implement AI in your business, visit FastSlow Motion.comslash AI. Hello and welcome back to AI in Action. I'm Eric Hush, joined as always by Fast Slow Motions director of AI, Zach Terry. Zach, how are we feeling today?

SPEAKER_00 0:34

Oh, feeling good, man. It's Friday. Excited for the weekend. Excited to chat about AI. You know what?

SPEAKER_01 0:40

And I'm excited about today because uh, you know, in the in the 20 or so episodes we've had leading up today, we've kind of talked about theoretical concepts. We've talked about these frameworks where organizations can bring in AI. But today, the rubber is meeting the road. Like we're going to talk about a real case study of a client of ours where we where we were really able to take some of these concepts, leverage the AI tools of Salesforce with the context of their organizational data and drive some real results. So this client is a technical uh trade association. On the surface, uh, it looked like these guys were cutting edge with AI, right? They had a standalone LLM instance that was getting quite a bit of use. Um, but despite them being forward and having that LLM usage and sort of enabling their people and allowing their people to leverage AI, they were still drowning in support tickets. So uh what we're gonna see, and we're gonna dig into this today, that system was a silo. So their business is running over here in Salesforce, but um they're they're they're they're they're uh trying to use the LLM over on this side. So, Zach, this is kind of what we're we're seeing a lot, especially with these companies that are that are AI forward. There's a sort of AI gap, right? They they had the brains, they had the tooling, but they didn't have the context for their business. They didn't have the body.

SPEAKER_00 2:10

Exactly. This is super common. It's super common to have a business where users are working with a standalone LLM on the side, and then they're doing all of their work inside of the CRM and it's not really talking to each other, or they're having to take information out and put it in and then take it and put it back in. And so uh there's a lot of issues with that. I mean, it's it's inconvenient, it doesn't work well, it doesn't scale, but it's also probably not the most secure model in terms of hey, you're a business leader and you know your teams are doing things like this. And so anyway, if we if we look at this specific client and and what we did, the the organization had a smart tool, right? They had this AI model that they were using, but it was a stranger to their data. It wasn't connected to their data sources. They're living and working in Salesforce. Their customers are living and working in Salesforce because they have an experience cloud portal where they can come in and they can ask questions and they can open cases and they can look at some of the information and the knowledge articles that they have. And so all this information is in Salesforce in some way, right? And so, really, that AI could answer some general questions, but if they wanted specifics, they were having to take manual processes and copy it over. And uh it couldn't really understand information about a specific customer because it wasn't connected to their data source. It couldn't help them find audits or invoices and all this other stuff that they were tracking. And so um the standalone tool is definitely useful, but they were finding that it wasn't giving, well, it wasn't really reaching its potential, is what I would say. And so we were able to help them bridge that gap by leveraging agent force on top of the CRM, but specifically putting a service agent inside of their experience cloud portal so that their customers could interact with it. And so I think the key realization as we as we discovered the problem with this particular customer is they were already using AI. They were at the forefront of this, they wanted this to be a part of their business, but they just didn't really have it connected in a way that was scalable, that was automated, that was working for them. So they were kind of having to fight it a little bit. It was adding more work, even though they were getting more information by using that AI, it was still adding more work to their overall sort of process and work stream. And so if they really wanted AI to help reduce their workload, it couldn't be a separate tool. It had to be integrated and embedded where the work actually happens.

SPEAKER_01 4:27

And you know that was just frustrating for that team, right? Oh man, we've got this AI tool, but the you the support cases aren't going down. It's not quite working. And I think what what we're going to say, and and people will probably get tired of hearing us say this, is that context of the business was missing, right? So so there's there's a huge volume of queries, massive volume of support cases, standalone AI, just incapable of doing the work because it does not have the context of business.

SPEAKER_00 4:57

Context is king. It's super important. And AI tools on their own are super useful, but you we talked about taking information and copying it over and getting that result. Well, that's that's our attempt to manually bring the context in, right? So there's this sort of process that we have to take to go and do that, and it takes time out of our day. When you have it embedded in the system and it has access to that context automatically, you're no longer having to go and think about how to provide it. So you still have to ask good questions, right? You still have to prompt it well. And if there's context that's specific to like a given situation and it's not just the business or what's going on with the customer, you still have to do that. That's still good practice. But what you're not having to do is go and search for all that context on your own and pull all of that down, take it out of the system and put it somewhere else and then bring it back in, right? It's really all about having AI have access to all of that context from one central location. And I mean, from a business perspective, they're paying twice, right?

SPEAKER_01 5:49

They're paying for AI tooling, and then they're paying for their team to kind of manually do this effort. Absolutely.

SPEAKER_00 5:55

Yeah, you you you you've got the model that you're paying for. It's a separate tool, it's off-platform, and obviously you're you're paying for seats in Salesforce, you're paying for users to be able to access that, you're paying for your customers to be able to access that experience cloud portal. And then you know, you get you get what I would call like labor inefficiency. And so it's it's not a direct cost necessarily, but if if I could be spending less time doing the same thing, that's an opportunity cost, essentially, right? And so you you've got that labor inefficiency that's getting created, where if you have AI integrated in the right way, that should create labor efficiency, which is the same people doing less of the work that takes up more time and focusing more on the strategic.

SPEAKER_01 6:36

And that's that's a more scalable solution, of course. So let's uh let's descend into the geekiness, shall we? The big pivot here was moving into Agent Force, integrating this data layer with a conversational AI assistant that customers could then bypass that sort of uh human effort, interact with the AI directly. Let's let's kind of start there. Let's talk about the data 360 aspect of this.

SPEAKER_00 7:01

This is kind of like the secret sauce. And I say secret, it's not a secret. We talk about it all the time, but that's that's how you make it work. And that's why this particular offering that Salesforce has, I think, is so powerful, is because you get all this tooling sort of embedded directly into the platform. So if we kind of take a step back, look at the problem, and think about okay, I've got customers. They're putting in all these tickets. Most of these tickets are pretty quick and easy questions that we can answer. And so now we're wasting time getting these emails, responding to emails. Customers are having to wait for a response. Um, they're having to search through knowledge articles that are on the portal themselves, as opposed to having somebody intelligently answer it, or our human agents are actually going through and finding the answers and returning them back. They don't have the ability to go and find information that they need really quickly, specific to their situation as a customer, right? All of that is what we solved with Agent Force. And so essentially, from the customer perspective, now instead of them having to go and ask questions or create cases, they have an agent that they can chat with that is connected to all of that knowledge. And so they can get those answers almost instantly, as opposed to having to send an email in and wait for somebody to reply. How can they get those answers instantly? That's where Data 360 comes in. So all these knowledge articles that they have, these technical manuals, PDF files of information about their certification process and their audits and what that looks like. All that information is now stored in Data 360. It used to be called DataCloud, Data 360, right? And that that is turned into embeddings so that this agent can semantically search for information. And listen, when I started working in this AI stuff, I didn't really know what semantic meant. So I'm gonna explain that because I didn't know what it meant. But it's it's essentially the meaning of the words, the meaning of the query, right? And so it's it's not being constrained to a specific keyword. It understands that there are similarities with different types of words or even phrases, or if I take the same word and I put it in a different phrase, it carries a different meaning. And so all of that gets stored in these embeddings and data cloud, and it allows the agent to go and semantically search through that information so that you can get information quickly and accurately because it's grounded just on that business data. And so you get that semantic search that's empowered by Data Cloud. It's all grounded on your data. And not only is it grounded on the knowledge articles and the files that you have, but you can also ground it on your CRM data. And so this company is tracking a lot of key information about their customers in order to help them self-service. So not only is it a contact record, an account record, but you've got audit records, you've got invoice records, you've got all this information about the actual activity that's going on with the customers in the business. And so if they're asking questions like, when is my next audit? What should I expect with this audit? Can I reschedule an audit? Uh, what was my last invoice? Hey, when is my next invoice? All these questions that, yeah, you could go and click around and you can probably find it, but you may not know where to look, especially if you're a customer, unless you've been there for a long time. So now this agent is connected to not only that unstructured data and the knowledge, but also to their specific CRM records. And it's secure because it's running under the Salesforce security model. And so it only has access to the data that it needs because they're in an experienced cloud portal, they're logged in. And so the agent knows who they are and knows to only get the records that are associated with that customer. And so now all of a sudden, you you've got essentially what you you might hire a person to be able to come in and do. Uh, but now you've got an agent that can answer all these questions, can do it very quickly, doesn't have to go and search through a knowledge base and then send you an email back. It can just answer it directly in that chat window, can answer questions specific to your customer journey without having to go and get a support agent involved and you know look through that yourself. And so ultimately it leads to a better experience for the customer, but it's also a better experience for the business itself because now they're not getting all of those surface level questions as cases or emails or even phone calls. That agent is now able to answer it much quicker and deflect. I mean, now deflect is kind of a weird word because you know that I think it's like it goes against the the helpfulness of the customer experience and and what a service center really does. But if you look at it from a metric, the idea is, hey, if I can have somebody get an answer before they open a case, it has two primary benefits. They're gonna get an answer quicker, which means the customer's happier, right? If I have to go and email someone and wait, it's not a great experience for me. But then the business and its employees are happier too because they're not having to field all those low-level questions. They're getting less case volume and they're really only focusing on things that that actually need their help.

SPEAKER_01 11:47

Yeah. So I mean, it's a better experience across the board, right? I was just reflecting as you were unpacking that, like how frustrating it is to send an email in and have to wait for someone to assign the case and then wait for resolution on the case. Nobody enjoys that experience. Uh so obviously the the customer's experience, the employees' experience is is a you know night and day here, but even thinking thinking on this on another level, like the maintenance as the company evolves, as the customer's journey evolves, now we're not having to keep uh uh uh any human sort of up to date with what's going on, like the system's kind of self-feeding and self-maintaining.

SPEAKER_00 12:30

Yeah, there's so much here because let's say you update a knowledge article or you add a new knowledge article, well, it's feeding directly back into Data Cloud, which means it's feeding directly back to the agent. So as you add new information, as you update existing articles, if you remove articles, if you recategorize them, all of that is getting updated behind the scenes without you having to go and do anything other than change the article, right? All the CRM data. So everything that's being captured from a customer perspective, if they change their contact information, if you get a new record for an upcoming audit or an invoice gets added, all that information is stored in the CRM as it normally is, but it's immediately reflected. And so that agent can query all the most up-to-date information. The customer can even ask the agent to update their own information or provide other details that can then get written back to the CRM. And so really you kind of have this system that's self-reinforcing. Otherwise, if you kind of take it back to the initial state, which is CRM's over here, LLM's over here, they're not connected. Well, anytime you were going to make an update to a knowledge article, or if you were gonna make an update to the CRM, or if you needed to get a current snapshot of what's going on in a particular customer's journey, you've got to go and grab that information, figure it out, aggregate it, put it in a format that makes sense, describe what it is, put it over here in the LLM, get the response. And so it's just much more work. But if you've got it all embedded, that whole system, it's already your system of record if you're working in Salesforce and this is sort of your source of truth. And so as just natural business progresses, the agent's already connected to it, right? You don't have to worry about going in and updating that context. Whatever updates are being made and written to the CRM are reflecting back to the agent, and it's doing that according to its permissions and its access policy.

SPEAKER_01 14:12

Let's uh let's let you we we've we've kind of hit on this, but I want to stay here for just a minute because I feel like when people hear these conversations around how Agent Force and AI solutions can drastically improve a customer's experience from a service ticket standpoint, that there may be some trepidation, like, hey, we're we're talking about taking away jobs. That's not the case at all. So let's let's talk about now with this particular client. They have Agent Force, they have it on the Experience Cloud portal. What how does the world look different and better for the that support team that's actually been in charge of taking care of these clients up to this point?

SPEAKER_00 14:52

For the support team, it immediately it's reduced ticket load, right? So they they are not having to deal with 70, 80, 90% of these tickets that are coming in through email, through phone, through cases that are, you know, getting written through a forum on a website that are just basic questions that could be answered through a knowledge base or basic questions that could be answered by grabbing some information from the CRM, right? So that that's the biggest impact from the actual business's perspective. And and what that also means is more time for those agents to focus on solving difficult problems as opposed to having to triage all of those problems, or maybe they're ignoring some of the harder problems because they're they're dealing with all these urgent requests that aren't important. So giving you the ability to focus on the important as opposed to the urgent. Um and then I think also it may lead to not having to hire more people. And so um you may think of that a little bit as job loss because you're you're you know able to scale that business without having to bring on additional headcount. But I think if you're if you're you know a small business and you're looking for the ability to grow without having to add significant investment, this is one way that can help you get on a path to do that, right? So I, you know, I think I we can't say that like AI isn't affecting jobs. I mean, it it is. We've seen this happen, but I think the ideal scenario when we're talking about how do I use AI at my business is it's empowering your existing employees to be able to do more of the important things and less of the unimportant things, right? And so I think if if we look at the impact, it's definitely going to be that reduced ticket load. It's it's consolidated spend as well because everything is on one platform. And so you're no longer having to manage invoices from you know multiple different systems. Now it's all inside of that one single platform. And really that's that's Salesforce as a CRM. So you got Service Cloud, that's experience cloud, that's Agent Force, that's Data 360 with the vector database. That's all you know now consolidated into a single contract, a single invoice. It's increasing your knowledge velocity as well because you're able to just make updates and then it's immediately reflected in the agent. And so you're not having to do anything other than just maintain the knowledge base. All the updates are automatically propagated and it already understands it. The customer benefit is the self-service side. And so I don't know about you, but I don't really love digging through knowledge articles on a help center. I've done it, I've done it many times. And you know, a lot of times it's trying to Google it and hoping that it shows up or you know, using the search that probably doesn't work very well on the website. So having the ability to just ask whatever question I have and getting a direct answer. And even if that answer is, hey, I don't have an answer, I'm gonna connect you to a service agent to help you. That's still better than me having to go read through it, make that determination myself, and then go and open a case myself, right? And so it's it's giving the customers the ability to have hopefully a much better experience, quicker time to resolution, and less frustration having to search for things on their own. And then, of course, because you've got everything built on one platform and it's built on top of your business data, it's grounded on that data. And so you can be more assured that the responses are going to be accurate, as opposed to using an LLM that is not grounded on your business data, in which case it can basically reference answers from the entirety of its training data. So you do get the added benefit of using a tool like this where you're grounding it specifically on your CRM records, you're grounding it on your knowledge, you're grounding it on technical documents and files that you have. It's able to respond from that information, which can lead to reduced hallucinations and higher accuracy.

SPEAKER_01 18:29

And you gotta feel like, I mean, yeah, we're talking about agents doing the work that humans used to do, but you got to feel for those support agents now, having reliable, up-to-date, correct information and being able to work on more meaningful cases versus someone just needing something that could be easily uh answered by an autonomous agent is more rewarding work in general. So let's uh let's kind of land the plane here, uh, Zach, for you know, this particular client, they're involved in construction, it's a certification heavy industry. Uh, what's the lesson here for someone who may be listening to this and and if this sounds familiar to them, uh, kind of what do we want them to take away from from this time with us today?

SPEAKER_00 19:14

You've heard me say it before, but it's worth repeating. And that's your AI is only as powerful as the context that it can access. Context really is king. So if your customer data lives in a central location, it's much easier to build an AI strategy around that central source of truth. And tools like Data 360 really make that easier than ever. If all your data is in silos across different tools and different parts of the organization, it's going to be much harder to build AI that's going to benefit the entire organization. Now, that doesn't mean you can't build things in those silos. You certainly can do that. But if if your goal is to create something like this, which is able to take the entirety of the customer journey and get all of that context together and answer questions specific to a customer situation. That is one of the main benefits of having it all consolidated. So I think the questions that you can ask, is your AI embedded where your customers already are? If you don't have an AI but you're thinking about it, then think about do you have a central source of truth and a central location where you can get all of that context to provide an AI? Do you want your AI to perform actions? So not just referencing data, but being able to update a customer profile or checking on a specific thing like an invoice or the status of an audit. That's a great benefit of having AI built on top of a tool like your CRM and giving it the capabilities to do that. Agent Force just you have the ability to do that. You have the full capability of the metadata model, of things like flow and Apex to be able to go and not just retrieve information, but also make updates to the information that lives in your CRM. And then the other piece would be the unstructured side. So not just your records in the CRM, but um do you have things like technical documents and knowledge articles and maybe PDF files that are swimming around on a Google Drive somewhere? Um, do you have things like that where you could benefit from semantic search where an AI can basically go and find answers from that information and return that either to your employees or to your customers? So be just be thinking about, hey, where is this stuff stored? Where do I have all of the information that a lot of our employees probably have in their heads, but we've got to have it documented somewhere, right? Be thinking about how how could an AI tool benefit our employees and how could it benefit our customers if it had access to that information and could retrieve answers without having to go and search for it themselves.

SPEAKER_01 21:32

So one of the phrases I like to use around here is putting the cookies on the bottom shelf. So that strategy piece, obviously thinking about where the data is and and how we consider the data as the evolution of the business is one thing. But if we were to boil it down, three things a business leader needs to do, let's just say this week, this month, whatever, in terms of thinking about this pivot from using the AI tools over here, having all my customer data over here. To that more unified approach. Give me those three practical rubber meets the road takeaways that people need to be thinking about doing right.

SPEAKER_00 22:10

First, if we assume you're already using AI tools in some capacity, then do an audit and identify where you have silos. So are different departments using different tools? Are different employees in the same department using different tools? Figure out just what's going on in general. List all of those tools and then figure out how you can consolidate it. Also, you can you can check whether each one has direct access to your CRM or not. So some of these tools will allow you to directly connect. I would say you need a policy in place. So if you're not sure, you need to be sure, right? You shouldn't be allowing connections to an employee's personal Chat GPT, right? Like you need to have sort of an enterprise governance process in place there. And that's a that's a benefit of embedding this inside of your CRM, is that that's where it is. You can control it, you can control who has access to it.

SPEAKER_01 22:58

So anyway, this may be a bigger conversation than if you're wondering if your employees are using AI, they are. Exactly.

SPEAKER_00 23:08

And if you're wondering if they're probably taking some of your business data and putting it in there, they probably happening too. So you might as well give them a tool that's trusted and uh isn't gonna open up extra risk like that, right? So that's the first thing. Find those silos. The second is just assess your unstructured data, check out any of the manuals, the PDFs, the knowledge article, documentation you have in file services like Google Drive or SharePoint. Just go and look through that and think through hey, how do we use this? And is there some value that we can mine inside of this documentation if we had an AI tool that could just access the answers directly from that? I think that's one of the really great use cases of AI right now is the ability to go and derive value from unstructured data. And then last is just check out your knowledge level of effort and the labor and the hours that you spend assessing, updating, maintaining your knowledge base. How much time do you spend creating articles and if you're using AI for the AI to read? And how difficult is it for you to provide those updates? Do you have a seamless system where the AI is automatically connected to updates, things are published automatically, or is it more fragmented? Is it difficult to manage? And if the answer is more than an hour or two, you know, I would say there's probably an integration opportunity where you can have all the maintenance of the knowledge articles automatically feed into AI. And yes, of course, we're talking about Salesforce and Agent Force here, and so obviously all of these are benefits of using a platform like that. But if you're not, that's okay. Just think through how how could we benefit from some of these tools or some of these strategies. Um ultimately it's yeah, go find those silos, think about your AI silos, assess your unstructured data, you know, how how is that stored and could you get some benefit from it? And then review your knowledge systems and assess how much time you're spending and determine if you're using an AI tool, does that AI have access to that knowledge in a seamless way that doesn't require you to manually provide it?

SPEAKER_01 25:08

Zach, we're gonna spend the next several episodes unpacking similar case studies to this. I think if we have a takeaway for this one, it's probably a theme we're gonna hear a lot of times, which is don't think about AI tools in terms of features. Think about the power of AI in terms of business context, of your specific business context. And that's the value it can bring. So if you're out there listening, uh, you're ready to stop sort of managing the siloed chat bots. Maybe now we've sufficiently freaked you out that your employees are dumping sensitive company data out into Chat GPT, and you really want to think about how to bring this all together in a secure, scalable way. That's what we're here for. Hit fastlemotion.com slash AI. We'd love to have a conversation with you about uh some of these concepts and what they mean for your business. Uh Zach, bring us home.

SPEAKER_00 26:00

Eric, don't let perfection be the enemy of great. I think we see a lot of leaders get stuck in analysis, paralysis, especially when it comes to AI. You know, you want to have 100% perfect data. But the truth is it's it's all about that context. So when you've got AI that's integrated where your team already works, it can better understand the customer's story and their journey, even if the data isn't completely perfect, right? And so that's what's actually going to get results. So I would say become a context engineer if you want to drive measurable AI results. Don't get too bogged down in getting everything perfect, right? Go ahead and experiment.

SPEAKER_01 26:34

Love it. Zach, always a pleasure, my friend. To our audience. We'll see you next time. Thanks, Eric.