From Data to Action: Build Your AI Retention Play

[Speaker 1]
Thank you very much for being here. Yeah, and like, so I posted a month ago about my newborn, Olivia, and actually LinkedIn is a little bit broken. So like got way more posts and like way more engagement than any of my other posts.

But do you guys know who the three people are that actually shared my delivery post? Like I'm still looking for three people that have the guts to actually re-share and re-post one of my most personal stories ever. So like maybe you can help me in the discovery.

So my name is Mick Wiers, I'm from Amsterdam. Wiers is very, very difficult to pronounce. So Chana, you did really, really good.

And today I have the pleasure of hosting an interactive workshop. So we only have 45 minutes and I'll try to do the least amount of talking I can. It's a full house, so I'm really, really grateful, especially after Chuck and the whole video production, we're gonna focus on collaborative learning.

But it's all about from data to action today. And the timer is already counting down. So why am I here?

So I founded Customer Success Snack five years ago and we do over 40 events in Europe in all the big like tech hubs from Amsterdam to Paris to Copenhagen to Munich. And I wanna bring those learnings here today because we always do round table discussions. But next to that, also we do an annual AI research study.

Currently, you might have already got an invite either via LinkedIn or via email to participate because the state of AI is just always changing and we wanna be the independent benchmark for that. But mainly why am I here today? It's because of the third one.

So over the last, what is it, 10 months, as a consultant, I built three to five digital post sales orgs. We're currently still building two. And I wanna bring those learnings here.

And four, also a little bit about customer excellence, but I really wanna focus on point three. So 10 years ago, I did a data analytics course. I think it's been the most humbling experience in my life next to trying to learn Chinese.

So what I did learn is that I really have to respect data engineers and people who actually understand AI modeling because everyone can vibe code. So how many people are using Chachibitty, hands up? How many people are using Claude?

How many are using our Breeze? How many are Atlas? Boom, everyone.

What's the danger of all that? You said something? They're collecting data, yes, that's dangerous.

But what's the danger if I steer my strategy based on their recommendations? They tell you what you wanna hear. And the word I'm looking for is false positives.

Like we can just steer, I can create a full deck, a full presentation to my CEO on why we need to change everything right now. So what did we learn? We learned that language is a tool and common language is the most important thing.

So that's what we're here today because we don't have a data problem. We do have a data hygiene problem. But we do have a problem that everything you guys want and it's in your head just doesn't get translated to the people who need to build it and execute it, the engineers or your data models.

So it's all about getting a shared language. So you've seen all kinds of like, like what is it, worksheets on the table. Like I challenge you for the next 37.15 seconds to not be on your phone. Please don't go on your computer. We go old school. We go whiteboards and markers.

So the faster I talk and the quicker I can get you through the boring stuff, we can go into a workshop. And afterwards, I will invite a few of you to actually come to the stage and to share your main learning or your main finding of this half hour on that whiteboard. And you all know me or some people know me, I like to reward people.

So for the table that actually is the most engaged and I see writing the most, I have some prizes there. So let me tell you what we're gonna do today. You're all conductors.

Like we're on discovery. You're all have agents around you, but we need to avoid false positives. So it's all about languages and really sharing what you need, why you need it and how you need it.

Your AI or your engineer should not have to interpret what you want. You need to be able from A to Z to really describe what you want, how you want it. So it's all about writing up as clearly as you can.

So I hope you guys can still actually write with a pen. So like the problem is we have the data, we have the dashboards, we sort of get signals. However, translating signals to actions, it's where there is a little disconnect.

That is because we go from data to context and insights to action. But every time I talk to the data engineers because you're constantly building agents, we have eight back and forths because they don't understand the context. They don't understand the insights and they just build what I believe I need, but it's never what I actually need.

So you need the ping pong. You need to have chats with other people. So every time we host an engagement like this, we have one success metric.

And that is that after today, I want you all to meet someone on your table that you think, ooh, next week I wanna call again because this person really had a good topic and that there's one thing that you learn from this session that you wanna implement with your teams on Wednesday. So just a little bit of context before we can go into the exercise. So far, I'm really scooping some time off to have more of collaboration.

So what is a weak and what's a strong signal? Like who actually is in CS operations in this room? Can I see some hands?

Are there any data engineers? Ooh, one. Do you mind sharing what you believe is a good signal definition that people came to you?

So I don't know, I'll just quickly grab the mic. Could I have the mic? So we are here with 99 CS or CS ops, but only one data engineer.

And how do they often communicate with you? Just test.

[Speaker 5]
The best way is just to chat me because I hate the males.

[Speaker 1]
And then a one follow-up question. So how many times is their first ask good enough?

[Speaker 5]
From three up to five, I think.

[Speaker 1]
Thank you very much. So like there is weak signals and strong signals. And that's the question that I want you to answer.

To actually focus on during the exercise. What accelerates a good signal? As I said, your data engineer, your AI tool cannot fix a meaningless signal.

Like in the end, we need to be very concise and very prescriptive about what we need. So we don't look for perfection, but we do look for longer sentences, smart goals. What's the benchmark?

What's the threshold? So in the end, we're gonna talk about signals. And I don't know if you look, please have a look on the worksheets, but there's two you will find.

So we have one bonus exercise. That's big white canvas. And maybe we get there, but I want you to have a look at the other one, at the journey map or at the discovery or the treasure hunter.

Because you need to start somewhere. And I'm not trying to push you to get to perfection. I'm just trying to get you started, trying to get you build the same language that everyone in the go-to-market or in the total organization can talk about.

So enough talk and let's go to action. I'm gonna quickly introduce the exercise and afterwards, I will be the timekeeper. I might walk around a little bit.

Just know that time is of the essence. So I do ask you in one table in the next 30 seconds to decide who of you is gonna be the main note taker. I'm just gonna give you like 30 seconds right now.

Please, among yourself, one volunteer. It's about who is gonna be the main note taker. So there's one thing I hate, that's interrupting people.

So I'm only gonna do it twice today. This was once, but I do need to actually get you through the exercise and then you can start going. So I created this scenario, a hypothetical scenario that's also on your sheet.

But please, if someone dares to step forward, make it a realistic scenario. This is a safe space. I don't want to see any AI note takers on the table, like for real, no note taking, like only in your brain because it's safe and I want you to be able to actually share your relevant customer case.

If you don't have one or if you wanna use the example, please use the example, but make it hyper relevant where you can. In the next seven minutes, I want you to focus on step one and step two and I want you to focus on making the longest sentence as possible. Don't focus on bullet points.

I love to communicate in bullet points, but my data engineer actually always says, Mick, what do you mean? If there's room for interpretation, you're not specific enough. Afterwards, I want you to focus on step three and four, another seven minutes.

Afterwards, I want you to focus on step five and step six and this way of thinking actually builds the framework that you need. So what's the signal? But the signal is nothing without a threshold.

What's the context? Because the signal without context is nothing, like what's their outcome, what's their objective, what's their success goals? The insight, this is what lives in you.

You have the outside in perspective. You are the benchmark of your company and then turn it inside, hopefully into AI assist. And if you can crack this code in Q4, then I know that the autonomous action will come for you in 2027.

So let's start with the exercise. The first seven minutes goes in to focus on step one and two and please, you can stand up for this exercise in your table. I'll be timekeeping and walking around, thank you.

So there are some very interesting discussions. So I'm very sorry, like I really don't like interrupting, but remember what we just saw in the presentation from Grant Clark, sorry, the new GM of Atlas and Atlas is basically the game site agent. And do you remember where the agent said, hey, there is a handoff or a guardrail and that's actually what you're writing up right now.

Because you can have a signal and a threshold, but if you truly fly blind on a signal and a threshold, you're gonna actually like get off targets very quickly. So your second brain or the context of the organization is what you're writing up right now. So what's the contact, what's the insight and what do you actually collect as a domain knowledge as your expert.

So this, you are the filter before you get to AI insights or AI action. So I give you 30 more seconds on step three and four and I want you to make something up. Like it's not about perfection, it's about getting it done.

So it's time to actually go to the last part of the exercise. And I'm very sorry, but try to simplify step five. Try to think in, I want an email generated or I want a webinar to be created or a snippet to be created.

Don't go complex, try to say a little message or a different content type or a nudge. Try to simplify for you and your table. So I also challenge you to write up step seven and eight.

So who should own this? I want you to write up one job profile. One job profile, seven and eight, one job profile.

So please also write up seven and eight. So I'm very sorry, I have to slowly wrap up. Please also write up in step seven and eight, the one tool where it should live and the one job title that should own it.

And like, I'm just looking for three people that would like to come to stage and to actually share what they took out of this discussion. Like for the brave souls who like to volunteer, I got a little prize. And it's all about collaborative learning.

We're a safe space, but I've heard amazing discussions. So who would like to come and share their main learning of the table with me? Like I know that there will be someone.

Otherwise I need to pick someone. So please, can I ask for some brave volunteers? Who would like to come to stage?

So Mounir.

[Speaker 4]
Yes.

[Speaker 1]
And Alyssa and Harry. I'll be on the board and just writing out your main learning. And please hand over the microphone to the next person indeed.

And like one big round of applause again for all of your efforts. Thank you very much. Thank you.

So like, Mounir, I need to be here. So what was the main finding of going through this exercise? Like you don't have to go through the full eight steps, but what did you learn from having an interaction with your peers?

[Speaker 4]
Yeah, I mean like we all like sort of had the understanding that multiple like signals of those are linking to each other in a way. So like for example, the key user activity dropping significantly. The champion role changes and the increase in support tickets.

For example, these all together can link into, for example, drop in adoption in concept or maybe lack of champion handover, lack of a champion enablement plan that should have been in place. And also the major signal was of course the three times increase in support and et cetera, et cetera. I can then give it to you.

[Speaker 1]
Thank you, Mounir. So in the end, so what the main thing that you learned was that all signals are connected and that you need to measure signals against something. And you said, okay, we're gonna measure signals against a leading indicator that might be increasing adoption and then a lagging indicators and RGR.

Thank you very much for sharing. Like a round of applause. And like please don't leave.

[Speaker 3]
So we had a few assumptions that we made for the scenario. We said it was a high touch client within the pharma industry. The main thing that we really took away is the main action.

It's kind of a medic perspective, but the change of the champion without knowing who that new advocate is inside the business, it's hard to really dig into the macro and the micro situation at the business to understand why adoption and usage has gone down. So we need to really identify who that next successor is and partner with them on understanding the business.

[Speaker 1]
And like in your case with high touch pharma, you said like the champion leaving. What was the action you wanted the AI assist to do in this flow?

[Speaker 3]
Yep, so we said that we can run an agent on a weekly basis to scrape the pharma industry news and understand if GSK had a drug failure, then that whole department might be depleted or move to another business or another business within GSK. So just understanding how that industry news can affect what we're gonna do going next. So it's whether did the champion go elsewhere or is there a new opportunity for sales or where did that champion go and what's their successor for success?

[Speaker 1]
It's all interconnected.

[Speaker 3]
It's all, yeah.

[Speaker 1]
Like one quick question, who in this room already has this agent running? So I believe this is one of the things like you can either do it with or without LinkedIn, but the daily news uptake and translating that into insights also for your domain knowledge is a really good insight.

[Speaker 3]
We have staircase as well. So this can help with that as well.

[Speaker 1]
Yeah, thank you very much. Like, I love your CS Ops badge. I don't work for Gainsight.

Barry.

[Speaker 2]
Yeah, so we kind of realized quite quickly, I guess in our discussions that we all have different tech. So it's quite difficult to sort of pin down one size fits all strategy around this. So we kind of, I guess I volunteered my company since the numbers were negative and they kind of represent our current state.

But yeah, for all 10 offline. But we sort of landed on key user activity as the sort of the main driver here. So what you both have pointed out already, it's all connected, everything's connected.

Sounds like a Christopher Nolan film, but it's true, is that it all leads back to a root cause. So there is a chain effect that starts from somewhere, right? For me, I was drawn to the miners and the logins, but given what Gainsight is saying, what other people are saying is that we're moving away from this sort of seat space model into everyone can have access.

It is even more important to understand who your key user base are. So when you see a big number like that, the red flags start to kick in, but it is not really that realistic to kind of say as soon as there's smoke, there's fire. So you don't say as soon as the number drops to 25%, suddenly you've got to run to data analytics and to ops and products and go, everything's broken, we need to fix this customer.

They're not high risk, sure. But if you start like recession, you don't measure recession as soon as the economy drops 0.1%. It's over a period of time. We landed on based on six months from renewal, roughly two weeks or so, you see sustained.

They might be sick, they might be on holiday. If they don't come back, that starts to be a problem. So we looked at kind of across the board there towards the context and insight.

And then ultimately landed on the AI. I'm looking at your clock there. The AI sort of augmenting and assisting in terms of analyzing the user behavior.

So one thing in CS that's missing maybe a lot in companies, you don't train on the soft skills, you train on the hard skills, you train on the software. You become product experts, not people experts. So you want the AI to analyze and across the web, do all that work for you to understand who you're talking to and who you should be talking to.

[Speaker 1]
So I don't want to interrupt you, but like I'm just looking into this room and we just had like an amazing speech. Like you're in the one size fits all doesn't exist. You say usage is the main driver, but are we not also moving away from usage towards consumption or outcome?

So is it not very, like are we measuring the right thing?

[Speaker 2]
Consumption requires access.

[Speaker 1]
Okay.

[Speaker 2]
So regardless of whether you are a seat or a person in the system, you still consume when you access. With the AI credit system right now, every time you consume, you are accessing the system to consume. So the same logic applies in my mind, at least anyway.

But yeah, as far as like the AI agent, it should just scrape, find the information you need to make informed decisions, but give you as the human, the control over how you communicate the relevant information that you've acquired. So if your champion is leaving or changing role and a new one's coming in, find out who's leaving and who they are as a person. Meetings, calls, emails, understand the person, the profile.

You augment that information with your own communication and personal style.

[Speaker 1]
Big round of applause, Harry. Thank you. I have one more minute.

So I'll leave you on stage. So like, we have a lot of these worksheets here because we, yeah, we accidentally printed like a hundred. So if you want to grab one after the show and you want to do this exercise with your team, please do.

But otherwise I would say, make a quick photo of this one. So now you can grab your phone and either you use this one or you rip something out or you change it, but try to have these kinds of exercises with your team. Sorry, I'm in the way.

Because I've seen these frameworks in different ways. Like you've seen data, context, insights and action. That was too simple.

Now, maybe this one is a bit too complex, but I'm also still trying to find the best way to bridge the gap between what we need and what engineering needs to build. Thank you for your insights. Like the other thing that like, I would love it if you could all add me on LinkedIn.

Like, I would love to see your journey. Like, thank you very much for your time. Like I do ask if I could have one or two extra minutes.

So like, I want to give one round of applause for the volunteers. Thank you very much. I just want to explain the other exercise and why I think it's really important that you do one with your team.

So we don't have enough time to do it. So take a picture of it. So why is it so important?

It is because everyone talks about the AI hub or your core. And then it's around mapping your tech stack. So where does the context live?

Where does your playbooks live? Because if I now ask Entropic to create a CS strategy, I get generic fluff. And like you have your secret sauce.

You have your years of experience. So your secret sauce needs to live somewhere. It needs to be accessible by the agent.

Otherwise you get like generic fluff. So try to map out the tech stack. And what we have seen is that in many organizations, you actually need 10 of those connected, but you only have three or you have four.

So in very little organizations, I have seen a tech stack where it's all connected, where it's hyper relevant. Because if you want an agent to build this, it's all about your data foundation. So if you don't have the tech stack connected, start there.

If you have your tech stack connected, work on data hygiene, work on data foundation, because your foundation is gonna make the difference. Like, thank you very much. And like, yeah, like one round of applause for yourself.

Thank you. Thank you.