The Customer Success Playbook

Customer Success Playbook Season 2 Episode 46 - Dickey Singh - Cast.app - CS AI Agents

• Kevin Metzger • Season 2 • Episode 46

Share episode

Send us a text

In this episode of the Customer Success Playbook Podcast, hosts Roman Trebon and Kevin Metzger engage with Dickey Singh, CEO and founder of Cast.app, to explore the transformative potential of AI agents in customer success. The discussion reveals how AI agents can revolutionize customer engagement by handling routine tasks while enabling CSMs to focus on high-value activities, potentially generating millions in additional revenue without adding headcount.

Key Themes and Insights

  1. AI Agent Implementation
    • AI agents can handle mundane, repetitive tasks across the customer lifecycle
    • Focus on 100% account coverage without additional headcount
    • Integration with multiple data sources and systems (Salesforce, Gainsight, Snowflake, etc.)
  2. Role Evolution of CSMs
    • Shift from routine tasks to strategic activities
    • Focus on four key areas: empathy, relationship building, solving new challenges, providing expertise
    • Enhanced efficiency through AI support
  3. Technical Implementation
    • Sophisticated hallucination prevention mechanisms
    • Real-time confidence scoring for AI responses
    • Secure data handling without synchronization
    • Integration across multiple enterprise systems
  4. Business Impact
    • Pure Storage: $1.6 million in additional annual revenue
    • HPE: Significant improvement in customer engagement
    • Route: 3x industry average engagement rates
    • Substantial ROI improvements (1000-4000%)

Please Like, Comment, Share and Subscribe.

You can also find the CS Playbook Podcast:
YouTube - @CustomerSuccessPlaybookPodcast
Twitter - @CS_Playbook

You can find Kevin at:
Metzgerbusiness.com - Kevin's person web site
Kevin Metzger on Linked In.

You can find Roman at:
Roman Trebon on Linked In.

Roman Trebon:

Hi, everyone. Welcome back to the Customer Success Playbook Podcast. I'm Roman Trebon, and I'm here with my co host, Kevin Metzger. As always, we'd really appreciate it if you could rate, subscribe, and share the show with your network. Kevin, we're talking about your favorite topic today. AI and specifically AI agents. What would AI agents bring to the table, Kevin, your opinion?

Kevin Metzger:

Yeah, Roman, I think AI agents are really an exciting development as soon as LLM started getting good at conversing the on topics. The topic of a agent started to kind of gain relevance. AI agents are looking at tasks that people do, which require both process and decision making and really provides the ability to automate those those tasks and enable AI to help drive them. I'm extremely excited to talk to Dickey sing it. Cast app about this today, as they're a leader in the space and really focused on how they're using AI agents and customer success. Dickie is the CEO and founder of cast dot app, where he's pioneering the integration of AI agents and customer success with over two decades of experience and. Silicon Valley, including roles as CTO and VP of products at customer sat and leadership positions at several venture backed companies. Dickie has a unique perspective on how AI agents can augment human customer success managers. His current work at CastApp focuses on helping B2B companies grow and preserve revenue at AI scale. With proven results including generating 1. 6 million in annual revenue for pure storage through AI driven customer success initiatives. In today's episode, we're exploring the transformative role of AI agents and customer success and how they can amplify the effectiveness of human CSMs. We'll dive deep into practical strategies for implementing AI agents to handle routine tasks, analyze customer data, and identify opportunities for expansion, while enabling CSMs to focus on high value strategic activities. Dickie will share insights on how organizations can successfully integrate AI agents into their customer success operations, measure their impact, and maintain the crucial balance between automation and human touch in customer relationships. Dickie, welcome to the show. Let's get started with the basics. What specific tasks and responsibilities can AI agents effectively take on to support CSMs in their daily work?

Dickey Singh:

First of all thank you, Kevin and Roman. Thank you for having me. AI agents is a very exciting topic. as you remember, Kevin, we geeked out on it quite a bit the other day. Yes. to answer your question, how can AI agents, what can the agents do? Every mundane, repetitive, and follow up task. How about every task that bogs down a CSM today? You know, so, inefficiency is killing the CSM. And there are several examples, like, they have to manually execute playbooks, they have to triage, and they have to, prepare and present content. So, Imagine a CSM jumping between three tools to understand what the customer purchased, how the product onboarding is going usage, adoption, churn, renewals building a deck, chasing down executives for calendar availability, praying the executives show up, delivering the presentation and answering the same question again and again Then email the presentation to people who did not show up, which happens a lot. Let me paint a different picture Now imagine you could scale to 100 percent of your accounts without adding a single headcount In customer success, account management, onboarding, renewals, sentiment analysis, and churn mitigation. And obviously, I don't mean do more with less mantra. That is well known to worsen customer and CSM satisfaction, growth, and churn. Let's think much bigger. Not only engage and influence 100 percent of the accounts, but also engage and influence every user and every decision maker, wherever they are, without wasting a minute of their time. Again, without adding a single headcount. How? by placing AI agents in the middle of your business and your customers backed by skilled, lean, and focused teams on the side, you can not only scale to every user, exec, and account, but also improve the team's overall satisfaction.

Roman Trebon:

so Dickie, you had me at hello there. I love what you're talking about here. And I will say that Kevin, when Kevin said that we were having you on the show, he was speaking very highly of the solution and how it works. and, you know, it's, I think real game changer, Kevin, what you said, not to put words in your mouth, but you were, you were super impressed by it. Dickie, I'm curious, for a company that doesn't have these AI agents in today, how would they go about doing this? Like, it sounds like it's their operational change, like even organizational change, you need to support this. How, how do you go and it sounds great, big picture sounds great, but how do we actually tactically get there?

Dickey Singh:

depends who you ask, right? If you ask a product person or engineering person, everyone wants to build the AI agent themselves, right? if you remember a few years ago, they'll say oh, we do this AI and everyone's, oh no, we are doing AI too, right? without quite understanding, and it's the same thing 20 years ago, oh, we do SAS as well. Right. things have changed. there will be an AI agent in every operating system, every product, every website, every portfolio, we used to say there's an app for that. Now we're going to say there's an AI agent for that. Right? So that's what I feel and agents will do different things like how we are embedding an AI agent within the presentation that the AI agent presents and answers questions. Similarly, there will be agents for SDR roles. There will be AI agents for, customer support roles and AE roles. So this agents are going to help a lot of people do a lot of mundane tasks that require like manual work today.

Kevin Metzger:

following up on the question from Roman. with the agents, as they get propagated, how are we want to go below the covers a little bit? Right? So, are you looking at when you build agents? Are there going to be agents that coordinate? Are you going to build specific agents for specific tasks? how do the agents get trained to know what to do?

Dickey Singh:

So what we do is we have AI agents for customer facing teams, right? Which, in my opinion, is the 4 out of the 5 S. So support being the 1st S. we don't do anything about support, but then look at services, onboarding services, technical services. we look at expansion sales. Obviously success is there, but there's another often overlooked one, which is automated sentiment analysis. For example, somebody gives you a score of nine or 10 on an NPS score or a score of seven on a customer effort score or a score of seven on onboarding effort score. You in the same breath want to ask for a referral, not five minutes later, not an email that goes up 20, 20 minutes later in the same breath. So we are able to do that. So an AI agent across the board. you don't need an AI agent for renewals. onboarding, upsells and cross sells and churn mitigation. You need just an AI agent that does everything for the lifecycle of an existing customer at the user level.

Roman Trebon:

and I was on the website earlier checking it out, great website by the way. I was reading through the customer testimonials and there's a whole bunch. So I'm curious from your experience, for companies that are about to go on their AI journey, maybe they're looking to implement AI agents across their customer facing operations. What are some best practices or things they should be thinking about heading into that journey and maybe some common pitfalls that they should avoid.

Dickey Singh:

Yeah, there are so many to list. every product engineering team wants to build their own agent. it takes a few hours to get an agent up and running using these tools, but it takes a long time to perfect it. that's where the real challenges lie. you have to take care of hallucinations. I mean, like AI. Hallucinates, AI agents hallucinate even more because you know, our, our AI agent, somebody asked, who's Morgan Freeman? And it tried to answer that, we had to train it not to answer these questions and we have built a lot of mechanisms around that. So we have built a lot of mechanisms, so they don't listen to it and we actually detect it and connect them To the CSMs account managers or onboarding specialists. The other thing is not every data set is available as a vectorized data set, right? So we deal with enterprise customers and, you know, large or small. But let's say we take an example of HPE, We have 17 different programs with HPE now. Every different program uses several data sources. Somebody is using Gainsight, Salesforce, and Snowflake. Someone's using Totango, HubSpot, and something else. not all the data that we have is available as a RAG based system. You cannot use that in real time. So we have to internalize a lot of knowledge in our AI system. Yes, of course, we can use. RAD based systems when available from our customers, but a lot of times their data lives in Snowflake, in Databricks, in other systems. our AI can write queries, across Snowflake, Gainsight, Tatango, and Salesforce reports automatically. We're using the exact same things like what AgentForce is doing behind the scenes, we are covering multiple products. they can say, hey, write me a query that goes to this system for this, this system for this, and this. And once the query is there, the A. I. agent can, answer questions, to summarize, I think it's very easy to get started with an A. I. agent, but it takes a really long time to perfect it or deploy it in the field.

Kevin Metzger:

So, Dickie, as you mentioned, the agent force and what Salesforce is doing, that's specific to Salesforce Absolutely. Test app actually sits outside of any specific application and kind of grabs data from all the applications to bring it in and act on data wherever it exists, how do you ensure that propagates properly when you do that? So if you have an action that moves a case from one state to another state. How do you make sure that's getting propagated back properly? How, how do you ensure all of the data flows both in and out?

Dickey Singh:

So you're touching on a mode that we have, which is we taught our systems how to unlearn. the case status could change from new to working on it escalated to closed. Similarly, the user's journey could change. For example, a customer may be a late stage prospect. They could switch to onboarding where you start working on. Now they have some usage. Now they have adoption. Now they have become an expert in your system. Now you can ask for a referral because you only ask referrals from experts. you have to Continuously learn and unlearn the state of the person. And remember I said, person, not account. the best way to do enterprise grade onboarding is to separate the account onboarding from the user onboarding. Users will come and go. They will hire a new executive. They will hire like new users two years after the product was deployed. You have to onboard that user. You cannot onboard the account because the account was onboarded two years ago. A salesperson sold something bundled. the customer started using two things, but not the third one. They want to use the third one six months later. How can you do that account onboarding if you have onboarded the account and user as a whole, right? So those are kind of. Things we have to teach and unteach the AI to respond to the customer questions.

Roman Trebon:

Vicky, I'm curious, do you have to do a lot of client education to overcome fears of implementing these AI agents across their customer lifecycle? I hear what you're talking about, Kevin, you have to be careful of hallucinations, you got to train and retrain. How would a client have the confidence that, the AI is performing as anticipated? how much overhead is needed to just to make sure there's that AI is performing the tasks needed and, and then when do you get to a comfort state where you say, I feel good. My AI agent knows that he's we're good to go.

Dickey Singh:

Yeah that's a very deep question. It could take a whole session to answer that, but I will just answer it in two different ways, right? So how do we detect hallucinations? We basically calculate an on the fly confidence level of the answer that's coming. If the confidence is between zero and 0. 7, we will answer it with a lot of confidence. We will say here's the answer. This is exactly how things should be done. But if it is between 0. 7 and 0. 9, perhaps, we'll say, I think the answer is this, but you should cross validate it before using it. But if it's worse than 0. 9, we have a reverse scale, Instead of 1 to 0, it's 0 to 1. If it's worse, then we say, I'm so sorry, I do not know the answer to this. May I connect you to a CSM account manager, onboarding specialists, And then we look at what segment they have. If it is a high touch customer, we will put the calendly of a CSM. If it's a high touch customer and the C level executive, we'll put the phone number. If it is a low touch customer, we'll generate a form on the fly and put that form in front of the customer. when I say we, the AI is doing it, but I'm taking credit for it. But that's exactly how it works.

Kevin Metzger:

Yeah, I like that. So if we've got the AI agents doing all this work, what's the role of the CSM becoming in this world where the agents really driving a lot and now the CSM. When it's exiting to the CSM, what's the CSM need to know? How's the CSM know what's happened most recently? Are we presenting that data and what's the skill level? Is the skill level of the CSM now having to increase a little bit because some of the lower skill level stuff is being taken care of by the AIs.

Dickey Singh:

Yeah, so what we are kind of taking care of, like as I mentioned earlier, the repetitive and mundane stuff. And to be honest, these are the kind of things that CSMs actually don't want to do. Who wants to keep on writing follow up emails and saying hey, When are you available? Every month for a meeting. And then and then who wants to like, presented again to the champion who helped you create the presentation earlier in the day, and the other executives did not show up, right? CSMs don't want to do a lot of these mundane tasks, right? But there are four things the CSMs can do a really good job of that the AI will never be able to do. one of them is empathizing with the customers. AI will not be able to empathize. Building customer relationships. AI is not going to build customer relationships. Only a CSM can do. AI can only solve problems it has seen before. But there are lots of new challenges. That the CSM's account managers, onboarding experts sentiment analysts should focus on. And, and, and, and the, and the last thing is CSM should focus on providing expertise to their customers. So, if they take care of these 4 things, the rest becomes easy to do for the agents. And, and that's why I kind of, I, you know, when they talked about pool model and pod model and why pool and pod model are never successful. And the reason for that is like simple. You added maybe 10 more accounts to a pool of three, right? What you did is you reduce their ability to solve problems in our case, what we are doing is just like two or three hours of work from one person is usually enough. In a week to handle all your customers, we can talk more about it.

Roman Trebon:

Yeah, it's great. I mentioned going to the website and checking it out. And like I said to Kevin before the call, I'm like, wow, there's a lot of success stories here. can you share with our audience, some of your favorite success stories, how customers are getting value out of the solution?

Dickey Singh:

Yeah. it kind of depends on your business. I don't want to use the word B2B2B because it confuses me also, let's take a customer. HPE is our customer. HPE uses our product to communicate with their customers like Coca Cola, Lockheed Martin PwC and others, we are not a co pilot. Co pilot means a productivity tool for team members to communicate, to like do things better. So they get like what 20 percent improvement, 30 percent improvement at best, right? That's what productivity tools do. What we are doing is talking directly to the customer's customer, like PwC in the case of HPE or PepsiCo in the case of some other customers, first of all, expect that the return on investment will be way higher. Right. So it's not 10%, 20%, 30%. You're talking like 1000 percent to start with to 4000%. that is the reason, we get such incredible numbers. one division of pure storage, which you mentioned is adding 1. 6 million to the bottom line. Another division is sitting on 100. 50 hours per customer by using AI driven education compared to an LMS system in which you tell the customer, here are 10 videos of 30 minutes each. Once you watch them, we'll give you a certificate, That doesn't work. So, HPE customers like PwC, Lockheed Martin and others, gave a score of nine out of 10. for the value it provides. their customers love what they see route this Canadian company gets three X, the industry average for engagement. Aruba reported a 6 percent increase in licenses. one of our customers, we come solution sites that the casts prospect pre loading pre boarding was the reason they won a really large multi million dollar customer called Skyline. since we start from pre boarding for late stage prospects, all the way to off boarding, every, every, and everything in between, kind of you know, so we're not just renewals. We're not just doing upsales. We're not just doing adoption or usage. We are doing everything kind of together, right? That's why they get, high numbers.

Kevin Metzger:

Dickie, how are you managing data security in this scenario?

Dickey Singh:

yeah, we, we have a cyber security customer that we, they don't want us to list on our website. Just let's say that they send information to, they send information to Will Smith, who's one of the celebrities, They are behind the scenes. taking care of a lot of like systems and then they have to notify customers what they did, right? And my favorite saying is a tree falls in the forest is nobody. No one to like here. Did it really make a sound right? So if you have a product and you're doing good things for your customer, and if you're not sharing that product with your customer as insights and giving them recommendations, your product will be churned, right? Because people don't know what, what value that they're kind of getting. And I lost my train of thought. What was the question? Sorry. We were in security. Oh, yes. So yeah. So I was saying that we have cyber security. Yes. So one unique thing is like you might have heard that a lot of CSPs do synchronization between Salesforce and their systems, They copy data from Salesforce and they take their data and copy back. We don't do synchronization for one simple reason. We looked at every CSP synchronization does not work. So we don't copy our customers data into our systems. And that's why we are inherently secure. When we are creating the presentations, we write queries and everything for each slide, for each variable. We'll get the data into and create a presentation, but we won't copy their data over into our system. And that's one of the reasons why we have cybersecurity companies as customers.

Kevin Metzger:

I'm going to follow up on that. just follow up on that for a sec. but you are doing a learning process and I'm assuming you're using data from these systems through the learning. So there's some fundamental information that's in whatever LLM that you've trained and learned on. Is that protected through, like, I don't know whether you're on Amazon. I'm not sure how you're hosting, but is that protected through how it's hosted and the LLM itself or yeah, this

Dickey Singh:

could be another podcast to be, you know, so as I said, we don't copy data over and we use something called LLM masking. What that means is we don't share who the customer is, what the user is, when we talk to the LLM. We use seven LLMs, as I told you the first time we chatted, we recycle them depending on which has the least latency All these LLMs are Co pilot oriented. If you go to OpenAI, cloud or LLAMA and ask questions, they're very good at answering Why? Because you have to write the prompt. In our case, we are doing this at such a long scale, like 100, 000 customers, 10 personas each, 10 users each. You cannot expect the CSM to write those prompts. So we generate the prompts on the fly. Since we are able to generate the prompts, we can do the masking on the fly. we will not say PwC, we will say customer CX423 and the second time we will ask customer someone else on the same thing. So the LLM does not know who we are asking about.

Roman Trebon:

So Kev, any more questions for Dickie before we get to the hard hitting stuff with the rapid fire?

Kevin Metzger:

Let's go to the rapid fire.

Dickey Singh:

is pass an answer for those? No, just kidding.

Roman Trebon:

Alright, Dickie, Early bird or night owl? You go to bed. Get up early or you staying up late.

Dickey Singh:

If I have a customer meeting, I'll get up really early, but I'm usually don't go to bed till like two 2:00 AM or so, so Ah, gotcha. Maybe 7, 7 30 or so. Gotcha. You're on the West

Kevin Metzger:

Coast, so you're gonna bed about the time we're getting up, That's right. So do you enjoy cooking?

Dickey Singh:

I have a smoker. I barbecue some stuff, like I'll take salmon and put it for like three hours. I'll take ribs or beef and put it for like six hours I put small slices of oranges on top of salmon and then put them in foil, which my wife doesn't want me to do, but, and then cooking. That

Roman Trebon:

sounds good. That sounds good. Alright. It's almost Halloween here. You have a favorite Halloween candy?

Dickey Singh:

No. We, I don't, but my kids do I think they give out a lot of Kit Kat and other things. Oh, kit. Yeah. My favorite candy would be a Toblerone, but I think it's too expensive.

Roman Trebon:

get a lot of those in the Halloween basket.

Kevin Metzger:

And you're where, I know you're out in California, what area?

Dickey Singh:

I'm 20 minutes south of San Francisco in a city called San Carlos on towards the hills on 280.

Kevin Metzger:

Okay. And so if somebody came out to San Carlos, where should they eat? What should they visit? I'm still in Robins question.

Dickey Singh:

Oh, yeah. Every good restaurant that shuts down in Palo Alto ends up opening in San Carlos, but you know, there's so much variety this town. But the restaurant I would recommend is a couple of cities over. It's called Aspetas. It's a Brazilian steakhouse. It's my favorite. even if my customers You know, don't don't like Brazilian steakhouse. I'd make it a point to take them over there because I love it. So It's in san mateo, asparagus. It's a brazilian steakhouse. It's like amazing food. Oh, awesome.

Roman Trebon:

All right, dickie, where can our audience find more about you more about cast on? App where can they find more?

Dickey Singh:

Yeah, if you want to I'm on LinkedIn as Dickie Singh, D I C K E Y S I N G H. the best way to connect with me is either on LinkedIn or just go to CAS. app and scroll down and there's a form you can just fill in some information and just connect with me to open Calendly automatically for them.

Roman Trebon:

Awesome. Awesome. Well, Vicky, thanks so much for coming on the show. I really enjoyed, you know, talking to you and learn more about the product and, and, and what's happening in the world of AI. I really appreciate your time.

Dickey Singh:

There's a lot happening in AI, but this is just barely scratching the surface.

Roman Trebon:

Yeah. We'll have you on again. It'll be completely different. I can't wait to hear what will emerge. thanks for being on We appreciate you listening. If you found this episode helpful, don't forget to subscribe, leave a rating and share it with your network. We are also on LinkedIn. You can find me at Roman Trebon. You can find Kevin at Kevin Metzger. Make sure to check out our customer success playbook page on LinkedIn. You'll see our upcoming episodes. You'll see a clip of this interview and others that are upcoming. Reach out, let us know what other guests and topics you'd like to see on the show. as always, we appreciate you listening and keep on playing.

People on this episode

Head Shot

Kevin Metzger

Co-host
Head Shot

Roman Trebon

Co-host