AI Hot Takes (2025 Edition)

If you listen to our Hot Takes episode from last year, you know that the predictions we came up with were sizzling.

So we’ve decided we had to run it back for 2025.

What spicy AI predictions do we have for this year?

So grab a glass of ice water or whatever you would take on an episode of Hot Ones and settle in.

Hi, I’m Courtney Baker, and this is AI Knowhow from Knownwell, helping you re-imagine your business in the AI era.

As always, I’m joined by Knownwell CEO, David DeWolf, Chief Product and Technology Officer, Mohan Rao, and NordLite CEO, Pete Buer.

We also have a discussion with Dr.

Henna Karna about how companies like yours can make sure you’re solving real business problems with AI.

But first, grab your milk.

Don’t forget to not touch your eyes, okay?

That’s going to be important.

And let’s dive right in and get some Hot Takes.

It is hard to believe, but it has been nearly a year to the day since our last Hot Takes episode dropped.

I know.

We’ve been talking about that Hot Takes episode like all year.

It’s naturally time that we kind of update our Hot Takes.

So David, Mohan, no pressure, but hopefully, we can recreate the Hot Takes energy for this year.

And hopefully, you brought your milk, your ice cream.

What is your, if you are on Hot Takes, what are you bringing with you?

I’m a milk guy.

I love milk.

Hot chocolate.

Hot chocolate.

You’re eating hot wings and you’re going to add hot chocolate.

This, put this on the top of the scale.

This right here that Mohan just dropped, that’s the hottest thing you’re going to hear on the episode today, okay?

He’s the winner.

That is Scoville, I don’t even know the rating.

I think it goes up to like 2.2 million.

That’s a Carolina Reaper.

That’s the kind of hot you just brought, Mohan.

So now that you’ve brought it, let’s keep that energy.

So who wants to go first?

Who’s got their hot take ready?

I’m just waiting for somebody that I can say, oh, that’s just a bell pepper.

Do you want me to give you a bell pepper?

Yeah.

I got a bell pepper.

You got a bell pepper?

Okay.

Yeah, I do.

Okay.

So I think in 2025 that this whole AI-powered, kind of human-delivered style of positioning your product will end.

Basically, I feel like AI was like a hot word that marketers, love y’all, marketers, you’re great.

But it was just like flung around.

So the buzz will go away.

The buzzword aspect is going to…

The AI-powered, that’s not going to be a thing.

We’re not going to see this…

It doesn’t differentiate anybody.

It’s just kind of the…

Yeah.

Right.

No smack dots.

I agree.

Yeah.

That’s a bell pepper.

That’s your palate cleanser.

Is it a red, yellow or green bell pepper?

Well, green is the hottest.

I don’t think there’s any heat in any of them.

Just different types of sweetness.

Don’t worry.

I got my Scoville scale up.

I will let you know momentarily.

A bell pepper is a zero.

So all wrong.

So there you go.

So what’s the scale?

So the Scoville heat units.

So it takes the different peppers.

Bell pepper is a zero.

A pepper chini is between 100 and 500.

Goes all the way up to, let me see, like a jalapeno would be 2,500 to 8,000.

A ghost pepper would be 855,000 to a million.

Like that’s the range.

All the way up to Carolina Reaper, which goes up to 2.2 million.

Well, I’m going to shoot for the stars here.

I’ll let you guys grade me.

I’m saying that data is dead.

I think we’ve been living in a world where data fuels these algorithms, machine learning and everything.

And if you actually stop and think about it, when we actually talk about data, I think there’s two realities that we miss.

The first one is that over 85% of the information in the enterprise is ignored and not even considered data.

And that is the communication that just swirls around and around.

And then the second one is I think that communication is actually a much better representation of the real life that we try to represent with data.

Data has come about because we’re trying to pull out little pieces of information that can represent a reality.

We can do math with in order to fuel these models, fuel predictions, all these things.

And AI is now at a spot where it can actually understand the real world better than the data has ever been able to represent it.

And so my prediction is I believe data is dead and the systems of tomorrow are going to be fueled by actual human conversation as opposed to raw data.

Wow.

Habanero.

That is, Mohan, what do you think about that?

I think that’s a significantly hard take.

I don’t know if it’s accurate, but it’s a hard take, right?

So I think it is, help me out.

It’s four out of five.

Yeah, that’s habanero.

Is that habanero?

Yeah, yeah, yeah, yeah.

Yeah, that is, I think that is spicy.

Can I ask a follow up question though, David?

How this data is dead?

Is this a this year, five years, 10 years?

What’s the horizon for the hot take?

When you set the rules for the game, you did not give us a horizon.

No, no, I didn’t.

I didn’t, but I would like to hear when you’re saying that.

I think this is an eight to 10 year thing.

I think we’re going to start to see signals of it.

I think we’re going to start to see some progress towards it, but it’s probably an eight to 10 year thing, and that probably means that in reality it’s more like 15 years.

But yeah, I’m going to say eight to 10.

So that also means that the fighting about whether or not a certain piece of data in an executive team meeting will also go away.

Wouldn’t that be wonderful?

It would be so wonderful.

Mohan, are you ready?

What are you serving us up?

So I’ve been thinking about deep seek and in general the open source models out there.

I think the open source models are such a double-edged sword that they’re just going to diminution value and go away because they’re open source.

I think open source models by definition are so ripe for commoditization.

All right, everything about that can be seen, known and therefore, the value it provides is going to be so different from traditional open source software that it won’t be a leading edge player as we go along because it may have mass use, but I don’t think it’s going to build competitive advantage for the company.

I’m going to name that a jalapeno.

I think that’s hot.

I think there are some people that would disagree with it.

I think there’s several people out there that believe that though as well.

What I’m trying to parse and I can’t figure out is, there are some people that just don’t like the open source model, especially these proprietary folks, and so they’re always throwing reasons why it’s going to fail.

I don’t know if they actually believe what you just said, or if they just say what you just said, because it’s one of the 13 reasons why they say open source is going to fail.

So that’s the piece to me that I’m pondering a little bit is, yeah, there’s some interesting rationale there.

I think you’re right.

As I think about it, there are differences between open source software and these open source models.

But I’d never thought about it deeply that way like you did.

And so that’s, yeah, that kind of burnt my tongue a little bit in the back.

Good thing you’ve got that hot chocolate.

No, that’s Mohan.

I’m drinking milk.

I’ll throw one out there and I’m going to stretch on this one.

You might be laughing at me in a year or two.

But I think in the next two years, Open AI is going to implode and all of that equity value is going to evaporate.

I think that they are just playing with fire.

I think number one, the LLMs are going to be commoditized.

I think you see it with DeepSeek.

I don’t believe a lot of what I hear and read about DeepSeek.

I don’t think it was nearly as efficient to create as they say, and I think there’s different things there, but it doesn’t matter.

What it’s done is set a new bar and it has pushed others to innovate further, and ultimately, the LLMs themselves will be commoditized, and they’ll be commoditized more and more, and they have built that business on the commoditization, and they’ve raised so much money that I don’t think that is going to be the asset, and I think there’s plenty of other players.

I think the first mover advantage doesn’t play out here.

I also think that there’s been so much chaos with the leadership, and this transition from a nonprofit to a for-profit, I think is going to cause them so many more problems.

We see Elon Musk making an offer for the assets.

I just think we are going to look up in two years, and OpenAI isn’t going to be the clear leader anymore, and the assets will still exist somewhere, but they won’t be seen as the leader.

The assets will be somewhere distributed.

I don’t know if they’ll get acquired.

I don’t know if they’ll merge somewhere.

I don’t know if they’ll go back to being a pure not-for-profit, and the profit engine will be spun off somewhere.

I just think it’s going to implode as the AI stalwart.

Do you feel like that is a good thing or a bad thing?

I’m just curious.

I think it’s probably a good thing.

Is it because Sam Alman kind of gives you the Ick?

That’s me.

Yeah, I don’t have a strong opinion.

I don’t have too strong of an opinion on that.

I think they’ve done a lot of good from just pure innovation standpoint of pushing things forward.

But I guess the part I struggle with is the drifting away from the original mission, right?

When you are a not-for-profit and you’re founded that way, to drift away from that feels a little bit disingenuous at best to me.

And it makes me question some of the underlying principles and character of the entire operation and wonder, can they be true?

And in a world where trust matters so much with AI, will people ever really trust them given that reality, that they’ve been able to switch and pivot from their expressed original intent?

And so, yeah, I don’t think it helps them.

So David, let’s say your hot take comes true, even partially true.

Play Satya Nadella for me.

What do you think is Microsoft’s signature move if your hot take is going to come true?

You know, it’d be really interesting to read those investment docs and see what they actually have, right?

What kind of controls they have and different things.

I mean, I’m really curious and even Elon Musk too, right?

I think, you know what I can really see happening with the state of things right now is I think Satya and Elon get together and do something together as partners.

If the two of them weighed in together, I think that wins the day.

You know, if you think about the largest models out there, right, so there is, I mean, you know, there is OpenAI, then of course, Gemini has moved up really well, right, so in every sort of thing, there is the Lama open source from Meta, right?

And then there is Anthropic.

So it kind of falls off very, very quickly, and right?

It does.

You know, whenever this question comes up, I always wonder about what’s going to be Microsoft’s move, because they’re not in the list.

Well, and interestingly, Apple isn’t either, and everybody talks about how Apple is behind.

People don’t think of Microsoft is behind because of the investment in OpenAI and how they have rights to that IP.

But I actually think there’s something to, we are seeing more and more the LLMs get commoditized and I really believe they’re going to be.

And I think Apple’s move has been to, we’re going to be better than anybody at using the LLM, not creating our own.

And I like the differentiation there.

I think that’s smart.

You actually see Gemini starting to get on that bandwidth as well though.

Like the way Google has rolled out Gemini, so it’s not just this LLM, but it’s embedded in their products.

I mean, they have, in my mind, in the last month and a half, they have commoditized the video transcription and summarization market in like so fast.

Like it’s gone.

If you are a standalone transcription service for video calls, you’ve got no hope, all right?

If you don’t have some really differentiated domain expertise or something that you’re putting on top of that, I think you’re in trouble.

Google has driven that by saying, hey, it’s not just about the LLM, it’s about how it’s integrated into product and how it’s driving end-user value.

That’s where Apple started out.

I really see Google poking its head up as the leader in the space right now because of that, whereas OpenAI is more about the LLM purely itself, and I think Apple is at the onset and just getting started and focusing in the right place of where they can differentiate and add real value versus just providing another commodity.

Everybody else has products to integrate the LLM into, OpenAI does not.

That’s right.

Right.

So yeah, that’s going to be the differentiator.

And I wonder if Microsoft is going to just acquire OpenAI.

It’s always interesting to see what the head of Microsoft AI says.

I think his name is Mustapha Sileman.

And Mustapha’s tweets are always very interesting because he’s walking this tightrope.

By the way, just on the Gemini transcription, their one problem is Google Meet is so behind on video.

It depends who you ask.

I know you’re a Zoom lover, but there’s a lot of people that prefer Meet.

Hate to say it.

I’ll believe that when I start going on more external calls where people are using Google Meet.

I think we’re the only one.

Actually, I’ve been really surprised how Meet’s market share has steadily increased.

I mean, still Zoom is the leader for sure.

But Meet’s been moving up.

What is it now?

Do you know off the top of your head?

I think Zoom is 55 percent, so it’s very large.

And then Teams is like…

Google Meet is up to 31.38 percent.

That’s shocking because, again, I am pretty sure we’re the only ones using it.

But listen, we got a bad Gemini.

That’s awesome.

And it’s smart.

You know what I mean?

We don’t have to pay for Zoom.

So, hey, that’s everybody’s, the opposite of a hot take.

Don’t pay for…

Okay, David, Mohan, everybody listening, take a big gulp of your hot chocolate.

Hopefully, you’re cooled back down now.

Hot chocolate.

And thank you for this episode.

Don’t worry, we’ll keep talking about these hot takes periodically, I’m sure, throughout the year.

Thank you, too.

Appreciate your thoughts on this.

Adios.

Bye.

Everyone wants to keep the clients they worked so hard to win in the first place.

And it makes sense.

It’s much easier to grow with your existing clients than it is to win new ones.

That’s why we created Knownwell, to deliver real time insights into things that impact customer attention, like service quality, perception and commercial alignment.

Say goodbye to guesswork and managing based off your gut.

Visit knownwell.com to learn more and let us show you your data in the platform.

Go to knownwell.com.

Dr.

Henna Karna is a tech leader who has built AI products and teams for companies like Google and AXA.

She’s currently a Harvard Fellow, and she sat down with Pete Buer to talk about solving real business problems with AI.

Henna, hi, welcome to the show.

So nice to have you.

Hi Pete, how are you?

We’ve been looking forward to this conversation.

So you are a Harvard Fellow, you’ve held leadership roles working in data and AI with Google, AIG and more.

Can you give us a quick introduction to your work and where AI fits in?

Sure, a quick view of it would be three concentric circles if you can give it that way.

The world of digital where we’re looking at behavioral analytics, human interaction with technology, the world of deep tech.

So what does it look like when you scale analytics?

What does it look like when we’re applying security protocols across the value chain of a company?

And then I would say the third is much more cognitive.

The connection between how our ability to discern information, whether it’s real or fake or the longevity of that information.

Sort of that’s the area of focus I’ve been on most recently.

But connecting those three things are really what I’ve been doing the last, I don’t know, 28 years.

One of the biggest concerns we have when we’re talking to executives who are going through strategic thinking around AI transformation is that too often it’s treated like a solution in search of a problem.

Do you have recommendations for leadership teams to make sure that they’re taking it in the right order and solving the right problems with AI?

That’s a great question because that comes with hype, right?

Anything that’s hyped up so much, we want to not be behind, so we try to find a way to fit it in, and sometimes that is a solution looking for a problem.

I would say that the most critical thing is our organizations, if we can trust our planning process, if we can trust the fact that we do know how to prioritize, what is the most important for our customer, what makes us differentiated, and taking those priorities which we do year on year, organizations do this every year, and they spend months doing it, looking at that list of things that already exist, and then applying an AI lens to that list.

Said differently, we don’t want to add, if I have top five priorities in my company, we’re not trying to add a sixth one that says we’re going to use AI now.

Because that is already a strain on the company as a whole.

So we already have five things, assuming we’ve budgeted for it, we’ve maintained the workflow around it, we’ve got people working on it.

And then we add a sixth one, it’s going to be begging, borrowing and stealing to get that sixth one up and running.

All eyes are going to be on looking at it because it’s straining the back of the organization now with a net new thing to worry about, which by the way, we’re not all very well equipped to know yet.

Most typically in such new things, people are learning on the go.

So I think the most critical thing is to not add an additional to do, but to look at what we have and then say, okay, pre-AI era, we were doing it these ways.

Now, with the new AI capabilities, we’re going to augment it x, y, z different ways, and that augmenting equation is really what we’re spending time on.

So don’t strain us even further.

Let’s just look at it with a discerning lens.

Do we need to do these five steps still?

Are these the right five steps in this particular prioritization or not?

Got you.

Okay.

So it’s kind of trust to your current strategic planning, operational planning processes, and use that as the mechanism for then figuring out against the priorities that we’re trying to tackle, where does AI fit?

Do you feel like leadership teams are aware enough or educated enough on AI and how it can help in order to be successful in applying it in the strat planning process?

That’s a great question too.

Yeah, so the caveat to my answer earlier would have been that, how well equipped are we in our technical literacy with the way that we’re headed to be able to look at that in a different lens?

So, no, we’re not very well equipped, but I think everyone is in that state.

We’re not behind anybody else yet.

There’s a short window where we don’t feel like we’re behind industries.

Some industries are a little bit more ahead than others.

More B2B industries are a little bit lagging than the B2C industries.

But I would say that skill set, first and foremost, for something like artificial intelligence is different than cloud computing.

Cloud computing was a very pointed area of focus, where the technology departments and the data departments, they were really focusing on that, the system security departments.

AI is a way of working.

That’s everybody.

It touches everybody.

So really looking at how are we getting everybody’s taxonomy to be lifted in the company?

Are we speaking the same way?

So that we don’t have miscommunication, misfiring, all sorts of different places.

So first and foremost, we’re learning a new language.

Can we learn it together?

Can we get on par together?

I’m championing with boards that I’m working with right now to try to get everyone to be trained on fundamental artificial intelligence.

Everyone from the HR department to the accounting department to, of course, the technology teams and the P&L owners.

That’s a key thing.

To the board, right?

And to the board.

That’s right.

To the board.

No one left behind because this is a new era.

No one’s off the internet.

No one’s off the social media.

Well, I don’t like it so much, but no one’s unaware of it.

And we shouldn’t think of it like that for AI either.

Have you seen best practice in getting an entire company to a level of technical proficiency on a new technology like this?

New way of working?

Only very small ones.

They already had momentum down that path.

So taking a slightly fork on the road is not so complicated.

Where organizations are multi-threaded and multi-dimensional and lots of factors come in play, meaning even their customers have to be AI enabled for them to be AI savvy.

There’s a little bit of the ecosystem who goes first.

That’s definitely happening.

But I would say that the companies that are doing super well, Pete, are the ones that are fully aware of where the customer is intending to go.

And then leaning into that, not necessarily saying, I’ve got to fix all my documentation issues.

Well, we’ve been doing that for years now.

Now it’s really much more, the gravity of the situation is so much higher.

And working with the customer on where the customer wants to be, that’s really going to give us a prioritization that’s meaningful, that’s going to sustain the times.

As an all-thing, start with the customer, huh?

Yes, that’s correct.

Don’t forget to start with the customer, even with the tech.

Okay, so shifting gears, cost has been a hot topic, especially these past couple of weeks with DeepSeq and maybe the massive cost differential that it represents.

How can leadership teams get past the concerns around cost and actually flip the script and think about AI as a way to bring cost down?

A more non-technical answer to that, I think, because there are technical answers to that, but I would say that the most non-technical answer is, I don’t know of any company, at least that I’ve been working with, that doesn’t have a bottleneck, a series of bottlenecks.

They are not over-saturated with people that have so much time on their hands.

So I think hidden costs have always been a dilemma in organizations.

They’re harder to look through.

Data is a severely complex hidden cost, which AI can then easily resolve.

Really being thoughtful about the fact that we are not as optimal in our expense ratio in most industries.

We weren’t so fine-tuned before artificial intelligence was invented, or not invented, but rolled out broadly, even in artificial intelligence in 1994.

But I think that ownership of ourselves, how much can we drive this ship correctly more towards the fact that we now know better about our technology and how it can be leaned into, that’s going to be more powerful.

And as that’s happening in companies, that’s more internal than the external factor of the storage, the compute, the copper, the materials required, all of that is going to start to solve.

And so we want to be ready as a company when that stuff is ready, but we don’t have to wait.

And we certainly know that that cost is coming down.

So the food of our labor is going to be much more meaningful as soon as we can fix our internal expense ratios, ready in time for that scalability and sustainability of it going forward.

I guess companies’ abilities to realize the return on their AI and related work stream process changes sort of dependent on the degree to which they can get teams adopting, accepting using new approaches to work AI-powered solutions.

What have you seen in that space?

How are companies succeeding in getting folks to pick up the ball and run with it?

There’s several examples.

I mean, there’s quite a bit on the security-compliance side of things, on the legal side of things.

One that I really like, and it goes back to the customer.

But to be honest, customer identity, know your customer.

I think those are great use cases that with the empowerment of artificial intelligence in our architecture, we can do this frictionless.

We can do this in a much more meaningful way where we are really getting the 360 view of the customer as we’re trying to enable the customer’s growth, as we’re trying to be more valued with our partners in our ecosystem.

I think an example right now is there’s an organization that I’m supporting and consulting with.

I guess it’s consulting, but it’s really a little bit more than that, where we’re designing out the net new construct of what a customer looks like.

What’s that bi-directional 24-7 exchange that I can have with that customer, such that my diary of the AI agents, which are really just task masters, how are they going to empower me further in my moments of engagement with the customer?

That preparation, but also more than that, the benchmarking, the assessments, the things that took me hours to do pre-call.

Now, I can actually be more thoughtful about it when I’m in a meeting with a customer.

I’m able to really get them to be the priority.

Not that segment of the market, not that potential of who could be there years from now, but it’s very much about them and us and how to be improved together.

That use case where it’s starting to bridge the edge of the network and making the edge of the network more powerful for our organization.

We could do it, but we could do it in bootstrap models.

We could do it in one-off.

The top 10 customers would get that kind of support, but the remainder could have been top 10, but we wouldn’t have even known.

Now we have that ability to really roll out the red carpet and talk to those that have a mission that’s aligned to ours and can go forward together.

I think that’s going to be amazing.

You got me.

Realizing that the uptake and acceptance question doesn’t just apply inside the business, but it’s a customer question too.

Is one or the other, the harder of the two to accomplish?

I’m finding it more and more surprising, but it seems to be that the customer, it’s a little bit of a stalemate mode where if the customer isn’t ready, what do we build for?

If we build for it and the customer isn’t able to adopt it, we’ve just shot ourselves in the foot.

So there is a little bit of that, how do I play the chess game not too fast?

I can compromise some works, but I got to be careful about how fast I go because the industry has to comply with that.

So there’s a bit of that.

That I think will iron itself out, to be honest.

The more expedient our awareness of the technology gets, the more cost-effective it gets, the more prepared we are internally.

We’ve wired ourselves more appropriately.

It’s going to be an incredible journey ahead because of that.

Thank you.

It’s been delightful to have you on the show.

We’re so grateful for the fact that you’re willing to be here and share your insights and appreciate it.

I appreciate your time.

Thank you, Pete.

Thanks as always for listening and watching.

Don’t forget to give us a review on your podcast Player of Choice.

We’d really appreciate it if you can leave a review or even share this episode on social media.

At the end of every episode, we’d like to ask one of our AI friends to wait in on the topic at hand.

Hey, Jim and I, how’s it going?

Welcome back to the show.

This episode, we’re talking about some of our hottest takes on what to expect in AI in the coming months and years.

Any hot takes you care to share?

I think AI is going to get way better at understanding what we mean, and we’ll see it showing up in some really surprising ways.

Plus, I bet we’ll have some serious conversations about what’s real and what’s AI generated.

And now, you’re in the know.

Thanks as always for listening.

We’ll see you next week with more AI applications, discussions and experts.