In this episode of PROMPTED, Kyle James sits down with Mike Lemire, a longtime customer success leader and advisor, to unpack what is really happening with AI inside CS teams.

AI was supposed to make customer success more proactive, more strategic, and more human. Instead, many teams are chasing “AI strategies” without clarity on outcomes, customer experience, or value.

Mike shares what he is seeing across CS organizations, including:

1. Why AI strategy has become theater instead of problem-solving
2. How investor pressure quietly shapes AI adoption in CS
3. Support teams as the earliest AI testbed and the risks of optimizing for deflection
4. The danger of over-indexing on sentiment and the “customer watermelon” problem
5. A hackathon story where vibe coding replaced a six-figure SaaS purchase
6. Why better product experiences could ultimately reduce the need for CS itself

This is a grounded, honest conversation for customer success, CX, and GTM leaders who want to use AI thoughtfully rather than chase the latest tool.

🎧 Listen for practical lessons, cautionary insights, and a clearer way forward for AI in customer success.

Learn more and Connect with Mike:

1. Connect with Mike on LinkedIn: https://www.linkedin.com/in/michaellemire/
2. Learn more about Harmonic Leadership: https://harmonicleadership.com/

digital revolution, and then the cloud revolution, and the mobile revolution, and sort of you see these waves of tech that produced these massive returns for investors, and AI is now sort of the next wave to ride, and they've all placed their bets.

Anytime you're shopping for an AI product, what you'll hear the sales rep say is, you know, this will free your team up to do more strategic work, right?

Like that's, you just sort of put that on your bingo card.

Oh, AI will impact us in the future from an enterprise sales standpoint.

Like it is

happening today.

Kyle, I do feel like I'm coming across as anti-AI though.

I want to make sure, like, I do like it.

I think there's place for it.

I just, there's some cautionary tales I want to make sure people understand.

And I have not yet seen.

anyone fully articulate what that more strategic work is, at least in the world of customer success.

If they are bringing in AI technologies, they're not freeing the people up to do more strategic work.

They're just adding more accounts to their book of business and increasing the capacity of a CS member or a support member right there.

The ratio of ARR to CSMs is going up instead of sort of maintaining and then having the CSM provide more value to the end customer.

This episode sits at the intersection of customer success, economics, and AI reality.

Mike Lemire has led customer success teams at multiple companies, including Toast, HubSpot, and Overjet.

CS leaders are being asked to have an AI strategy, often without clarity on outcomes, ethics, and even ownership.

So without further ado, let's jump right into with Mike.

Mike, so talk to me a little bit.

You work with a lot of CS leaders across many different organizations.

What are you hearing most often right now about

how AI comes up in this conversation.

Like, what are the leaders thinking about in the CS land?

Yeah, it's interesting.

So probably jump back five years ago, the question that came up in every sort of CS leadership meetup, whether it was sort of like formal or informal, just over drinks or something is, you're buying Gainsight?

Should I buy Gainsight?

How's Gainsight going for you?

Like, that was kind of the product we all had.

It was the only one.

Or, well, we didn't get Gainsight, but we got Totango or True Zero or sort of another sort of

version of Gainsight.

Today, the conversation is, what's your AI strategy?

What are you guys doing with AI?

How are you figuring out AI?

That's the topic, kind of the only topic that's coming up in customer success.

And it's really interesting for me because it's less, hey, how are you solving this problem with AI?

And more, my leadership team, my board is asking us to figure out our AI strategy.

And so I need to go figure out my AI strategy, right?

Which is so disheartening and such a...

perverted way to come that purchasing a piece of software for your team.

Well, yeah, and how many times have we all seen the Simon Sinek conversation start with why, right?

Like, why do we need an AI strategy?

Like, that seems the place to start, but how are you kind of interpreting that and like breaking it down with them?

Like, where does it go from there?

If I'm sort of just a fly on the wall listening, it will go directly into products.

These are the products that I purchased.

This is what we're doing with it.

And there's no real focus on like, oh, what were the problems you were trying to solve?

Why did you bring this in?

What's the ROI of the solution?

Just, I'm using AI so I can sort of wave my flag to the board and put in the board deck, this is what customer success is doing with AI, which is a bummer.

So if I'm getting into the conversation with them, I will first ask like, well, what are the problems you need to solve this year?

And how can AI support that?

And then secondarily, I'll ask, we are the defenders of the customer experience.

How is AI going to enhance your customer's overall experience?

That piece of the puzzle at the moment is very frustratingly gone.

Like no one is really talking about enhancing the customer experience.

They're just talking about sort of how can we bring AI into the team?

And if any outcome is desired by them, it's how can we do more with less?

How can we sort of make sure that the team is doing more with less, fewer resources or fewer teammates?

It sounds like there's like a, how do we consolidate the stack?

We don't want fewer tools, but it's also like, what AI tool can we go out and buy, right?

That seems a little bit contradictory.

Yeah.

I don't even know if it's, let's consolidate the stack, right?

So it's do more with less people, not less tech.

The dirty little secret people don't want to talk about, right?

Yeah.

The dirty secret right now is, you know, I think there's this phrase that comes up in all the sales pitches for AI.

Like anytime you're shopping for an AI product, what you'll hear the sales rep say is, you know, this will free your team up to do more strategic work, right?

Like that's, you just sort of put that on your bingo card anytime you're taking a sales call.

That's the buzz.

That's the buzz phrase.

And I have not yet seen anyone fully articulate what that more strategic work is, at least in the world of customer success.

If they are bringing in AI technologies, they're not freeing the people up to do more strategic work.

They're just adding more accounts to their book of business and increasing the capacity of a CS member or a support member right there.

The ratio of ARR to CSMs is going up.

instead of sort of maintaining and then having the CSM provide more value to the end customer.

All right, so what would that look like to actually see that successful, right?

How would the AI fit in there the right way?

Is there certain workflows that they might solve for, or problems, or how do they fill in the gaps, right, in a constructive way?

Have you seen anybody do anything like that?

Or have you thought of, I'm sure you've thought about it, but.

Yeah, I mean, I have.

And I think the way that it can fit in the best for CS teams is around, I think the biggest challenge we've seen and heard over the past few years is I want the proactive CSM.

How can I get a more proactive CSM team?

How can I identify customers that need help?

And how can I get the right type of help to the right customer at their

at the right time, right?

That's probably a catchphrase.

If you have a bingo card for a brand new CS leader coming into your organization, what are they going to pitch in their first 90 days?

It's going to be that exact phrase, right?

They'll all pitch it.

But they're not wrong, right?

That is what we need to move towards is using data to identify which customers need the right level of support and move away from the old model of, yeah, we'll touch base with our customer every 12 weeks, right?

That's sort of the very, very old school model.

So that is where I think AI can be supportive.

There's a couple products I'm seeing, like one of them, Reef.ai, helps develop customer health scores to better identify customers in need of help who are trending at risk or to better help identify customers who have a high growth opportunity that you could be spending time with to grow your revenue.

There are some products like that I think

are helpful, that are letting CS teams focus their energies in the right place and sort of do more of that strategic work.

But instead, what I'm seeing is let's have call recording identify the sentiment of our customers.

Let's have a sort of front door experience, whether it goes from, I mean, support is, tier one is basically all chatbot at this point.

But how can they move that upstream to the CSM?

How can we automate the e-mail writing for all the CSMs and sort of take away that human level experience?

I don't love that as much as helping me identify the right customer to spend my resources on.

Well, and how big of a lift is that anyway?

Because what team doesn't have templates they pull from anyway?

Like every good CS person, just like every good sales rep, has their own e-mail templates that they pull from and just swap out certain things, right?

Sure.

And I'm not opposed to templates.

I mean, the way that I think about it is, SMB to enterprise is a spectrum.

And so SMB on one side of the spectrum is science and enterprise is art.

And on the science side, like you get a consistent input and you run a play against it and have an expected and consistent output back from that, right?

So you should always be tweaking and adjusting.

It's much, it's very similar to SMB style.

marketing approach.

In fact, I think running a modern customer success team today with an SMB-focused segment is more in line with running a marketing team than it is running sort of a consulting team, which a lot of CS leaders say, we want to be the trusted advisor.

But it's really almost a marketing team.

On the enterprise side, though, it's much more of an art.

to understand the human beings that you're working with, understand the stakeholders, the relationships they have.

I need to pre-wire this person ahead of that meeting so that everyone that I need bought in is bought in ahead of time.

There's some tooling that AI can, or some support that AI can provide for customer success there to sort of make them smarter, make them better informed.

But at the end of the day, you're dealing with people.

And I wouldn't trust an AI automated response on a high dollar value at risk situation.

I want a person handling that and who knows the people and have developed a relationship with those bad customer for multiple months.

Yeah, totally.

But it does feel like what you're leading to is...

There's no reason the AI couldn't like be the briefing of like, hey, before you jump on this call, here's what you need to know.

Like, look through all the recent notes, look through what's going on, understand the power players.

Hey, you might actually get this person on the call.

Recommendation from AI.

And here's the, here's what you need to go onto the call because let's be honest, if you've got 4, 6, 8 calls a day, like,

You need that focus.

Give me 5 minutes before this call to like get on cue before I jump in.

And like, that's probably a huge opportunity there with AI, right?

Absolutely.

And I sometimes feel like in these conversations, Kyle, is like, I feel as though I'm waving an anti-AI flag and I'm not.

I just, I haven't seen the industry respond to AI in a way yet that I think maximizes the value of the CSMs

instead of sort of just overloading them, giving them a permission to overload them or focus on enhancing the overall customer experience.

I think that those two North stars needs to drive the decision making here for CS leaders.

And right now it's more of a FOMO.

Like I don't have AI.

I need it because everyone else is.

I'm worried that I'm falling behind.

I should just go buy something.

It's driven by FOMO or cost reduction, which I get it.

That's part of the game.

You have to be doing that as a good leader.

Am I as efficient as I can be?

But I'm just not seeing enough of the focus on the areas that are going to move the needle, I think, for the customers.

That makes sense.

I mean, let's think about it.

The reason everybody does call transcription is because AI is just naturally good at that, right?

Great.

You take audio, you turn it into writing, you can pull together and pull together notes from it.

Like that's AI's superpower right there.

So it's easy to do.

So everybody has that.

But to your point, it's like, what are you doing with that information from that point?

Because there's like all these different playbooks or use cases from that.

And those have to be bespoke and custom to the organization and maybe even the person too.

And we just haven't got to that piece yet, right?

Yeah.

So here's sort of a, this isn't true in all organizations, but in a lot of organizations, the CS leader is usually less powerful than the sales leader in terms of demanding budget for software for their team.

So what you sometimes see is a, we'll just

We'll just jump on board with whatever products the sales team got and sort of shift it, mold it in a way that works for us.

So you're seeing that with the call recording, right?

Like the Gongs of the world.

Salesbot Gong, we're going to use Gong too.

Yeah.

Exactly.

So we'll just use it for our calls.

And then you're starting to see this, exactly, the Sentiment.

And Sentiment comes in almost every new CS platform that has AI, has a huge spotlight on Sentiment.

because it's easy.

I think it's because it's like baked into a lot of these models.

It's an easy feature to bring out that's sort of hard-coded into what a lot of these large language models are already developed to produce.

The worry that I have about that is it's then raising the value in the minds of a lot of CS leaders on what sentiment is telling you.

And in customer success, we have this concept of a watermelon where the customer looks green on the outside, but is red on the inside.

And I would so much rather have a customer who is grumpy and their sentiment seems bad and they're always frustrated.

They're always saying, hey, I really need, where's this feature?

I talked to you about this last year.

I can't believe it's not in the product yet.

but is deriving a ton of value from the product that like they would never rip away the product from them because they're getting the ROI significantly.

That matters to me so much more than a customer who doesn't use the product but has a great, happy, cheerful relationship with you, checks in on your kids, you check in on their kids.

That's the watermelon risk.

So sentiment is becoming a

more consistent topic that CS leaders are focusing on, and it's diminishing what they should be focusing on, which is recognized value.

That's really what it's all about.

Yeah.

So what are ways you think AI can help do that?

Right.

We talked about health scores a little bit earlier, right?

Like some of this data probably could go to like customize that, but I

I guess maybe it even makes sense.

Like let's dive into health scores a little bit more because how unique is that across organization versus is there some universal elements that every company should care about when it comes to what a health score should be for a customer?

I don't know that there are universal elements.

Well, I guess there are, right?

Like it comes down to adoption.

And I think health score is a proxy to value.

is what it is, right?

Like, why do we care about adoption?

Why do we care about usage?

Because the assumption is the more they're using it, the more value they're deriving.

There are some softwares where you're able to see the actual value that you're producing, right?

If you, HubSpot is an easy one.

On the marketing side, they're helping generate leads.

And if you're using their CRM, you're able to see what those leads convert to in terms of dollars in sales.

So the articulation of the value for a HubSpot CSM is pretty easy.

Hey, we helped you generate X amount of leads, which produce this much in revenue for you as compared to our cost, you know, our ROI is X.

If you don't have that sort of bottom of the funnel visibility into what your product provides, if you're a B2B SaaS platform, for example,

you then have to assume adoption equals value and make sort of a logical jump to, here's what I'm seeing usage.

It must equate to some sort of ROI.

And there are calculators that a lot of teams have that'll help transfer

or translate the adoption into an assumed ROI.

But I think that's possibly where AI could get smarter is sort of like taking that from an Excel spreadsheet calculator that they have and adding a lot more sort of unique elements to that business that comes back from the discovery that the sales rep did that might be lost in their gong notes that are just living in a open text field in Salesforce.

But use that to populate the ROI message.

which is sort of the focus of the renewal, which is the moment that matters the most in a CSM's year.

Yeah, that makes a lot of sense.

It's like, how do you backward to that currency piece, right?

Like you mentioned HubSpot and Leeds, like that's currency, right?

It just is.

You sign a dollar amount.

So what is the thing that you can send a dollar amount to?

Usage or activity or adoption in the platform.

And then how can you use

enrichment through AI to like map that back, map that better or get that data in a way that's automated or faster and not manual.

Exactly.

I mean, another element that's I think should be included in all of the health score development.

And again, like health score as a concept might die.

as AI starts to make it smarter.

Well, it might die as AI gets smarter about identifying who to spend time with.

That's really what the value of a health score is, it tells you who to spend your precious resource of time with the most.

Who should you prioritize reaching out to?

Who and what type of support do they need?

Am I reaching out to them because they are ready to buy and we could expand revenue significantly?

Or they're on the precipice of churning and we need to spend time with them on that side.

That's what health

score does.

I think AI could help teams be smarter in identifying that without the same sort of rubric of a scoring system and sort of simplify that down a little bit.

But the one of the, you asked about the elements here.

So adoption, certainly a key piece of it.

I think support calls are another interesting one.

There's this assumption that we want all of our customers to never have to pick up the phone and contact our support team.

I remember a study we did at HubSpot that said customers who are the happiest, most likely to grow their revenue with us and have the highest net revenue retention, call support 1.2 times per month on average.

And it's because they're using it so much, they're getting so much value out of it, they're bumping into walls, maybe they didn't learn a certain feature, how to use it, and so they're calling support almost for a training resource, or as a feature request, or how might I,

as opposed to troubleshooting a bug, right?

There are some support calls that are painful and that you wouldn't want, but categorizing them, being more thoughtful around sort of the data you're pulling from a customer support ticket and making sure they're not all viewed as equal and not all viewed as a negative experience, I think is a differentiator between sort of the more advanced CS leaders sort of building out this scoring system and the more

So those who are sort of starting their journey and building this out.

Let's stay on support for a little bit, because you said something interesting here, and we've kind of talked about it earlier with the pressure to keep costs down, headcount down maybe too.

But let's be honest, support has been kind of a test bed for AI.

Because, like you said, how do we keep you from getting on the phone with somebody?

And we've seen it deployed more heavily in chat bots and, writing knowledge centers and things like that.

Like, is that true?

Like, are you getting that sentiment?

Are people saying that?

Like, is it because it's kind of...

the lowest hanging fruit, it's just like, I think we have to provide, or I don't know, dig into that a little bit more with me.

Yeah, it's a highly repeatable process, right?

Like to the point we were talking about templates, I mean, it was before so templated, right?

Like, let me copy and paste this response from the knowledge base and hand it back to the customer.

So obvious opportunity for AI there.

As a user of customer support chatbots, I've not been a fan for a long time.

I think we're starting to see them hit their stride.

I've seen some enhancements.

I've seen some customer engagements where there is a benefit to that tier one sort of support chatbot.

And I'm sort of...

slamming my fists on the keyboard, screaming for an agent or a person a little bit less frequently.

Okay, why?

What changed?

Is it more the natural language?

Is it because it's getting you to answers faster?

Or what has started to change?

I think it's more the natural language.

I think it's more sort of the recognition of context and the investment in sort of these tools.

going from a if-then model, I heard this where I heard you say billing, so here is a billing article for you and understanding the context of the question and providing a more thoughtful answer based off of the full consumption of the knowledge base and full consumption of previous tickets.

So it's less trigger oriented and more contextually oriented, I think is the big change there.

Within the support round, though, one thing that I've noticed that has been troubling to me is a language shift and a language shift away from ticket resolution.

And the term now being used more often with AI investments is ticket deflection.

And it's almost this blending of ticket deflection was most commonly used in the realm of customer education, right?

Like we set up our

insert software name here, Academy.

And they have the video series and we have the documentation and we're including that in all of our e-mail links and closing all of our tickets with a reminder to check out help.whatever.com.

And the resources there and the page views there, was a lot of, I would say, a lack of consistency around how to actually

calculate page views to ticket deflections.

There are formulas out there.

I mean, everyone would do it, but...

Customer education as a ticket deflection tool, that language makes sense.

But using a chatbot and measuring ticket deflections from the chatbot, I don't love as much.

And moving away from that focus on, did I resolve what the customer wanted as opposed to just, did I save the resource of a high-cost human being to get the customer to go back to what they were doing and sort of

deflect them as opposed to fully resolving them.

Now, I do think some of the chatbots are resolving them.

It's just not the language being used inside the walls of customer support leadership.

And that is troubling me a little bit, the way that sort of we're moving further away from caring deeply and maniacally about the customer experience.

Yeah, like we want to educate and solve the problem and give them the solution before it touches a person.

the more we can do that.

Because I could see, like, that is a win all around.

It's a win for the company.

It's a win for the customer because they're getting immediately, they're getting solved their solution immediately instead of having to wait to either talk to someone or then re-explain it to someone who then has to like

process it and figure, oh, this is the answer you're looking for.

If it's actually getting resolved, right?

And like I said, some of them are, but if it actually is.

Some are just getting frustrated and going and doing something else.

Yeah.

Exactly, And so I think there's sort of a, I'm hopeful that, you know, one of the changes we'll see as we get more mature as an industry and these tools become

more normalized for us when we get used to them is a refocus on customer experience, is a refocus on customer resolution, customer happiness, and not just deflection.

Let's go here a little bit.

Like I'm curious with some of this, you mentioned it at the beginning, the need and the desire, like how do we get more efficient?

Hey, I've got to have an AI strategy.

What is it?

Like where do you think that pressure is coming from?

Like I'm curious, is that coming from

executives outside investor pressure?

Is it because they're seeing it and it's FOMO, we've got to figure it out and try it too?

A little bit of all of the above?

And how is that playing out?

I think a lot of this is coming from the investor class where they see the language it has been, we went through sort of the digital revolution and then the cloud revolution and the mobile revolution.

And sort of you see these waves of tech that produced these massive returns

for investors, and AI is now sort of the next wave to ride, and they've all placed their bets.

And so they're using their megaphones to say, if you're not using AI, you're doing it wrong.

So there is this sense of FOMO happening for a lot of the leadership.

Everyone is sort of seeing these

LinkedIn webinars or these updates or going to conferences and hearing panels talk about how AI has sort of revolutionized their workplace and everyone's sitting in the audience thinking, God, I haven't really done much with it.

I got to get smarter on this.

A lot of my clients in my consulting practice will just say like, one of my goals, Mike, with you is to get smarter about AI and sort of like know what the landscape is there and how to use it.

So many people feel this sense that they're already behind when

They're not.

Most people are still testing, experimenting, getting a feel for what this space is and should be.

So I think there's that FOMO creation.

I also think there is a natural desire for investors to maximize margin.

And the more that you can reduce cost by reducing the amount of

human capital you have to invest in, then you're going to increase your margins.

So can we be using AI to increase our margin?

Like that's kind of what they're looking for.

It's been the way of the world forever.

And there's such an.

Interesting, it's a hot take, but I like the thought behind it because

These investors have other investments and if they're all investing in AI, like it's a self-fulfilling prophecy at that point.

If we can also motivate the people down the chain to like start using this stuff too, well, you using these AI tools, help these other AI tools I've invested in, get evaluated, you know, evaluations go up, win, win, win.

Yeah, and we're more efficient here.

Yeah, like I love that hot take.

It's something probably most people won't talk about, but

if you understand how the world works and what really goes on, there's a lot of truth to that.

So I think that's, and I would just encourage any CS leader who's listening to this is to sort of make sure as you're making decisions, part of the criteria on decision-making you have is your customer experience.

Like if your customer was in the room hearing about the way you're going to use this technology,

to enhance your team's workflows or to make things a little bit easier for them, what would they say?

What would your customer say about your decision to use AI?

And I just fear so much that the customers are not part of this equation any longer.

Yeah, yeah.

Well, I'm putting on my time as a customer success rep, and much of that revolves around the relationship.

Yes, you want to solve their problem, but like,

Good CS people have a good relationship with their customer.

And so then I start thinking about, well, how does AI help me have a better relationship?

Well, it helps me come to a call prepared better, right?

Maybe it helps me follow up with the e-mail and the follow-up task in a more organized manner better.

Maybe it helps me schedule these conversations better.

It frees me up to spend more time present on the call instead of doing all these side tasks.

Talk about that a little bit.

Like, does that seem like a reasonable outcome?

Does that seem like, you know, if we're thinking about where this can succeed and where it can like not succeed, how does that fit into that?

Or am I totally like off off basis here?

You're right.

You're right.

That's, I think, the dream.

And I can talk about sort of where I think it can be valuable as long as it's

the CSM is then allowed to use these tools to maximize their impact with the customer and not, we've given you these tools, so you have to do less of that.

So now you can do less of that prep time or follow up time.

And so we're just going to give you more customers, which is a reality of what I'm seeing.

But, you know, some of the benefits there that I've seen, even with agent is, you know, I found a agent on the marketplace that

you would put your customer's LinkedIn profile ahead of a call and it would share back with you their DISC profile.

And for people who don't know what DISC profile, it's essentially like a personality four box that tells you sort of the professional style of a person.

The way to communicate with a person, how they like to be communicated and knowing yours, you can kind of see how it matches and yeah.

Exactly.

So if I know ahead of a call,

maybe this is a kickoff call, that my customer is a strong I for a DISC profile or a strong C ahead of a DISC profile.

That's a great way to use AI ahead of a call to really maximize the development of my relationship with the customer.

That's really using AI to maximize your strategic time with them.

But, there are other pieces, like the auto scheduling, great if we can move some of the admin stuff off.

The drafting an e-mail based off of what the conversation was, as long as it can sort of maintain that human experience, sure.

I just worry about sort of getting too templated.

I mean, we've all gotten those AI emails into our inbox and it's like, okay.

It doesn't, the same way that I would encourage CS leaders to get out of the automation into the inbox of their customers,

Because I've never met a customer who was like, oh, my adoption was low until I got that six e-mail drip campaign.

That really motivated me to increase my adoption.

I want them to be more in the product.

It's the same thing where don't get addicted to AI automated e-mail writing because it

the lack of authenticity comes through to your customers and the value of that e-mail is diminished over time.

Yeah, that makes sense.

Makes a ton of sense.

Kyle, I do feel like I'm coming across as anti-AI though.

I want to make sure, like, I do like it.

I think there's a place for it.

I just, there's some cautionary tales I want to make sure people understand.

I think everything you said is on basis.

I think maybe one way we could spin this and the kind of a thing that I'm thinking about to kind of throw it at you in a little different senses

It felt to me personally, and granted, you and I are probably early adopters at the beginning of the curve.

A lot of people are not there yet.

But 2025 was a year of like, figure it out, play with it, you know, see what these things are.

We're not really trying to pull ROI out of it yet.

And everybody says the MIT study where 95% of the, you know, 95% of companies have no value and no ROI in AI.

But that's expected, I thought.

Like, that's good.

That means that people were okay failing.

and trying a bunch of different stuff.

And I think now in 2026, we're starting to see, I hope that we're going to start seeing real actionable use cases and value add.

And I think what we're calling out here is like, don't make these mistakes.

We've already learned the scars on our back from like, it fails in all of these things.

Don't even try to make this work.

But here are the things you can focus on.

And I think that's a really important lesson that it's not that we're being negative Nancy's, it's like we're being like,

just like realist.

Yeah.

And on the ROI side, I think a lot of it is, there's been some magical thinking.

the term AI at this point for me kind of doesn't mean anything anymore because the industry has made it mean everything.

And it's like, it's just, it's magic.

AI is just magic.

It does everything, whatever you want.

And that is where I think a lot of the ROI challenges come out is because it's like, well, I just bought this magic software that should solve everything for me and reduce my cost significantly.

And I couldn't get my data piped into it to begin with.

So like the fundamentals still matter.

Do you have access to good data?

Do you understand what that data is telling you about your customer's experience?

If so, then these tools can be valued.

If you haven't done that,

upfront work to get access to it, clean it up, make sure you understand and have a hypothesis on how to use it, then this magical software you bought won't produce the value that you're hoping for.

Looking into 2026, though, right, if 2025 was the year of like, let's just buy it and hope it works, like if we look at 26 and beyond, you know, some of the things I'm interested in is less like, what are the, who are the big pieces of

like enterprise AI software that are going to win the market.

I'm more interested in sort of the what are the teams going to be building for themselves through sort of vibe coding to be able to like.

produce what they need internally.

Let's go there.

Let's go there.

Because you shared with me kind of in pre-talk about kind of a hackathon that a team you work with, like the buy versus build conversation, right?

And I said, let's just try to hack something together and tell that story.

Because I think it's a really cool example of like, you've got a very bespoke solution that you need.

You could build these things now and not necessarily need to go buy the big expensive software or set up an agent to kind of do some of it.

I had a client and we were

working on their CS strategy, the question comes up, as it always does, should I buy Gainsight?

I love Gainsight.

I think it's a great product.

It does a lot of things wonderfully, but it's expensive.

It's a significant investment that you're making in your team.

And too often I see teams buy it before they're actually ready for it, waste six months to a year

of the SaaS cost while they're getting ready for it, become frustrated with it and dump it.

That was the situation with this customer.

I said, I don't think you're ready for it, but what are the features you're most excited about on Gainsight?

Why do you want to buy it?

So we listed out the feature set.

And earlier in the conversation, I knew that they had a hackathon coming up.

So I said, let's do this.

Give your feature set that you're excited about for Gainsight to your CS team.

Don't have the one-to-one relationship of CSM to engineers.

Let the engineers go do something cool, sort of and faster, without sort of being paired.

Give all of your CSMs lovable credits and have them build, based off of those feature requests, your own custom version of Gainsight.

have one engineer at the end of the day come and tighten the screws on whatever you put together, and they're about five months in now to using sort of this hacky version of Gainsight.

Is it perfect?

No.

Is it sustainable?

Probably not.

But does this help them get to a much better place of extending the amount of time they have where they're not spending money on

on a Gainsight subscription and knowing with a high degree of confidence, okay, we've outgrown our homegrown solution.

We know we're ready for Gainsight and I have my data in a place where I can just plug it in and start to get value right away.

That was a total.

monumental shift for me where I saw, wow, the Vibe coding is now taking 100K deal out of the mouth of Gainsight today.

Like this isn't, oh, AI will impact us in the future from enterprise sales standpoint.

Like it is happening today.

Well, and yeah, and when they do go to implement that, it's going to take them a quarter of the time because they know exactly what they need to set up and exactly how they want to do it.

And they know their workflow and their process and what metrics they care about measuring because

They figured out along the way instead of like, here's everything.

Now you have to make all these important decisions at one time or you're not getting any value.

It's like, well, we kind of worked our way through that.

They're going to be a better Gainsight customer eventually when they buy it.

They're going to be a happier Gainsight customer.

They're going to drive more value out of it because they prepared themselves for that product and know with a higher degree of confidence that they're ready for it.

So I think Vibe coding,

It's a term I don't love.

I wish there was a different term for it, but it's what it is.

I don't decide the terms.

But I think vibe coding and sort of getting more and more people comfortable with creating their tools is kind of the future here, I hope.

And even not just for CS teams, right?

I think probably the biggest impact on CS teams will be products' ability to code quickly and get products to the market.

I think

Another kind of dirty secret of the industry is that the reason customer success teams exist is because product isn't perfect.

If product was perfect, you wouldn't need an onboarding team.

If product was perfect, you wouldn't need someone whose job is focused on reviewing adoption scores and reaching out to help customers use the product better, right?

Whether or not that is a

fountain of youth concept, like a truly attainable concept of a CS lists B2B SaaS product because the product has been optimized and perfected so much that you don't need to hire a CS team.

I don't know if that's a reality, but I think that as CS teams start to provide more feedback to the product teams and product teams can be faster and their output can increase as sort of the barriers of shipping code come down,

That is where I think CS teams will probably actually benefit the most from AI over the next couple of years.

I love that.

I love that.

What level, knowing kind of this is coming in, you can do these things now, like what level of technical ability or kind of data literacy do you think like CS leaders in the teams like need to have now?

it seems like there is a barrier that's being like leveled up here, or maybe not because they can kind of just talk to an LLM or cloud code or whatever and find a vibe code or lovable.

Like, do you feel like that's lowered the barrier enough or people still need to level up to kind of bridge the gap with these?

I think 2 things I think where in terms of the literacy, where I'm asking some of my clients to focus their time on is 1, vibe code.

Just get out there, pick your preferred coding partner of choice.

Pick a, it's kind of like learning Photoshop, where if you just said, I want to learn Photoshop, I'm going to read some of the documentation on how to use Photoshop, you won't learn it.

You need to learn on a project.

You need to create something on photo, have a reason to use the tools and want to understand how they operate.

So pick a non-work topic.

Build a gym app for yourself.

Build a tool that helps your, is an app that sits on your phone that your kids can use to get,

like digital stickers for when they do chores, like whatever it is in your real life, that's a slightly low, like important project that you've thought about, just vibe code that, just to get a feel for what it can do and how to write prompts and how to engage back with it and what's produced and what's possible.

Then I would argue, bring that into the workplace and start figuring things out.

I used to have a philosophy that if I had a problem or I needed to go buy a piece of software to solve something,

I would build out a model in Google Sheets and let the water run through the pipes on that model in Google Sheets for three months, and that would give me enough information to go RFP.

Now it's build a model in Vibe Coda model in Lovable or whatever, and produce that, let the water run through the pipes for three months in the prototype that you've developed, and then go buy something or commit to that prototype, which is now much more possible than committing to Google Sheets.

So I think that's one, is sort of familiarizing yourself with what is possible here.

The other, and what I've found personally as I've done some of the vibe coding work is the way that I think about product development, I thought a lot of it through the lens of the front-end design, and now I'm thinking a lot more of it around back-end database connectivity.

And so I would argue that

if you're working to become more literate in this, is understand the data set that you have, understand how it's stored, understand how data elements can or should relate to one another and how they can be joined to produce more value for your team.

I think if you understand those two elements as a CS leader, probably a leader of a lot of different business units,

you're going to be so much more successful on finding the AI solutions that'll produce the long-term ROI you're after.

There you go.

So tell everybody, let's go take a data schema class and do a little bit of vibe coding and lovable bolt, you know, agent AI, whatever you want to use for that.

And there you go, people.

As we kind of wrap up, any other advice you would have or things that you tell people like, hey, you know, if you're thinking about doing this 6,

12, 18 months from now.

Go ahead and start this now to be ready for that in the future.

Yeah, I think if you're thinking about it now, the one thing that people, I think, which they started yesterday is that data piece, is understanding their data and making sure they're talking to their engineering team and their product teams around instrumenting the data so that they can use it.

And where is it sitting and starting to get access to it?

I think that is

All of this AI tech is going to become significantly more powerful if you have confidence in the data you're feeding into it.

And the two things I hear from every company are, oh, we're so bad at prioritization.

Sure, everyone is.

And our data's a mess.

Our Salesforce is a mess or whatever.

Like everyone thinks that's such a unique instance, but it's a serious problem.

No one has a high degree of confidence in the data they have access to, the consistency of it, or the signal that's telling them.

Start with that, because that is the foundation on which your AI castle will be built.

Yeah, we talk about it all the time, like your framework, your foundation, like garbage in, garbage out, right?

Like if you don't have that right, nothing else is right.

Well, Mike, I really appreciate you taking the time to have this conversation.

We covered a whole lot here.

Tell everybody, what's the best way to find you, to connect with you?

And if they want to go deeper and any CS leaders that want to kind of connect with you and get help or whatnot, how could they do that?

So I run a consulting firm called Harmonic Leadership.

You can find me at harmonicleadership.com.

You can also search for me on LinkedIn, Mike Lemire.

Those are probably the best two ways to find me.

And what I do is I help customer success leaders

develop and optimize their strategy.

I also help go-to-market leaders across the spectrum and founders with up-leveling their executive skill set, focusing on leadership and executive coaching.

So if you need help on either of those, my favorite is when they need help with both, then I can help focus on strategy and up-leveling that head of CS who's not ready to make that VP of CS jump.

That's sort of where I specialize.

And if you want any more hot, spicy takes, like you could reach out to them for them too, because we've given you just a little bit of a taste of that.

And I think there's a lot more we can say.

You just got to keep going to get more.

Awesome.

Everybody out there, if you like this, if you want to hear more of this, like, subscribe, leave a comment, tell us what you think of some of these hot takes.

Are we right or have no problem you telling me we're completely wrong.

That's part of the fun of it.

So do all of those things.

And until next time, keep learning and go tape some database courses, people.

Y'all have a great day.