If we were to rank the top paradigm shifts in product in the last 10 years, I'd reckon AI and product -led growth would sit at the top of that list.
Both concepts really come down to amplifying efficiency, doing more with less.
But the real cheat code is bringing them together and using the time -saving power of LLMs to enhance your PLG strategy.
In a recent panel event, How to Use AI to Supercharge Product -Led Growth, we brought together three awesome PLG experts, Bramley John, the founder of DelightPath, Danny Grant, the CEO of Jam .dev, and Anuja Diah, the author of Growth Hacking for Dummies.
We got the three of them talking about the ways product teams can leverage AI technology right now to boost every phase of the user's journey.
What I loved about this event was how each panelist's expertise came together in such a complimentary way, offering lots of practical ways to adapt and adjust your growth strategy using AI tools.
And they were just super -nice people.
Let's jump in. So today's session is going to be focusing on how to use AI to supercharge product -led growth and we'll be speaking with some amazing voices in the space.
We have a really exciting lineup today.
I'm really excited to introduce them.
We've got Ronley John, who's the author of Product -Led Onboarding.
He's also a renowned expert in PLG strategy.
Ronley brings a deep understanding of how to turn prospects into passionate users.
We've got a little jeopardy question for Ronley today.
So Ronley, you've been called the onboarding wizard by some of the biggest names in SAS, and your book Product -Led Onboarding has been a game changer for countless product teams.
We're really just pumping your tires right now.
If you can wave a magic wand and instantly fix the most common onboarding mistake you've seen, what would it be?
Thanks so much, Hannah.
I would say it's actually not related to the product.
Often the biggest problem around onboarding is internal friction, not product friction, and what I mean by that is product that's the product thing inside of the product, marketing that's their onboarding emails, and then customer success that's their customer onboarding thing.
And they don't talk to each other.
This is what happened while we're even working at an onboarding company called AppCubes, where a product adoption software.
And we had that same issue.
So I think if I could wave a magic wand, it's if I can get the teams to talk to each other more and agree what success is like for the user, which is a hard problem, that's where I would start.
Man, siloing is just one of those perennial issues.
Well, thanks for that.
We also have Dani Grant joining us.
She's the founder of Jam .dev and a former product lead.
Dani is known for using product data to design seamless, high -impact user experiences.
And she's also just such an awesome person.
So Dani, thanks so much for joining us.
Question to pass to you.
So Jam has skyrocketed to 150 ,000 users at 32 different Fortune 100 companies, which is incredible.
But you're also speaking at events conferences and you're active daily on LinkedIn.
Honestly, I'm jealous of how you managed to make all this work.
So what's your secret to getting all of these things done and staying sane?
First, just everyone here should read Romney's book.
It is so good. My co -founder read it first in our company and then told me you have to read it.
I read it, now it's required reading for our growth product team.
And the thing that it will completely change your mind on is it will change your mind as to what onboarding is.
So product managers, we all think about onboarding all the time.
But we all think about it as from when you sign up to when you've used the product.
And Romney's book will show you that onboarding starts a lot before and ends a lot later.
And focusing on onboarding in that way will change the outcomes of your product like it has for us.
Anyway, Romney's brilliant read his book.
As far as speaking, we are so lucky.
All of our users are builders.
They are out in the world trying to change the world by using software.
And so we're over here building our company.
Our users are over there building their companies.
And so our job is to share what we're learning building jam with everyone else building their things.
And so we end up posting learning lessons online, joining things like this.
It's such a privilege and an honor.
So thank you for having me.
I appreciate you taking the time to pick up another panelist.
That's so awesome. See, I told you guys, she's a great person.
So we also have Anuj Adiyya joining us today.
So Anuj is a growth expert and author of Growth Hacking for Dummies.
He's got writing on productled .com.
And he's also got deep experience guiding SaaS companies on implementing product -led growth strategies.
So really honored to have you here with us today, Anuj.
During our pre -call, you shared that you're planning on setting up a Guinness book of world record by gathering, I think this is just so cool, the largest number of people wearing party hats in Boston.
I think we need some more context here.
Can you tell us more about what you're doing and how can people get involved?
Where do we send the party hats?
Thank you for having me.
This is such a fun group to be part of.
And yes, I think on the surface, it feels like it has nothing to do with product -led growth, but it really does because it's a large experiment.
So there's this party invite app that I'm consulting with called PartyClick, if you want to go check it out and easiest way to set up an event.
And we just try to think of like, what are ways we can put this in front of more people?
And a lot of ideas came around, especially things like, oh, we should do our own celebrity lookalike thing, like the Timothy Shalomay thing that just happened.
And I'm like, yeah, that's okay.
But what else can we do?
And then somebody on the team was like, don't they have world records for largest gatherings?
Yes, that's what we should do.
So I go and look up the Guinness Book and sure enough, there's a world record for people with party hats.
I'm like, great, this goes with the name of the product, you know, party hats, party click, great, we should just do this.
So literally this week, I'm in the process of getting through the application with the Guinness Book and, you know, begging and pleading the city of Boston to let me have 2 ,500 people be in Boston Common.
So, you know, let's see how close I get to achieving that goal.
So but and the real reason I'm saying this publicly is more to hold myself accountable and shame me if I don't make this happen.
So if you're in the Boston area, bring your party hats.
What was the date again in December?
With thinking December 22nd.
So okay, well, now you got it.
Where was in the area?
What about it? And, you know, if I don't make it happen, you know who to throw the brick bats at.
So moving on to setting the stage, the intersection of AI and PLG.
This question is for Romley.
As a general question, has PLG become essential for the SaaS industry in general?
We are we out with that right now?
I would say it depends on yeah, it depends.
That sucks. But it depends on how you define product lead.
I think if you say product lead is PLG necessary, if you have a free trial freemium, I don't think so.
I think like I actually suggest startups start with high touch, you know, get close to your customers.
If you mean by product lead is removing any unnecessary friction and create a great experience for users.
I do think if that is the definition that we're talking about PLG, then yes, the standard and bar for people in terms of what they expect from a product experience has exponentially grown a lot more than it was 10 years ago, where, you know, something annoying, it's like, Oh, that's normal.
But now it's like, if this is annoying, I have 100 other options that I can jump into, you know, there's a ton more products out there.
And I really do believe that creating a great experience for end users, for users who are going to be using the product is going to be critical in terms of retention and activation and all the other things that we're going to be talking about later, we're going to be touching upon how it will affect the rest
of the funnel. And if that's the definition of product lead, and yes, I do think it's essential.
Do you mean that's free trial and freemium?
Then no, I don't think so.
Wow, nuanced answer.
Okay. Does anyone have anything that they wanted to add about just like the relevance or state of PLG today?
It's so powerful. I'll just add that I see a lot of founders actually asking the wrong question.
Because the question they seem to ask is how can I be more like Slack or how can I be more like when really the question they should be asking is, how can I use my product better to serve the end needs of my customer or my user?
Right? And that can take any and many forms, right?
And that opens up far more powerful avenues for implementing a product lead approach than just trying to copy something that's just never going to work for you.
So true. So let's kind of move into where AI is sort of entering the space.
This one is for Danny.
So how would you say AI is poised to supercharge PLG strategies?
I'm not sure if you've got any stories from how you guys have leveraged it at Jam.
We're about to go into all the tactics, but at a broad level in PLG, your product does the selling.
And selling is something that is more effective when it's personalized, it takes a lot of content, it takes a lot of convincing.
And so this is actually something that AI is really, really good at.
Companies have been trying to do this forever, like machine learning for personalization, super segmentation with lots of content.
It's now easier than ever to do this really well, you have to understand large swaths of unorganized data.
And so really, really excited to dive into the tactics.
This just feels like something AI is really good at.
One last thing I'll say is that the best PLG strategy is just to have such a great product that people can't help but talk about it to others.
And with AI, you can now build even more powerful features for your users.
And so that's also a really big part of the puzzle, I think.
Okay, so we'll move on to building the foundation.
So AI's roll across the user journey, I think this is going to be the section that takes up the bulk of the session.
So definitely a less listen.
Now is the time to put away the phones.
So let's walk through each stage of PLG and show how AI can support each.
So let's start with the acquire Danny, did you want to chime in with how your team is using AI to power that acquire stage?
There are two things to mention.
So the first is maybe the most obvious, which is in PLG, your customers find your product, because the product has sold itself in a way, and they onboard on their own.
And one place where you have to generate a ton of content is in SEO, and AI can be really helpful.
Unfortunately, all of us have seen what this looks like when it's bad, it just looks like AI mumbo jumbo, and it doesn't rank actually.
But AI is actually really, really helpful in SEO.
So here's like tactically how we use it.
It's AI writes the outline using SEO best practices, you literally take an article about SEO best practices, you give it to the LLM, you say, write me an outline that would follow all these best practices, AI that edits your SEO, or gives you recommendations for best practices, using perplexity to search
for the tools that you're going to mention, like, get me all the prices of all these tools, you know, whatever.
So very, very helpful in that way.
But beyond that, we're always thinking about like, how do we build content that engages our audience online, our audience, our developers, and in web development, there's so much going on, the space is moving really quickly.
And so we wanted to be able to do content around that for our audience.
So we thought, Okay, let's create a newsletter and a podcast about what's happening this week in JavaScript.
If that's interesting to you, by the way, you can go to this week in JavaScript .com and every week in four minutes or less, get the news of JavaScript in the week.
And we're a tiny startup team, we can't record a podcast each week.
And you know, AI is really, really helpful for us.
And it allows us to write great content, we have like over 50 % open rates on these newsletters, when the podcast is really well listened to.
But AI really helps us do that.
So those are two ways that we use AI for the acquire step of PLG.
Super cool. I think the other place that I've been using AI for in terms of the acquire is like repurposing content.
So like taking like the podcast episode and then plugging into Cast Magic, which is a tool or summaries.
And it like outputs like blog posts and Twitter posts and LinkedIn posts that you can share and newsletter posts a bit that it's, it needs a little bit of massaging is not perfect.
But I think that's another area in content that I've seen AI this work in terms of acquisition is how do you squeeze more juice out of the lemon?
Is that the saying?
How do you get more bang out of your content essentially, is one way I've seen AI be used for acquisition?
We actually do the exact same thing.
So one of our core beliefs is that founder led sales in 2024 is happening largely online.
So it used to be the case you'd have to get like a warm intro and then like you have a first meeting and the founder like introduces themselves to the customer, that's happening passively online as people are scrolling LinkedIn today.
And so you're trying to constantly share how you see the world and what changes you're trying to make in the world passively online.
And so we use AI for part of this, and it's just the repurposing content.
What we'll do is if someone on the team goes on like a podcast somewhere, we grab the YouTube link, we'll put it into a tool like Opus Clip, which will clip up the YouTube link.
And it either gives us video clips to share online, or it pulls out video clips of things that were like aha moments.
And that gives us nuggets of things to talk about and write about later.
That's been really, really helpful in the acquire stuff too.
I think that's that's a huge thing right now.
I think both of you kind of spoke to this being able to repurpose content as much as possible and yet get as much juice out of the lemon as Ron Lee put it.
That's super awesome.
Okay, I think we can move on to the activate stage then.
So Ron Lee, you're the onboarding guy, what would you suggest as far as AI enabled practices to kind of support that activate stage?
Yeah, this is actually something that I found another company do while we were in a conversation around how they're using AI in terms of their onboarding.
So one of the challenges with product lead is challenges or opportunities as well, is that you really have to have a strong like user research or customer research muscle.
With sales that approach, you're like in direct contact with customers, you're hearing objections.
When your product is the one selling, you don't hear those objections, they just leave.
How do you get those valuable insights, the wide that the data, the quantitative data can't show?
One way to using AI, which I thought was cool, is they take all their sales calls.
They have some sales, they have like a hybrid approach.
They take their sales calls recording, and then they have like also a high touch onboarding experience for enterprise companies.
They take that recording, they plug it into chat GPT to train the large language model.
Once it's plugged in, they've kind of started using it to create things.
One of them, which I thought was cool, is create a sales to customer success handoff document.
And here's kind of the things that you should do, you can help craft, can you help sequence, create a five email onboarding sequence based on the pain points on the sales call and customer onboarding calls.
So since it's been trained on the objections that the sales calls have made and the kind of question and confusion during onboarding calls, it's more aware and more trained as to what your customers are actually going through.
I suggested that they also plug in their most top requested support tickets into that chat bot or that chat GPT and then kind of feed it information that is valuable to that activation phase.
So that when you write that sequence, or you write that tour, like what is the tour and the sequence and that can really help out with that.
I think that's one way it's around kind of like plug into chat GPT.
I've also seen more advanced tools, AI tools kind of plug in your data and it kind of figure out like retention metrics.
So I'm seeing like amplitude or things like June that SO where you plug in your data and it kind of tells you, hey, your most of your successful users, if they do X, Y, and Z, 30 % stick around, which you know, I'm going to take that with a grain of salt of course, but it is some valuable information to
kind of like go through the data you have, especially if you have a large enough user set, rather than like, you know, if you have 10 users, you plug it in to this model, it's not going to have enough data to figure that information out.
So I would say, I think that's a caveat to that kind of analysis where if you do have enough data set, you know, over time, then you could probably plug it in and kind of figure out like some of those information and data like that.
Yeah, that makes sense.
Does anyone else have any kind of tips that they would want to chime in with on supporting the activation stage of PLG?
This is kind of obvious, but like once you figure out, okay, good activation, the user has to take the following steps, then you have to figure out like what are a bunch of ideas we should try to get those users to do those steps, right.
So for us, we consider a user activated if they create four bug reports with jam.
And so we're always trying to think like, How do we push more of our users to create their third and fourth bug report?
Like, how can we encourage them and show them the value?
There's just a big brainstorm, like lots of ideas moment.
And it's actually just fine at like, coming up with a lot of bad ideas that trigger you to come up with with the good ideas.
And so that's just another obvious way to use AI in a step maybe.
I do like the idea of generating bad ideas that help you come to the conclusions of good ideas.
Here's the thing about humans.
It's really, really hard to have pure imagination to come up with an idea from zero.
Like if you're ever tasked with being the first one to write the first draft of something, it's really, really hard.
But humans are really, really good at reviewing and reacting because like, that's just how our brains are wired.
We see something and then it triggers other ideas.
And so as long as you have one bad first draft or one bad first idea, you're suddenly really, really good at the second draft.
And so AI is good at giving you the first thing that triggers your brain to think the way that it does.
Well, sorry, I think I might have cut you off like a little bit there.
No, I was just adding on that.
I think the great thing about this kind of large language models is like you can train it.
So it's like, Oh, this is bad.
Here's why, you know, like here's why I think it's bad and kind of give it like some kind of feedback.
I heard this really great quote from a podcast episode with Nathan Berry.
He's over at Kit and he's like, you have to treat AI like a very early stage intern, like an intern who's like a like a beginner.
And by you telling it good feedback, it actually learns what is good and what is bad so that as Danny mentioned that no, it comes with a bad idea.
Give it feedback. And it'll actually start kind of producing better and better output and results based on that.
Or maybe it'll at least the bad version will be a baseline better.
So your ideas are even better.
I don't think it works that way, but maybe.
Ramley, I've been calling it like a drunken ton, but you know, Oh, that's a better day.
But also the only other thing I will add here, and maybe this is a segue into the next thing that we talked about is that activation isn't really a separate step from a user perspective.
Because once they're in the product, they're in the product, and I think what's helped me sort of connect the dots downstream and upstream is just thinking of activation as just short term retention.
Because once they're in the product, that's the game, how do you keep them around longer?
How do you monetize them better?
More all of that, and connecting what they do as Ramley and Danny were talking about initially to what happens later.
It's all happening along the continuum of retention.
It's not a flip a switch in a user's mind.
It's like, Oh, I've been activated.
Now I've been retained.
It's all the same thing.
Yeah. You kind of have to be thinking about things, not just from what you see as the stages, but where the user sees themselves.
It's possible, right?
It's like you're using the products and you use it in the beginning stages you use at the end.
I do think like as a, not as a product person, but as a user, there's a point at which I'm checking something out.
And there's a point in which I'm using it.
And I think activation happens in the middle.
So I do think they are somewhat different from the user's perspective.
Anuj, what would you say is or some of the possibly uses, as you've seen already or some ideas that you have around using AI for that retention stage now that we're kind of in that, okay, we've got our user in the product, where we go in from here.
Yeah. And some of this I'll say is upfront, it'll sound like an expansion of what's already been said in many ways, because I think what's most fascinating to me about retention and PLG is how it's sort of completely transforming our ability to understand and act on user behavior patterns.
And I'll share an example.
So many teams that I work with, unlike, I think everybody even potentially audience, just sitting on mountains of product usage data, right?
Everybody's got amplitude and heap and mix panel and whatever, right?
But for whatever reason, maybe the stage of company, not enough data analysts, whatever it may be, there's struggling to extract enough of the right kind of actionable insights quickly enough, right?
So I think what's changing the game is sort of combining these traditional analytics platforms like your mix panels, amplitudes, heaps, with these sort of modern AI analysis capabilities, right?
And you can just see it, right?
It's almost like people were operating in the dark and now the lights have been turned on, because the thing that started to emerge to me, right?
And I'm sure many of us have used these product analytics tools for eons.
And I thought I was reasonably competent at them.
But when I started plugging in some of this data into whether it was private versions of chat GPT, or things like that, it started identifying these very sort of fascinating micro cohorts, I think that would be almost impossible to spot manually, or based on your level of expertise with any of these
tools, right? And so kind of going off of the examples that already be told, right, I think you might discover that users who perform a specific sequence of actions in their first week become power users at three times the normal rate, right?
Like, I think there's one example from one customer, where you know, they found that users who both export data and share it with teammates in their first few days are showing much stronger engagement patterns, like, this kind of thing is, is the kind of thing people are looking for, right?
So I think what's really powerful here, I think is that combination of analysis and automation, right?
Because I think what some of these companies now are starting to set up better behavior trigger journeys, right?
Because people have always wanted to do this, right?
Because that when the system detects a particular user following some sort of high value pattern, it can automatically sort of shift the content or the experience into getting them faster to that path, right?
And so that, I think, in one case for me has gone even beyond what's happening in product, because a high potential users, in one case, sort of got invited to a beta program, or they got invited to a customer advisory board based on their sort of actual usage patterns, right?
And that's the thing people want to do to run these earlier point of getting closer to the customer, but they just haven't been able to do it fast enough.
Also, I think for me, the real game changer is sort of using all of these tools to connect the quantitative to the qualitative, right?
Because machine learning, yes, you know, you can analyze patterns across support and community and product feedback, but connecting those insights with usage data, right?
So like when you maybe combine that with something like GONG or your product analytics, right, you get a much sort of deeper 360 degree view of sort of not just what they're doing, but why they're doing it, right?
I think that's that of the insight and the speed of that insight, I think is invaluable because you're relying less and less on assumptions and upping the percentage of what's actually happening because of user patterns and needs.
Okay, so I'm curious about this, is do you or any other panelists have kind of anecdotes or stories of kind of like seeing some of this in action?
Because I mean, it's a perennial issue of qualitative or combining the quantum call data and being able to kind of tell that story more effectively.
I'll jump in with another example from my end, right?
For a lot of teams, I think these exercises are disconnected.
There'll be a team that will, you know, run the product market fit survey, then there'll be a completely separate database that will host all of the support tickets, the completely separate database of sales conversations, right?
And sometimes it's not clear even whose job it is to put all of this together.
And it's just that uncertainty and indecision, not because of somebody's fault, it's just because nobody was told that this is their job because they're like, oh, I'm a user researcher, I'm a salesperson, right?
So I think it's power and sort of assigning it as your, let's call it your AI employee, to make the take that decision out of your hands and just collate all of this data for you and then present it to all the stakeholders involved.
I think is a trend I'm starting to just see more of where I guess, you know, the opportunity cost of waiting or indecision is just so high, right?
That this is sort of the most MVP way of let's just connect a few data sources or even if you can't automatically do it, let's just take some exports of these data.
Let's just throw multiple spreadsheets into this thing and let it elucidate these patterns for us.
Even if you do the most MVP way of starting to speed up your rate of insights and being able to act on it.
Web designers, this one's for you.
I've got 30 seconds to tell you about Wix Studio, the web platform for agencies and enterprises.
So here are four things you can do in 30 seconds or less on studio.
Adapt your designs for every device with responsive AI, reuse assets like templates, widgets, sections and design libraries across sites and share them with your team.
Add no code animations and gradient backgrounds right in the editor and export your designs from Figma to Wix Studio in just a click.
Time's up, but the list keeps going.
Step into Wix Studio and see for yourself.
What I'm hearing here is kind of like using AI as a way to sort of break down some of the silos between all these departments.
So are you suggesting sort of like generating a cross -departmental state of the union report kind of thing to kind of help people give everybody in different departments like a bird's eye view of what's going on across areas?
No, absolutely. I think it's not controversial to say that growth in general is a multiplayer sport.
It's cross -functional by definition and certainly true in the product -led world because you're trying to bring all of these pieces of content and education and community and product and sales and all of these teams have to work together.
And if you have a growth team of any size, the entire purpose of that is to bring all of the key stakeholders together to understand what's the current state of growth?
Where are the problems?
Where are the opportunities?
And it's always a challenge for everybody to understand not necessarily what is the role I play, but what is the best, most impactful thing I could be doing right now?
And communicating that because I think being a person that leads growth is as much about your point storytelling and communicating insights rather than here's what the data says for people to understand why should they care about this thing.
Don't rely on your own abilities, right?
Like get help from an AI tool to help you understand how do I communicate this data point better to sales versus dev rel versus product, whatever that may be, right?
And so I think it's power and breaking down those silos and getting everybody on the same page is sort of highly underrated.
I really, really like this tactic.
I want to kind of upvote this tactic.
We had a podcast episode with Michelle Ronson, who is a well -known UX researcher.
It's a very similar kind of process that she recommended that was a little bit more analog.
It wasn't necessarily AI enabled.
I like this as an AI enabled counterpart, but her recommendation was more in terms of UX research.
When you've got conducted all this research, that's great, but then how is it usable to all of the different departments that it connects to?
And her recommendation was just to frame it department by department.
What are these learnings kind of informing as far as the next action items for each department?
So I kind of see it as sort of a similar and equivalent to that in which you're kind of using the AI to kind of translate like what are the findings that we're seeing across departments and what are the action items that we can kind of derive to kind of help us work together better as a team?
Because yes, you're right.
I think that this is an issue that we struggle with in every startup and every company of any size is, you know, how do we work better cross -functionally?
I love this discussion point.
Does anyone else have any notes on even using AI to work better cross -functionally, I think is a whole other area that we haven't explored.
This is not AI for cross -functional, but look, why do companies hire product managers?
As everyone on this call knows, it's because you want one -point person who is going to get to the bottom of things and owns the whole problem and owns the success of the product and is going to do literally whatever it takes to make the product succeed and the project go well.
And so like what Anuja is saying where something is not someone's problem, like someone understood it at a spreadsheet level, but not at a user level, like it's so true.
But that's the beauty of PMs, right?
There's no PM in the world that says that's not my problem, they're just there to make the product succeed.
So not AI related, but one tactic that's been very helpful for us that maybe is helpful for you all is sometimes it can be a leap to go from a spreadsheet level to a user level if you don't have user calls set up anyway.
And so the best thing you can have is a drumbeat of like you're talking to customers all the time.
And then you just plug in a few questions that you have.
But that's tricky, how do you do that?
And so we have a little bit of automation around this that I recommend for everyone, which is every single week, the top 100 users from the week before that haven't heard from us recently, get an email from a DM co -founder asking how things are going.
And if they reply, and their reply is like very thoughtful, we'll say like, oh, like we'd love to learn more, like would you be up to hop on a call?
And so we have this drumbeat of calls.
And what that means is that if a PM is looking at something and saying, this doesn't make sense to me, or I'm curious to learn more, there is something easy to join instead of having to invent a new motion of customer calls from scratch.
So one tactic not AI related, but highly recommend.
Awesome. Yeah, yeah, it doesn't have to be AI.
We're just looking to grow.
All right, so we'll move on to the expand stage here.
So Neusch, if you want to take the lead on this one as well, anything you wanted to lead with on expanding, then we can move into Danny's tactic as well.
Sure. Like I said, everything's sort of a bit of a continuum for me, right?
But if you think about sort of the traditional expansion playbook, like another work, like you wait for usage indicators, maybe you have a customer success team to do some quarterly reviews.
Quite frankly, I've worked with some teams that are just hoping to catch expansion signals.
But I'll give you a practical example.
This is one SaaS company that are using sort of gong, right, to analyze customer interactions across sort of every touchpoint, right?
Sales calls, support tickets, QBRs, everything, right?
And sort of the key is sort of the trying to train the system to look for very specific expansion indicators, right?
So it's not just picking up on like very obvious phrases, like we need more licenses, like sure, right?
But I think it can start to identify patterns around discussions of say, like adjacent use cases, right?
A mention of a new team member or department or a conversation about a pinpoint that could be solved with some additional capability, right?
So I think when you start to combine that sort of conversational data with product usage analytics, that's when it's allowing this team to create, let's just call it expansion intense scores, right?
Based on sort of multiple signals, right?
So then when they start to hit certain usage thresholds, right?
Like in this one case, like they were approaching like an API limit, or they were attempting to access some premium features, right?
And this sort of thing correlates with very specific sort of conversation patterns that they've also analyzed, right?
And that sort of started to create a really clear signal for expansion opportunities.
All of this to say it's not just to identify expansion opportunities, but to I think make the process also feel more natural and sort of value aligned for the customer.
Fair enough. Yeah. Danny has a process that she was going to walk us through as far as using call transcripts to generate first reference follow up emails, which I kind of alluded to earlier.
Did you want to walk us through that process, Danny?
I mean, pretty self explanatory.
I'll say that as product people, we really like, at least, I really believe the product should do all of the heavy lifting, but really to change people's behavior, you also want a face that you can trust.
And so a lot of the expand step is very human led versus product led.
In a really well running motion, you have a lot of back to back calls, helping teams expand their usage.
And you want to follow up from those calls with something thoughtful.
And again, it's hard to start from no draft, it's a lot easier to start from a first draft, taking a call transcript, giving it to like, what were the three main points this person communicated to me?
And then using that to draft like, here's what we heard is just a lot easier than blank slate, especially if you've eight back to back calls, then you're doing it at the end of the day, you don't fully remember, or like, brain is a little fried.
I appreciate that. Anything to save the old brain at the end of the week.
Okay, we'll move on to section three here, which is from Vision to Execution.
So this is all about building AI enhanced PLG strategy.
I'm actually going to throw it to Rony, because we haven't seen him for a couple of minutes here.
I would love if you can kind of take us through some of the common pitfalls and misconceptions around PLG around using AI to support it, kind of what are some great pitfalls or detrimental pitfalls that you've seen in your work?
I would say in terms of AI itself, the output is only as good as the input.
So like, if you feed it, nothing, garbage information, then you're going to get garbage for output as well.
And I think just based on the conversation we're hearing, what Daddy is saying, like, don't start with a blank page, you really got to feed it with the right kind of information things, specifically around training them with your best customer experience and the best calls that you have.
And like giving it the right kind of context, especially during the expansion that we just heard, where here are the calls that we've had with them.
And here is the documents that we have around them, like give me the output around that, and kind of giving it like the right kind of context as to the right kind of prompt in terms of like, Hey, we're trying to do this, here's what we're expecting, give me three points.
And kind of being goes back to what Anush said earlier, he called it treated like a drunk intern.
I would say treated like an intern who's just starting the job.
So like a little bit more verbose or a little bit more upfront at the beginning and saying, Hey, here's what I'm looking for.
Here's the output I'm expecting.
And here is the context that is in that in terms of the information.
I think that goes a long way rather than being vague about it.
I've also seen prompts where pretend that you're director of customer success or customer director of product for this company.
Here's the information about the company.
Here's the LinkedIn profile of this company of this person.
And what would I write to all the three bullet points that this person might care about based on the calls that we've had with them?
I think that kind of like being very clear and upfront and being succinct about that as well.
It could be very helpful in terms of using AI.
I think one pitfall to add is, especially with really junior people on your team, if you remember being a new person in the workforce or new person in your role, one of the hardest skills to learn is what does done mean?
And the more junior you are, the more done just means complete versus understanding the impact of what really good work can create.
And so I think one of the pitfalls with using AI in any of these processes is that for a junior person, done can just mean complete.
And so, oh, the AI ran some analysis, but a more senior person would be like, and that has surfaced some new questions for us to dive into that are the next logical things to look at, or this actually doesn't quite make sense to me, this must be a hallucination.
Or this draft of an email reads like mumbo jumbo and like, obviously, we shouldn't send it.
And so I think that's a big pitfall.
And I think the way to avoid that is to actually just have a lot of conversations about what does done mean in this team, what does that mean in this company using it or not?
Great career building advice in general, it's just be constantly asking that question, what does done mean having that conversation constantly?
Sorry, guys, I had a couple of things, one to what Danny said, and one to what Ramdi said.
So the point Ramdi was making about garbage in garbage out.
So a pro tip for myself now that I have begun to implement is I've started to realize most people don't even know how to Google properly.
And that's almost my way of getting an early indicator of whether, if I'm going to hand off a task to somebody, will they even be able to prompt properly, because prompting is so far away from, like, it can get to this point, like super verbal.
But so I literally have people run what I think are, you know, basic Google searches, and see how they respond to that.
And I like that almost gives me a signal of, okay, you know, maybe this is a person that needs a little bit more training, you know, before I let them lose on this sort of a system versus, okay, this person inherently gets how to query a system and let them into the system first.
The other thing about what Danny was talking about, right, where things might seem complete, right, or done.
I think this is why I have found great value in one of the first exercises I do is have everybody on a team understand what is the product North Star metric, right, because this is critical for everybody to understand that this is how we deliver value to our users and customers.
And everything we do is in the service of growing value.
And that applies as much to if you're going to prompt a system, that is the perspective the system needs, not the perspective of the specific task or the specific analysis, is this is in the service of that greater thing of growing value to our users.
I think the common theme that's set in running through all of this is I think there's a little bit of more context setting and background that whoever is going to interact with these systems should have, you know, in terms of that sort of greater user perspective before they start interacting with these
systems and extract whatever they think is an insight.
Yeah, some definitely a few common themes emerging here.
That is as well as challenges with regards to working together as different themes.
We're just gonna go through a little pre close here.
Normally we take this time to just tell you about our next session.
For those who are kind of engaged with us month over month with sessions, just so you know, we are not going to be doing a panel in December, we're taking a bit of a break for the holidays.
So if you see, you know, there's no session, you know, where are they?
We are going to be back in January with a basically a more of a career focused session.
So it'll be about transitioning into a career in product management, that'll be the focus of the panel.
So registration for that will be starting in December.
So we'll send out a link to our subscribers, when that is available when the registration is open, but it should be a really great session.
Those who are here probably won't be relevant as much since you're all product managers already.
But if you do know anyone who's interested in the career, who's been asking you a question about making that jump, please let them know.
We'd love if you could help us spread the word.
We'll get right into questions.
Q &A, our most voted question is from DM.
It is AI can exhibit bias and interpreting nuanced data leading to misleading conclusions.
In the context of PLG, how can we strategically leverage AI to enhance user experiences and drive growth while mitigating these risks?
Anyone want to take the lead on that one?
The thing about AI is it doesn't solve human problems, and you really just have to use your judgment.
So it's a great tool, but you need a in the pitfalls of junior person thinking something is done.
Just having an answer doesn't actually solve your problem.
I just, when you're using AI especially, you have to be even more really cognizant, really there, really present with your work, really thinking.
I totally agree. I think this goes into the whole ethics of AI.
There needs to be human intervention.
I really do believe that where that bias is, there needs to be human to catch those kind of biases, especially.
I think it's just an input to your decision -making process, and you could have multiple inputs.
You can have qualitative data, quantitative data, you have AI suggestion, you can ask your CEO, you can ask somebody from customer success, but these are all input to eventually it needs to be a human to make the decision.
It needs to be somebody's like, yes, that's what we're doing.
That's where we're going because based on all of this information, including an input from AI, and I would be very cautious if we let AI make the decision for us and kind of go from there.
There's obviously there's some biases there that it might not catch on its own.
What I would say is, I think it's really important to not forget how things were done in the past.
Let's just call them like ground truth baselines.
How did we do things before AI?
Just so that we can have a clear comparison point, right?
Because if we don't know how we did it, how will you even know whether that system is hallucinating or even have the opportunity to ask that question as to, should I poke at this a little bit more or not?
And I think what's associated with that, right?
I think it's sort of important to have sort of this triangulation approach.
So I don't know, let's pick a situation where, let's say an AI system is sort of flagged like some users at risk or something, right?
It's like, okay, you can't just take that at face value, right?
You've got to sort of look at sort of the raw product analytics data.
You got to look at feedback, look at whatever the customer success team is talking about, right?
And see, like, are those signals manually also aligning?
And that gives you a little bit more confidence in the system as well, right?
I think a recent example I came across, I haven't personally used it, so disclaimer, I saw a heap analytics has this sort of great feature, right?
Where, you know, they call it like, you know, they've explicitly designed for diversity in their training data.
And sort of in PLG, what that means is they've ensured that their AI models learn from users across different company sizes, industries, use cases, things like that, right?
So what that sort of helps you catch sort of when your AI system might be over indexing on a behavior pattern from like a larger or a more active customer base, right?
While missing any signals from smaller ones, right?
So I think there are some tools that are catching on to this and trying to account for it as well, right?
But I think, just don't forget the way you've done it and you can always manually verify.
Good tips. So the next best uploaded question is, would you recommend using AI in place of traditional support or customer success teams?
I feel like we've kind of already answered this one passively, but we can maybe give some more context here.
So those roles are too typically seen as responsible for cohort retention, yet are two areas where we were seeing explosive growth rate in agentic AI tools.
This is true. This is interesting.
Okay, who wants to get into this one?
I think that the most important thing in 2024 is your users experience, because there's more competition than ever.
And it's easier than ever to switch tools.
So because most companies have some sort of PLG strategy, if someone gets frustrated with your product, it's not that much later in the day that they sign up for a free trial of another product.
And so user experience in 2024 is the most important thing.
And so I think that that's the way to make that decision for your users.
If the user experience that you want to enable is, we've got a lot of documentation, we want to make it really, really easy to search, and natural language is a more simple way to search our docs, then having some sort of agentic customer support is actually great, right?
Because it's just a nice UI for your docs.
If the user experience that you want to provide is that we're here for you 24 seven, there's a person on the other side who cares about your problem and is right there with you for every step of the product journey, then that's probably not the experience to provide.
It's possible that for some of your users, like for your developer audience, you want one, they actually don't want to talk, get on a call, but they want just quick query language for information.
And for your enterprise customers, you want something else.
So anyway, I think that the way to think about this is not through a tools lens, but rather through what's the user experience.
I totally agree. I think people are craving, I think with the rise, with more and more rise of AI, people crave human connection.
There's this quote, because I'm working on a new book, but from Josh Kaufman, he wrote the book, the personal MBA, he said, there's a paradox of automation, the more that the automation is efficient, the more crucial the human experience is.
And I think that's so true, because I think in talking about customer success and customer support and user experience, like what Danny was talking about, I think people would be delighted when there's an actual human chatting with them, there's like face to face, like right now.
And whenever I go to in person events, they just like, oh, you're real.
You're not like a chat bot, that's like spitting out things.
You're actually somebody who cares for me, you know, who I feel valued as a customer.
And I think that's going to go in a longer way, especially as companies automate more and more things through AI support, you know, marketing websites and things like that.
And even emails, I heard KFC is introducing AI emails to get you to buy more fried chicken, which is crazy.
But I think it's really going to be more important to have humans involved, especially in B2B products, I think where, you know, customer and face to face does matter in that value.
I think at the same time, the bar for what quality AI experience is getting higher.
Imagine like your own experiences using products out in the world.
If someone introduces an AI feature or an AI chat support, and it is a low quality experience, you actually hate them even more than if it was just like a normal feature or normal chat support.
There's something about like, oh my god, they're just on this buzz train, it's just hype, it's just marketing, it's whatever that just like feels bad in your gut.
And so I think that if you are going to introduce such a core functionality of your product, like customer support using AI, and that the customer knows that it's AI, it better be like really darn good, the bar is actually even higher to ship that than it would be otherwise.
Our last question for the day, how do you balance personalization through AI with user privacy concerns?
Follow your privacy policy.
If you're thinking this is wrong, for example, if in your privacy policy, it says we're allowed to look at top level usage metrics in order to improve our product, then an AI can help you in that.
If in your privacy policy, it doesn't say we're allowed to look at all user data to all like it just, you know, whatever, like user trust above all else.
I don't think that there's much else to say about that.
There's definitely some nuance response.
Like I've gotten emails for like, oh, Ramly, I noticed you did X, Y and Z in our product.
And like, I think there is some value there where like, oh, yeah, okay, I'm stuck.
I need help. But if it's like, I'm not sure, I think there's definitely I would agree with Daddy.
I think it really is like the privacy concern where like, oh, I feel like I'm being stuck, but I'm in your house, I'm in your product.
So maybe I do expect that a little bit for you to see what I'm doing.
I think this one goes to a question I've been kind of pondering, right, when somebody I don't know who said this to me, but they called it, you know, their attempt was at respectful personalization.
And so I think that that the key for them was to be really transparent about the value exchange, right.
And I don't think people are not going to expect that we use AI to personalize experiences moving forward.
But I think it'd be useful to just follow, you know, take simple rules like, okay, if we're going to collect a particular data point, we internally really should have a really clear demonstrable benefit to the user.
That as your guiding light first.
These are all very insightful.
Unfortunately, we are so out of time.
We always do this. We always leave ourselves not quite enough time to get through all the questions.
Thank you everybody for such an engaged session.
Thank you for participating in the chat and for all of your questions.
And for joining us today.
Really great of you guys to make time.
I hope that this session was helpful for everybody who attended and I want to give a warm thank you to our panelists, Bradley and Neuge, Dani, you guys have been amazing.
Always a pleasure to have you around with us.
And thank you everybody have a great day.
Thanks for listening in.
For more great insights, how to guides and tool reviews, subscribe to our newsletter at the product manager .com slash subscribe.
You can hear more conversations like this by subscribing to the product manager wherever you get your podcasts.
