How a 5-Person Team Built 150 AI Workflows That Changed an Entire Company

By the way, what was step number one towards this 150 plus AI workflow output? You're listening to Unchurned, brought to you by the Gainsight podcast network. When Cloudbeds handed Colin Slade zero budget and a CS team at 120% capacity, he did something radical. He told his ops team to drop everything and bet the house on AI. Nine months later, 75% of grunt work gone, 7,000 hours reclaimed every month, and a company wide revolution sparked by pure fomo. I'm Josh Schacter and this is Unchurned. Hey everybody, and welcome to this episode of Unchurned. I'm your host Josh Schacter and I am here this week with Colin Slade.

Josh Schachter [Host]:
Colin is the senior Vice President of AI strategy and Customer Success at Cloudbeds. Colin, thank you for being on program.

Colin Slade:
Josh, excited to be here.

Josh Schachter [Host]:
I start out just out of like, I don't know, unoriginality every episode saying I'm so excited about this episode because, you know, blah, blah, blah, and then I'm. Sometimes I'm really excited, sometimes I'm not so excited. But I am so excited about talking to you today. And I say that with a lot of emphasis because you and I have been chatting for a few weeks right now. As you know, in my role mandate at Gainsight, I'm the senior Vice president of strategy and market development. And the market development piece there effectively means understand the market, understand what's going on right now in the world of AI and the confluence of AI and retention, post sales, customer success, all the things. And I've spent the last several months since I joined gainsight talking to VPs, CCOs, leaders across the board in SaaS, understanding, you know, what they're doing with AI. And there's one person who in my opinion is leaps and bounds ahead of everybody else.

Josh Schachter [Host]:
And guess who that person is. Colin Slade.

Colin Slade:
I hope it's me, since they're asking.

Josh Schachter [Host]:
It'll be real. It'd be really, it would be really awkward.

Colin Slade:
Timmy.

Josh Schachter [Host]:
Yeah, yeah, yeah, yeah. My peer. Jenny. Yeah, no, it's Colin. It's you. It's you. You're like so far ahead of everybody, but in a way also where we know that, that things are in dog years right now with AI. So like within three months probably my hope is that Everybody gets.

Josh Schachter [Host]:
Not three months. It's gonna take more than three months, but I hope it's that within nine months we're talking to many more leaders that are at a place where you are right now today. And I know it's been a very concerted, intentional effort for you to get to where you are right now as my poster boy for AI leadership in CS transformation.

Colin Slade:
I appreciate that.

Josh Schachter [Host]:
Yeah, of course. So we're going to talk about all things. We are filming this before Pulse Europe, Gainsights, Pulse Europe. That being said, we're filming this partly in preparation for my entire Pulse presentation around AI is devoted to you, to a spotlight on what you are doing. There is that much stuff that you are doing in AI that I decided. I told the team at Gainsight, the marketing team, you know what, I'm not going to sit there and like, and give like, you know, key takeaways from my learnings in the market, blah, blah. I'm just going to showcase Colin. I just want everybody to know, unpack what he's doing so they can learn from him.

Josh Schachter [Host]:
So we're doing this as an addendum to that. Of course, of course. So, and this will come out right, I think around the same time or maybe a day before we do the presentation in Dublin.

Colin Slade:
And.

Josh Schachter [Host]:
And yeah, it'll be good content for people to follow up with, to get to hear this in your own words, since I'll be the one relaying it out in Dublin.

Colin Slade:
Okay.

Josh Schachter [Host]:
With that very long introduction. Welcome. And why don't you. Do you want to share? You want to give us like the high notes of what's been going on, like what the culmination of your efforts have been in AI?

Colin Slade:
Sure. I mean, I think we're both long winded, so you can cut me off and keep me directed. I mean, you know, I studied AI way back in the day. So we finally arrived at a point where I can actually do it. I led customer success for seven years, then I left, came back, led the people team and then cloud beds. Yeah. And so about nine months ago, kind of took back over the people team and pretty much was like, you need to kind of turn this around. You have zero budget, you have zero resources, you don't have any tools.

Colin Slade:
And I'd already been dabbling in AI on the side. Right. I'm definitely one of the users who's constantly, every night I'm spending watching YouTube videos about rag vector databases and, you know, joining all the different communities and trying to build stuff with lovable. And so I had Already had a lot of ideas. And so now it was a very much like, we need to try to apply this in an enterprise level at a company and see if it works. And luckily I had two co. The two co founders of Cloud Beds are very much in AI, especially the CEO. And so he was just sort of like, yeah, go figure it out.

Colin Slade:
Like just go try things. And so I took a cs. So we had a. No one was really using AI. We'd all been given Gemini. There was low usage, people using it for little bits and pieces, but not really anything. And so I took a csops team. I remember like sitting down with him and just basically saying, I know you've never heard of this tool called nnn.

Colin Slade:
I know you don't really know about how to. What agents and automations and differences are, but this seems like the best team to do it. Do you want to do it? They're all like, they're excited. And so we just started to build stuff and to where now we're building. I mean, they just launched a couple of new things today. We're building stuff in days versus weeks and months and it's been awesome to see and really fun to get real value out of it. And it's worked, which is fairly surprising based on where we started.

Josh Schachter [Host]:
What'd you launch today?

Colin Slade:
So the team was like listing off things. I literally stopped the team and I was like, I just want us to remember what this conversation was six months ago. So we launched our knowledge system to the entire sales team today. So we launched a little brand bot. So as of today, cloudbeds has a new brand. So we created a little AI chatbot that you can ask any question about the new brand and we'll tell you the answer based off of all the marketing materials and all the team and so. And then we also launched a new Zendesk macro search. So now you can find any template inside of Zendesk that you need to use when you're responding to a customer, whether you're account manager, an onboarding coach, or we call them coaches, but.

Colin Slade:
Or a sales support person.

Josh Schachter [Host]:
Awesome. So I buried the lead a little bit here because I praised you, but I didn't really explain like why I'm so impressed with what you've done.

Colin Slade:
You've.

Josh Schachter [Host]:
You've launched over and of course it's grew today, but you've launched over 150 AI workflows, slash agents. Sometimes there's a fuzzy area, right, gray area between what's an agent, what's the AI workflow, and we like to claim it's agents, but they're not really agents, but it makes us sound better, you know, blah, blah, blah. But you've launched 150 plus of those things. You were with Cloudbeds Pre ZIRP, right? No, sorry. You were with cloud beds during the ZIRP era. The zero interest rate. Like the glory days of SaaS when, you know, you had unlimited budgets and everything was off and to the right. And then you took another role.

Josh Schachter [Host]:
You came back postserp when there is tightening of the belt.

Colin Slade:
Uh, you came back and post launch of ChatGPT, and post launch of sort of AI coming into the space, which obviously had never really existed. Right, right.

Josh Schachter [Host]:
The company was still growing. I mean, they, they'd added 5,000 customers that previous year. Uh, you had zero. But, but, but, like there were still constraints.

Colin Slade:
Right.

Josh Schachter [Host]:
You didn't have budget. They didn't give you any budget for hiring or for tools. Uh, the team was overstretched. You told us they were at 120% of capacity, your post sales team. Um, and you'd also mentioned before that there was just a fragmented experience, customer experience, different silent information across the teams.

Colin Slade:
Yep.

Josh Schachter [Host]:
Okay.

Colin Slade:
Yeah. And I think we, and I think a key piece of this is that we had gone through a riff maybe six months before I took back over, and so we'd also lost people. So we also have 40% less staff than we had, like, you know, let's call it a year, year and a half ago with these, you know, additional customers, with this additional product lines. You know, it's not like everyone stops developing new stuff. So the complexity got harder, the customers got more complex, and we had less people. Not just. We didn't hire new people. We literally lost a decent percentage of people over time.

Colin Slade:
Yeah. Wow. Yeah.

Josh Schachter [Host]:
I mean, this is classic. This is not just cloud meds. This is a, this is a SaaS story. This is a SaaS industry story, you know, post sales story as well, for sure. And when you came in, like you said, you had the support, you've had the support of your. Of your executives, of the CEO, but basically they wanted you all in on AI as well.

Colin Slade:
Right.

Josh Schachter [Host]:
So it wasn't like you had to push uphill against that. You had that endorsement.

Colin Slade:
No. Yeah, but they weren't like, they weren't fully convinced. Right. So it took. There was a lot of executives at the company who were sort of like, I don't know. I don't know if this AI is a bubble thing, I don't see how it can help Me within my department. You know, I'm not convinced that you can, you know, augment real people. It still is very much like, yeah, it's cool to ask where I should go out to eat or act as a therapist or help me write an email, but, like, can it really do real actions? And so that was kind of my directive from.

Colin Slade:
From our CEO was basically like, you got to go prove to people that, like, this is legit and this is powerful. And. And now, as of a couple weeks, like, a week or two ago, I'm now the head of AI strategy. So now I'm taking over AI strategy for the whole company. So it took a while to obviously get to that point, but everyone has seen what we've been able to do with the small team of. It was originally five people. Now it's down to four people and me. And so that's the team that's been building all of this.

Colin Slade:
And one person knows a little bit of Python, kind of. He's the most technical of us all, for sure. But everyone is just excited to learn and just hustling to try things. Yeah.

Josh Schachter [Host]:
And you came in there. The goals from a business standpoint were to improve churn, create stronger activation during the onboarding journey process, and then again, unify that fragmented customer experience. So.

Colin Slade:
Yeah, and pretty much survived too. Right. It was like, you know, everyone was at a. You know, at a. Again, like an over. Like, we were asked to do more, needed to. So all those things were broken. And obviously our number one directive at post sales is to, you know, activate the new sales and to keep.

Colin Slade:
And, you know, keep the current sales and lower churn. And so that is always our number one goal. But there were also just other pieces that had to happen to the team. Needed to feel like they were part of something and change. Management's not talked about enough. Like, we gave everyone Gemini, and no one used it because they were all afraid that not all of them, but a lot of them were afraid that if they. If it showed that it could do a better answer than them, that they'd be replaced. Right.

Colin Slade:
And so it's hard to get everyone to actually use these things, even if you're paying for them or even if you set them up.

Josh Schachter [Host]:
How'd you know that? How did you know that there was that fear? I mean, that's not something I would imagine would come directly to your office. Uh, you know, people would say to you directly or. Or did it.

Colin Slade:
Not really, no. It's indirect. Like, you start talking to the leads and the managers and the smaller teams I started to, when I first came in, I went and met with a lot of the different people and leaders and, and teams and they don't say it directly, it's more like, ah, I gotta read a lot or I just don't think it does a really good job or. And you're kind of like, yeah, but it's, it's better for you. And I think what, what I realized two things. I realized one, I had to let them know that their job was not at risk. Mm. Like this wasn't going to replace them but if they didn't use it and try to like get the full value and like we could basically put, we could train them to be the best, whatever, CSM customer six, you know, customer support, like actually using AI better than anyone else in the industry.

Colin Slade:
They could go get a job anywhere. So I basically was like, give me your, like if you're willing to learn and push yourself, I will give you the tools, I will give you the training. And then if it doesn't work out here, doesn't matter, you'll be at the top of everyone's list because you know how to, to implement AI into your daily work. So that was one piece and the second piece was we initially tried to have people create prompts and like, you know, figure out how to use all the tools. And what we realized pretty quickly is that it's a very much a baby steps thing. Most people's mind is not opened up to what is actually possible with AI and so we had to show it to them first and we, we have the same that like we had to make the AI invisible, like it had to be within their daily life and it had to do things that made them their life easier and they didn't need to create a prompt or do anything crazy like, you know, contextual engineering or any of that kind of stuff. We were provide them basically the solution or the augmented, you know, partial solution and then they could add the rest. And we're in hospitality tech so we're never going to not have people who want to talk to people.

Colin Slade:
Right. It's never going to be a. AI is handling everything. This was trying to make our team, you know, two, four, five times more effective and have all the information that we store in our systems actually available to be used and do a bunch of the busy work and the manual stuff for them so they can focus on the relationship and the outcomes.

Josh Schachter [Host]:
And this, like you said, this wasn't a nice to have. This was a must have for you guys. As a business, in our case, for sure.

Colin Slade:
Yeah, yeah.

Josh Schachter [Host]:
A lot of people out there listening in their cases as well. And what you were mentioned about, like the, the psyche of the team, it actually reminds you. Was it, was it Franklin Roosevelt? The only thing you have to fear is fear itself. I hear that, you know, from folks. It's like it was at a conference yesterday with the sales leader, sales executive at OpenAI, and she was talking about, you know, future of the workforce with AI and it was the same thing. It's like the only people that really should be concerned about AI are the people that are, you know, not like that, are afraid to get into it now.

Colin Slade:
Like ignoring it.

Josh Schachter [Host]:
Ignoring it, you know? Yeah, like that. Those are the only people that should like the people that embrace it. Those are the ones that it's going to help amplify and augment. Okay, what was step one? Coming in here, you've got this mandate. We have the context now of what the situation was at the company, what your expectations and goals were with your CEO. What was step number one towards this 150 plus AI workflow output?

Colin Slade:
Yeah. So, I mean, we. Step one now would be a little different than step one, what it was. So initially I tried to go, you know, 100 miles an hour and go, all right, we're building out all these workflows. It's getting in. You just dive in there. And no one knew what the heck was going on. Right.

Colin Slade:
We didn't have any knowledge, any buildup. And so our very first thing we tried to build was this super complicated customer feedback tool where in Slack you could, you know, send in a ticket or send in a customer call and would automatically analyze it, check JIRA for duplicates, then go check this. The product manager for the PRD, like it would do like 17 different things. And it did partially work, but it was way over engineered. Right. And we spent, you know, weeks working on this and it was just sort of like. So then we changed tactics and was like, let's stop. Let's just solve small outcomes and start getting our feet wet, start getting comfortable.

Colin Slade:
And so the very first thing that we created after that was a little sentiment bot. It would monitor every single email and ticket that was coming into Zendesk and it would send it to AI, analyze whether the person was angry, neutral or happy. Right. And depending on what they were, it would add it to the ticket so we could track it for later and would put it inside of an internal note and then do something. So if it was an angry, there's multiple Angry tickets from the same person. It would send a slack message to ping someone that needs to go deal with it. And so that solved the problem that we were having, which was we couldn't see really quickly who was in trouble, right, who was at risk. And so that was sort of the first step and we started adding a few more of those.

Colin Slade:
So not just sentiment but what needs to be escalated. Then we started to try to solve the knowledge issue, right? So could AI try to answer the ticket? Could AI triage the ticket or the situation? And then we started to go a little bit deeper and say now let's download like all of our calls and turn them into sales handoff notes automatically for the onboarding team or for the account management team. And so that's really. So step one was just small, little, tiny, little wind. AIs getting used to it. Started showing the teams, they started to have better ideas. Because if you go ask your team right now, what do you want me to build in AI there? I. They're honestly, their ideas are terrible because they don't know what's possible.

Colin Slade:
So they give you like what's right in front of them. Oh, it'd be really great if you could help me write my emails. And you're like, okay, now our teams, I'm getting literally like full two, three page documents. In fact I could show one that came in today like I want. They want to build a risk analysis billing engine, right? It checks 10 different things and sends out emails and goes into our system and our billing system. Like that wasn't even a possibility for people to suggest like a couple months ago. But now that they've seen examples of things that can do a bunch of stuff, they have opened their minds and so you gotta show them the little bits. That was the first step.

Colin Slade:
Second step is we started to actually build workflows and automations using N8N. So multi step things like okay, now when a ticket comes in and we see sentiment, don't just read the sentiment, try to do something about the sentiment. Right. Um, then we started to get into more of an agentic, right? So there's certain use cases where this is where like you said the little gray area like everyone calls everything agents, right. But the way that I define agents is or multi agents is like the, the system needs to reason, it needs to make its own decisions. And in most enterprise business cases you don't want the AI to take its own decisions. It's too dangerous. Like, but there are cases where you can put very clear guardrails around it where you're like, yeah, I want you to make this decision.

Colin Slade:
Is this person happy or sad? That's an easy thing for AI to do. And if it's wrong, 2% of the time it's not the end of the world. It's not going to be detrimental. But like the other opposite would be, you know, you've heard all the stories about like the person was able to do code injection on a prompt and get a, you know, a Ford truck for free from the dealership. Right. Because they figured out how to prompt the agent that was reasoning and it reasoned its way out of the restrictions. So you got to be careful with that. But you go from real basic to basically agents and automations.

Colin Slade:
Then you go into some agents, like real agents making decisions. And then what we started now doing recently is you start putting those into different interfaces and building it into people's flows. So we've been building Zendesk sidebar apps. We've been building full on apps using Google AI Studio or using Lovable and we've also been building like you know, Slack bots, Salesforce interfaces. We are kind of now in the point where we need to figure out where do we put it to give the best value. And now it's doing multi steps and it's hitting APIs, it's hitting our systems, hitting all the other systems. It's calculating, you know, answers to questions from the kbs. But that's like way farther down the road, right.

Colin Slade:
Initially you just got to start getting wins and then your team will start telling you where you can solve some of these things to help them. Literally. I mean we're, we're saving thousands of hours a month now from a few months ago just because a lot of this stuff's being done for people.

Josh Schachter [Host]:
7,000 hours a month to be.

Colin Slade:
Yeah, something in that general range. Yeah, yeah.

Josh Schachter [Host]:
I mean that's insane. By the way, we didn't talk about this. What is, what's like the layout of your org size wise.

Colin Slade:
So we have. Yeah, so we're about, I think I just saw it, we're about 250, somewhere between 250 and 300 people. It's pretty much everything post sales. So I co lead the CS as of recently with Marshall Knoff has been a good buddy of mine for a long time. So he's got sort of the onboarding side and now he's recently taken ops. I have the support side, advanced support, training enablement. Now the AI Solutions team which was a CSOps Tech Ops team and then we also have an account management team with which handles renewals and relationship and then I don't handle. So one difference, there's some CS for some CS teams is I used to manage it.

Colin Slade:
But the CGM, we call them CGMs but our sort of more traditional CSM upsell team is on the go to market Org. So we're still building out things to support them, but it hasn't been our priority up to this point.

Josh Schachter [Host]:
So this is a large. I mean this is 250 in your. In the post sales. Org. This is just to make it clear to people this is not a 20 person team. This is at scale.

Colin Slade:
Yeah, no, this is at scale. Yeah. I mean we're getting, you know, somewhere in the range of 30,000 tickets a month. Right. Just on that side, just on the support side. So you can imagine that scales out to all the different areas and we're in 140 countries or 150 countries for our customer base and 20,000 plus customers. So with the. It's challenging.

Colin Slade:
There's lots of cultures, lots like that's the other thing too. REI had to not just handle English, it needed to be able to handle a whole bunch of languages and cultures and which is. Adds a different level of complexity.

Josh Schachter [Host]:
Sure. When I hear at scale generally I also start thinking about environments, technical environments, dev versus production. And oftentimes that becomes a bottleneck for folks at larger companies trying to do these types of things. Can you talk a little bit about how you, you know how that went for you guys?

Colin Slade:
Yeah, we. So it's always challenging. So I'm sort of like it's best friend and worst nightmare because in fact today they were asking me who created this random OpenAI API, you know, connection, which was me, but I forgot to name it. So they don't know where it came from. But it's in all kinds of places. And so I think it is challenging because I think from an IT perspective they want to make sure that all of our data and security perspective is secure and all that type of stuff. What we did is try to include them in the whole time. So they've been a part of this whole process and they were testing NNN at the same time.

Colin Slade:
You know, they've done some of their own applications and things like that. And so we've been in lockstep and at times I've made them feel very uncomfortable because this was all sort of a prototype. Right. These are all sort of POCs like will this work? Can we do this? Like can I get, you know, access to a pinecone database to try out this vector database solution. We never use Pinecone. What's their, you know, what's their, you know, security credentials and are they SOC2 compliance and all that kind of stuff? And luckily they were, so it was pretty easy. So we started testing as of recently because this has all worked so well. Now we went to more of an enterprise level nn.

Colin Slade:
Now we're locking things down, we're making sure the credentials are correct. But to be honest, for a long time we just, we're just trying to get things to work right. And so it's, I think if you can keep them in, in the loop, it's really good. But also I think you have to let, they have to be willing to let you take some risks and try out some stuff and connect to some systems that they might be a little uncomfortable with to get to those solutions and get it there. We were very careful, like, you know, if we brought in anyone on outside, they, you know, they signed obviously, you know, contracts and NDAs and all that kind of stuff to make sure it was secure. We're very careful what we send to the foundational models. But we also just didn't let that stop us. We just tried stuff and we just tested and like, we just tried to figure it out.

Colin Slade:
So I think to scale, you got to start small. And then when you get to the point where like, okay, now we need to really go to the whole company or we need to really go to large data sets, then you need to start looking at, okay, maybe we go off a pine cone to Snowflake, right, A much larger system. Like, maybe now we need to not have Collins log in as the API credentials. We need to have a set, you know, system credentials that virtual teams like you'd get there, but we're just getting there now. That should not stop you from building all these things. You absolutely can do it with still security but without having too much. I think the fear of things prevents people from innovating and I think you should be trying and innovating and then prove it that it's working and pretty secure. And then they'll, then they get more comfortable.

Josh Schachter [Host]:
So there's a little bit of like, I mean, there was nothing groundbreaking you just said there, Colin. It was. It's a little bit of cozying up to it, letting them know what you're working on.

Colin Slade:
Right.

Josh Schachter [Host]:
So this is not just some, some ran. There's a larger mission behind this. It's a little bit of asking for forgiveness rather than Permission and then just using your sensibility, having the right paperwork and documentation and, And I think you.

Colin Slade:
Need to have someone who knows a little. So I'm somewhat technical, right. I'm not. I, I was on my old days, I was a Java developer very briefly. Like, I know a little bit how to talk the talk. I know what their fears are. I know. So like, it does help if you have someone who's on your side who's like able to say, here's their, you know, security page.

Colin Slade:
It's okay. A lot of big companies are using them, right? Like, you know, if this huge company is using them to do this, we should be fine. That helped a lot. I used other, you know, examples and scenarios to be like, like, we're way smaller than this company. If they're going to, if, you know, gainsight or OpenAI or whoever is using this, it's probably like less risk than for us to try it. And so that helps to get it through the door. But yeah, no, it's nothing groundbreaking. It's just, you know, a little bit of what you said.

Colin Slade:
Ask for some forgiveness, push. And also once you start, once you start showing solutions and showing outcomes, it's really hard to stop that for the rest. Cause everyone is like, oh my gosh, let's go. Keep going.

Josh Schachter [Host]:
That's a good point. That's the best. There's the groundbreaking point, right? Just start to show the value and things will open up. You know, imagine that in a separate conversation that you and I had. You know, I got the impression that you came back to cloud beds and you, you basically transformed the, the Ops team, the CSOps team into AIOps and, you know, making sure that you had your, your roster of the different roles and responsibilities that you needed set in place. So talk through that. You know, when did that occur, how did it occur, what's the lineup that you have now, et cetera.

Colin Slade:
So, yeah, so when we, we had a CS tech ops team, yeah, they're handling all the normal things you would handle, right. So they were doing, you know, handling, creating QBRs from scratch. They were helping with different systems. They were running some bi. At times. They were sort of. We had, we had one AI tool called Forethought, right, Which is helping us with the selection, which we still use. And that's been a big, big piece of our, some of our solutions.

Colin Slade:
And so we had one person that knew how to use that kind of. But we're still refining it and so I think we got really lucky. But I think where you want me to go is sort of. I have this philosophy that I think you need a couple of different pieces to make it work. Right. And so I think you need a, what I call an AI visionary. It's usually an executive. They're sort of diving in deep.

Colin Slade:
They're like, you know, they're trying things at home, they're making things, they're, they're, you know, they're up to, they're. They're following all the people on LinkedIn that talk about all the latest models and the, you know, Sora and all that kind of stuff. Not because you're going to use all that in your business, because you can't use all of it. Sora isn't exactly useful for me other than I do. I do send some sort of videos to people to, you know, congratulate them for winning awards and stuff like that. But, you know, it's not really helping, you know, I'm not going to send a sort of video to a customer.

Josh Schachter [Host]:
If Sora were the, Were the cure all for customer success, then Gainsight would have the highest NRR and grr in the history of that. Because our cco, Brent Kremkes, is hooked on Sora and he's a phenomenal Sora, you know.

Colin Slade:
Oh, yeah, it's an. I spent a whole Saturday one time just completely. My wife was out of town, there was no one to keep track of me and I'm just doing, you know, I'd probably. I hit the limit. I didn't even know there was a limit, but there's 30, 30 or 50 videos a day. I hit that limit on that Saturday. So, yes, I think you. To get back on track, I think you need someone who, for a couple reasons, you need someone who can push through the budget.

Colin Slade:
If you need budget.

Josh Schachter [Host]:
To be clear, this is you.

Colin Slade:
This was you, this is me. In this particular case, does it have to be you?

Josh Schachter [Host]:
Does it have to be. Does it have to be a senior vice president?

Colin Slade:
I don't think. No, I don't think it has to be at the top level. But you need someone who. I think it has to be a director above. I think it has to be someone. It needs to be someone that's in these leadership calls or can deliver that information to the vp. They have to have a little bit of clout because I think you need to be able to convince a bunch of skeptics that this is a good thing to try, this is a good thing to move towards. This is a good thing to.

Colin Slade:
And I think it also helps to have that person as a little bit of a hype person. So I hype my team all over the place. I'm sharing what we're doing. I'm like. And that gets people excited and allows us to have more leeway, more resources, more credit in a sense. And so I think that's really important. I think the other piece you need is you have to have company experts, right? You need the company ethos, the company knowledge. It's really hard for me to go in and develop an incredible AI solution for a CSM if I'm not doing CSM work.

Colin Slade:
I don't know what their bottlenecks are on a daily basis. I might know some things, but I don't know the depth of their role and I don't also know how like cloud beds or gainsight or whoever does it. Right. It's a lot of tribal knowledge, a lot of tribal processes, a lot of things that have been done for years and usually those aren't documented very well.

Josh Schachter [Host]:
So, so this, that's the second architect archetype is the, the knowledge key master, let's call them.

Colin Slade:
Yeah, or you probably. There's probably a couple of those, right? This like. So I think you need to have a, you gotta have the visionary, you have to have this knowledge expert, right? This person that knows your company probably is a few people, depending on which areas. And then you need the operators, right? Ideally you have a technical ish type of operator. And when I mean technical, I mean they know a little bit of like either JavaScript or HTML or they've done some, you know, maybe they've done a little bit of Python or they know at least how to prompt Claude Code or ChatGPT to help them kind of massage their way through the technical pieces. And then you have just generally a non technical AI operator, AI implementer, whatever you want to call it. So I got really lucky because the team was made up of. And then there's probably a final one which is kind of the organized type.

Colin Slade:
A person who sort of keeps everything on track, right? Which isn't one that you have to have, but it does help to have someone that can do that. So in our case we had a manager, her name is Olga. She's now left to go do AI in opposite a different place, which is awesome for her. But she helped kind of set the stage for. Okay, it was her team, it was her team that I took and said, hey, we're going to learn NN, we're going to learn AI. I want you to dive into YouTube videos and check out Google AI Studio and we're not just going to use Gemini, we're not just going to do ChatGPT GPTs, we're going to like actually build agents and applications and automations and we're going to do it at scale. I'm going to focus on the outcome. And they were all excited and we just got lucky.

Colin Slade:
Like we have Gabrielle on our team. He was doing bi, he kind of taught himself Python. He'd been doing some level of this but not in the AI side of things. And so he helps us with our more technical stuff. Then I had arena and Monica and Bell, they are more of the AI operators and so they literally now every one of them can build an incredible workflow in N8N or Google AI Studio or whatever you want to do it in within hours, if not day, you know, days or hours versus months. That would take most people and you should just hear like they just talk in solutions now they don't even talk in like what we used to talk about, which is like we have no idea what the hell we're doing. How is this possible? What's an API like? Now they know exactly how to do it for the most part. And then there was the one piece that we brought in help.

Colin Slade:
We hired a team, a small little like two person developer team out of Brazil to help us with our rag system or like knowledge system that's fairly complex. It ingests from like 12 different systems. It's getting 95% accuracy. We built a lot of it, but we had them help us with like optimization and token usage and you know, refining our, you know, air. Getting it to that 94, 95% was huge. And now that's what we're launching to the whole company where you can ask any question, whether it's about product, process, people, Internet, like anything on the Internet and it will be able to not only give you an answer, it will analyze all the different sources and make decisions like, oh, I see that was mentioned in Slack and it was in your KB and it's in a Google Doc. But I know which of these to prioritize and I know which one is the latest. And so I'm going to give you a full fledged answer that you can give to a customer or that you can use for yourself.

Colin Slade:
And so that's kind of how we did it. And we just were like everyone was trying to learn as fast as possible. We had a CSAI Champions channel in Slack and we just post. I did this workflow, I have this prompt, I just figured this out. Have you seen this? And we just kept adding to it and getting to the point where we are now, which is now this team is turning into an AI solutions team. I'm changing the name next week and they will basically we're going to have a group of champions, like a task force from every department that will sort of dotted line report to this team and we're going to train the trainer when it train them how to use these tools in their world. So you know, in sales or in marketing or in finance and then we're going to basically kind of make sure they can't break stuff and baby steps, you know, same, same process we went through but they'll have guidance and now we've built a bunch of building blocks that we can just apply to all these places and so we're really excited about that. It's going to take quite a bit of coordination and you know, but all the teams have basically I got all kinds of people wanting to be a part of this and get access and try to build and it's been pretty cool because they've seen what we can do and what they can do and so now I think it's going to become a company wide thing.

Josh Schachter [Host]:
So just to be clear, this task force is a cross functional. This is basically folks that are going to come in and lead building out workflows for their respective function.

Colin Slade:
Basically. Yeah.

Josh Schachter [Host]:
Taking from the lessons and the stories of what you guys went through.

Colin Slade:
Yeah, I mean it's a mix of people. It's like directors and some teams, it's you know, ICs and other teams. It's less about the role and the level, it's more about the willingness to, to learn and to push. And these are all kind of the creative builders on their teams.

Josh Schachter [Host]:
It feels like you established some FOMO at cloud beds of folks that were like, for sure. You know, maybe you started out as like the geeky group. Right. That's just like working on this on any den and all these things. And then now it's like, oh wait, we need to get in here. This is, this is the future of the company, you know, you know, absolutely.

Colin Slade:
FOMO for sure. I mean in a good way. Like we want there to be fomo. We want them to be successful. It's awesome. We're really excited. But yeah, 100% we have proven it now they all want to be able to do it. Especially the amount of time and efficiencies that we efficiency we're gaining and time we're saving is just, they all want that.

Colin Slade:
It's just awesome and we want to be able to help them do it.

Josh Schachter [Host]:
You mentioned a couple of archetypes there. You mentioned the visionary, that was yourself. You mentioned the folks that have the knowledge, expertise of your own processes. Basically like the veteran CSM type folks, right, that also can think about these solutions. You mentioned the, the builders, the operators are actually, you know, doing all the vibe building and then a project manager, so to speak. What percentage of folks time was allocated to these efforts.

Colin Slade:
So for sir, the knowledge one is less. And that's also one thing. The knowledge worker, the, you know, company knowledge that was like a, maybe a couple hours per week as necessary, right. Depending on what we needed from them to figure it out. The operators full time. I said clear everything. We dropped every other project in CS tech ops unless a system went down, like a system of truth went down, source of record went down. We were only working on this for the visionary.

Colin Slade:
Myself. I worked on it as much as I could because it's really much funner than my normal job. But generally speaking it was probably, you know, 20% of my time, right? So, and then the. But the operator side of it was very much like a 90, I would say 80 to 90% because they still had to do some things that, you know, they couldn't drop everything. But as much as possible I had them drop it. The other thing we did do that I didn't talk about is one of the most important things about all of this is you have to have enough documentation for the AI to be useful. If your company has zero documentation or it was not written at all for AI and it's basically unreadable or unusable because they're lacking metadata or lacking context, you've got to fix that first. So thank goodness our training enablement team had been using AI to help it change it for AI, basically.

Colin Slade:
So we do a technique that I just learned about, we call it bouncing. Where basically like we go to each of the three to four models and say, tell me, you know, OpenAI or Anthropic or Google or X or what do they call it, grok and basically say, hey, I need you to take this KB article or this documentation and make it a five out of five for you to read it and then you bounce it between the different models. So if it says it's a three, you say rewrite it to be a five. Then when it's a five, you go to the other one and it says it's a four. Okay, rewrite that to be a Five. And until it's all five out of five, we don't say it's completed. And so that has meant that we went from literally this is, this is.

Josh Schachter [Host]:
Shaping those articles not only for folks that are going to the website to read the articles, but this is shaping the articles for. For the AI to interpret.

Colin Slade:
Yeah. And it's shaping it for both AIs, right. Our internal AI that we're building and all the big AIs out there. Right. So.

Josh Schachter [Host]:
Oh, interesting. This is a marketing technique as well.

Colin Slade:
Yeah. So like people used to get to our KBS, less than 2% came from like Google and the, and the, you know, gen gen AIs. Right. So the chatgpts now 80% of the people who get to our KBS articles come from ChatGPT or Google. So like it completely because you've optimized the articles for. If you search for anything for us, we are optimized for them. Right. Like we built it for them.

Colin Slade:
So 100%.

Josh Schachter [Host]:
I've never sworn on this episode, on this program since, since joining Gainsight, pre Gainsight, lots of. I was George Carlin of podcasting. But. But no shit. You had like 2% traffic to your KB articles for. From search and from LLMs. And now optimizing those help articles through the LLMs, you have 80% of your traffic coming from those sources.

Colin Slade:
Yeah. And like a combo of Google like so what everyone used to do is to go to our website or to go to our application to get our KB articles. Right. Because if you went to Google or to the ChatGPT, it was hit or miss. It wasn't written like it wasn't set up for it didn't have good metadata, didn't have good tags and labels. Now we're built for that. So now everyone just goes there because it's way easier. You get a better response.

Colin Slade:
It's combines it. It knows exactly what you're looking for. And so yeah, it's shifted like it's crazy. It was. It's a complete like flop now very few people go to our website or the app. Most people go through, you know, Google or Gemini or ChatGPT. Wow.

Josh Schachter [Host]:
How do you measure that though? How do you know? Because, because isn't. Isn't the LLM just feeding the information right there within its own application?

Colin Slade:
Yeah, because your KV article housing places usually have the source. So we can see whether they came from our app or our website versus coming from the search engines or coming from. So not inside of our apps but inside of our knowledge base tool. We can see where the source was. Yeah. So where did they come from? And it shifted about six months ago from like barely any ChatGPT or Gemini or maybe maybe eight months ago to almost entirely coming from that.

Josh Schachter [Host]:
So you're truly meeting users where they are trying to.

Colin Slade:
I mean it's always a work in progress. Like we had a meeting today to try to continually improve that and it's not whether we've solved it all. So like for example, everything we have documented we are really good with, but we're also missing a lot of documentation. Right. Because we have complex parts of our application that no one understands exactly super well. And so it's all in our product managers or advanced support teams heads, it's all tribal knowledge. And so we're coming up with strategies of how do we get that out of their heads into raw content and then put that raw content into our KBS and internal KBS so that the AI can read it and use it to help our teams. Because, and again that's to the point I made earlier.

Colin Slade:
Like if you have zero documentation or very bad documentation, it doesn't matter how good your AI is that you or agency build or automations, you will still fall very, very far short because you, it, it, it can't, it's not a magician, it's pattern recognition. If you give it patterns, it will do a good job of reading it and responding with some training. But if you wanted to make it up, it will make it up poorly. Like the hallucination is real and so you don't want them to be hallucinating on your more complex areas of your application or your company or whatever.

Josh Schachter [Host]:
Yeah, and what's the process that you have in place to make sure that the content is kept up to date?

Colin Slade:
So yeah, it's a semi manual, semi AI driven process. So anytime there's a release, it triggers a process where Basically the, the AI goes and checks all our current KBs, internal and external. It checks this. Basically it's got a whole prompt that we built plus some other tools. It checks to see what needs to be updated. Right. So what's changing? What are the, you know, the dependencies, all that type of stuff. Then it will try to change it, send it to a human to review to make sure we're not missing anything or didn't make up stuff.

Colin Slade:
After the humans reviewed it, then they'll do this bouncing technique right, to get it into a super high quality for the AI models and then they'll publish and usually they'll publish like right before the Actual release is going out and now we've gotten to the point where we were doing about 10 KB articles a day. Now we're doing hundreds a day.

Josh Schachter [Host]:
Okay, we've covered a lot of ground here, but barely any actually. There's a lot more I wanted to get into. I'm not going to make folks sit here and listen to an hour and a half episode. I do want to talk about a couple more things. So you've built, built, built, built, built, built, built. Do you think you'll ever come back and buy so?

Colin Slade:
Yeah. So while we've been building, the reason we built is because we couldn't find anything that wowed us for the cost and for the value. And we've been able to build ourselves, but we have done like 20 PoCs with all the major AI platforms and all the agent, a lot of the major specialized platforms. And so we've done it with anthropic enterprise and DevRev. And you know, we've even recently started looking at the game side. We will definitely buy so the dis. At some point in the future it might be a specialized buy. So for example, we use forethought.

Colin Slade:
Forethought does a really great job for us on the front end of deflection. For me to create that would be really hard, especially with all that we've put in and I don't have to maintain it. And I will say that one challenge with the build is that you do start to hit points where you have to maintain all these workflows and things that you've built. You have to scale the servers, have to get bigger. Your OpenAI API, you know, costs start to get more. So there are challenges to sort of navigate that. So because of that, at some point we might need to go to an AI platform and there's a few that I think are really good out there. At other points we might need to go more for a specialized like, hey, we don't want to handle the specific customer success piece of this, right? The signals and all that.

Colin Slade:
Because I'd have to build that custom. Let's do, let's go with someone who's an expert in that area. That's I think, where we probably will short term do. And that's what we've done up to this point. But we're very open also to be wowed by a platform that says you can come in and move all of your agents and things over here and still do some other stuff that you really need to do. So it's definitely not a I won't ever. We won't ever buy. And we've now proven that my budget has gone way up in what I could spend.

Colin Slade:
Right. Because we've proven the value. So I had zero budget eight, nine months ago. Now I have a much larger budget that I could use, but we haven't found anything quite yet with the timing and all the pieces and the funding and also all these AI companies. And I'm sure you all are still trying to figure out how to price this stuff. Right? The AI actions or the AI calls are still a funky gray area. And everyone's sort of, you know, some people are really good at knowledge part, some people are really good at the agent building part. I haven't seen anyone that's incredible at both necessarily, although there are some good platforms out there.

Colin Slade:
So, yeah, I do think we'll buy at some point. Maybe not in the short term, but I'd say in the midterm it might be. I'm probably more likely to be a specialized buy for certain use cases that we don't want to do or don't think we'll do as well, but might be an AI platform if we get wowed. Yeah.

Josh Schachter [Host]:
I want to leave us on this last segment here, this last section of our conversation, which is the impact that your efforts have made. There's a lot more to it that we didn't Discuss. I mean, 147 other workflows that we didn't discuss. Right. Heck, we didn't even discuss the, the, the, the dashboard that one of your teammates, you know, Vibe coded over the weekend to, to calculate to Belle for that one.

Colin Slade:
I told her not to do it and then she did it anyways. It was amazing.

Josh Schachter [Host]:
Yeah. And it tracks, it tracks all of your agents and shows the ROI and the usage and the performance metrics across all of your agents and how much. Incredible, right? We didn't even get into that. But before this AI revolution that you and the team led, your team was at 120% capacity, overflowed. They were burning out. They were spending 70% of their time on administrative tasks. There was that fragmented customer experience. You had tools that were sitting partially as shelfware, like you mentioned Gemini afterwards.

Josh Schachter [Host]:
Now you've had the same headcount that you started with, but you're serving 45, 40 to 50% more customers. 75% of routine work has been automated. 70% of time on relationships and strategy. I'm reading off something and I don't know where the rest of that sentence is, so there's something about 70% of time. You've the, we're not going to call it a knowledge base, but that knowledge hub that you built is 95% accurate in the Q and A responses it's able to field for people. You've had 75% of your support interactions have been deflected by AI and you've saved about four and a half million dollars on the cost of, of tooling and headcount, quite frankly.

Colin Slade:
Right.

Josh Schachter [Host]:
Of, of what the equivalent would be.

Colin Slade:
Yeah. And time. And time saved. Time saved, yeah.

Josh Schachter [Host]:
Time saved.

Colin Slade:
Right, right, yeah. And all that. Yeah. Altogether probably somewhere in that range. Yeah. By the. I would say that's an annual savings. Right.

Colin Slade:
Not like up to this point, but by the time we get a year in, that's probably where bubble would be. Yeah.

Josh Schachter [Host]:
And then 7,000 hours saved per month. And you know now again like you said, like your, your deployments of new workflows, agents, whatever we're calling them is daily, weekly. It's not a matter of two or three months as you roll out and you, and you've created this culture clearly with this new task force, this cross functional piece that you just mentioned, what stands out to you most? What are you most proud of?

Colin Slade:
That's a good question. I'm most proud that we've actually been able to really prove value, I think and I'm most proud of the team. Like I had, we had a moment where I was like, you just all need to clap for yourself because like literally what you just told as an Update today, like 10 seconds each. Right. Was what most people would love to have just this one agent for the course of a quarter and that team is busted out. So I'm super proud of the team. I'm also very thankful that the, you know, the company and our leadership has been willing to let us play these things and IT team has let us kind of muck it up and you know, like we wouldn't have been able to do without it. But I think the probably the biggest thing that I'm most proud of minus the team part is that like we have really produced real value.

Colin Slade:
Like it's not just talk, it's not just a poc. It is in people's worlds getting used right now. We just got stats today that are, we just rolled out a brand new like ticket answerer. Right. So it tries to answer for the customer and we got 50% usage in the last couple weeks that we rolled it out. So like that's what I'm really excited about. People are really using it, it's helping them in their job, it's helping them not be at 120% and hopefully it's getting them confident to do more and be better in their career. And so it's just fun to be able to say yeah, we are making a real difference and AI is real, it's not going away.

Colin Slade:
Sure there might be some level of bubble in some capacity but like the value that we can provide or that anyone who's using AI at scale can provide to an organization is just crazy.

Josh Schachter [Host]:
You know, we started this conversation and I was thinking we were going to instill some, some fud, some, some fear, uncertainty and doubt in our listeners of just feeling overwhelmed. The work that you've done has cast such a large shadow potentially on so many people that just are trying to figure out how to get involved. But my hope is that we did unpack a lot of what your effort was. There's still a lot more to unpack but that folks have a general sense of how you went about it. The fact that you started out in very similar footing as what many people out there are in that are, you know, fearful of really leaning in like you did. And so ultimately I hope that this is an inspiration for folks that are listening out there. Share it with your leadership, share it with your elt, share it with your peers at other companies. Share it with the folks that work for you.

Josh Schachter [Host]:
I really think that this episode is very important that telling your story, Colin, and the story of your organization is super valuable again to set that example for others in the industry to follow so that a rising tide can lift all boats here.

Colin Slade:
Yeah, and I, and I'm available to help. I, I, I'm probably gonna put you some type of community for us, for CS leaders and CX leaders to try to figure this out. I wish I'd had people to talk to. There's a few people that I could have talked to and we did talk at times but for the most part like this is really hard. You're being asked to make hundred to million dollar decisions for multiple years and if you make the wrong decision, your job's on the line. Right. And so it is not easy. It is very stressful and I always, I feel overwhelmed and behind too.

Colin Slade:
So if anyone needs some help or wants to know how to navigate, I'm absolutely happy to, to connect and like I said, I'm going to probably put together some kind of community where we can learn from each other because I want to see what you all are doing. You know, we're just trying to figure it out as we go. And hopefully I can learn from you, you can learn from me, and we'll see where we go from there.

Josh Schachter [Host]:
Colin Slade, senior vice president of AI Strategy and customer success. Congratulations on the new know part of that title as well on the AI strategy piece. That's really exciting. And that's the future of of a lot of executive leadership, too, right. That that role doesn't exist right now in many companies, but it will. So you're ushering that into and who.

Colin Slade:
Knows what the role exactly means. But like, we were trying to figure it out, you know, so that's why I put strategy in there, which strategize what the heck to do with AI.

Josh Schachter [Host]:
Well, thank you for sharing your story.

Colin Slade:
Thanks. Jo Sa.