Where’s the VC Money Going in AI?

A record $24 billion in VC investment flowed into AI startups in the second quarter of 2024.

So where did all that money go?

And what’s the significance of VC confidence in a technology like AI anyway?

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

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

We also have a discussion with James O’Brien of DECI AI about their AI-powered customer support and knowledge management platform.

What’s the significance of VC money pouring into AI, and what does it mean for your business?

I sat down with David DeWolf, and Mohan Rao recently discussed that very question.

David, Mohan, the second quarter of 2024 saw a huge influx of VC funding into AI startups.

I know you two were both paying attention, with more than 24 billion being invested.

So that’s a lot of money, and a really good thing when you’re in an AI startup, which the three of us all work for.

But today, I want to talk about where the money is going, and why is this level of investment important for business leaders to understand and have context for?

So why don’t we start with why this is important?

You know, the way to think about where the money is going, you know, it’s probably best to think of it as in three broad buckets, right?

Because if you think about where the foundational models are, there is going to be a handful of players.

It’s really, really, really, really hard to compete with them.

Because while some of these foundational LLM companies look like startups, they’re all backed by the giant, right?

So they’re really not startups in that sense.

So it’s also, there’s a growing realization that the general purpose AI models are not a sustainable business model, right?

So you just can offer chat interfaces, right?

That are general purpose.

So you’ve got to build something that’s specific.

So the way to think of this is there’s money going into vertical AI applications, right?

So these are specific for specific sectors.

There are horizontal AI applications.

These are multiple industries, multiple use cases.

And then there is AI development and infrastructure tools.

So if you just look at it from, if you remove sort of the big ones that you hear about all the time, money goes into these three buckets.

And in these three buckets, roughly it goes about 40% into vertical AI, 30% into horizontal AI, and 30% into AI development tools.

That is sort of my cheat sheet of how I think about where the money is going.

What do you think, David?

I think that’s a great frame.

I want to give a couple other frames, okay?

Because Courtney started off with the comment that 24 billion and obviously that’s a big rebound.

We’re getting back to more and more funding.

Absolutely agree with that.

One-fourth of that, 6 billion went to one company, right?

And it’s XAI, right?

And that’s Elon Musk.

Let’s put that in perspective and just realize it’s a big deal, right?

The 24 billion, but subtract the 6 billion, it’s not as big of a big deal.

But I think we should just be aware of that.

And anytime there’s an outlier, you have to take it out of the picture in terms of trending, right?

So that’s the first thing I would say.

The second thing, I love the question that you had, Courtney, about why does it matter?

Like how do businesses, if you’re not a startup, why do you care, right?

If you’re not in the VC business, why do you care?

I think you care because this is where trends begin.

And it’s not just they begin because somebody says, I’m going to start here, it’s because they’re actually finding smaller trends and different things to base their investment thesis off of.

And it is the beginning of the monetization cycle of AI.

And so I think as you heard Mohan say, there’s different categories where it’s going into.

There’s the foundational models that have already gotten their funding.

Those are billions and billions of dollars.

Nobody’s going to compete there, but it’s the application of AI that I think is really benefiting from this wave of new investment.

And what I love about it is that this is actually the second wave.

You had a first wave that was the emotional wave of AI has been democratized.

Oh, my gosh, I can’t lose.

So everybody started throwing money at tools.

Then you had a pullback of, oh, my gosh, all of those blew up and they were commoditized.

What are we doing?

Now?

I think you have more thoughtful, intellectually honest investments that are based on ROI and really not just the emotions of investing.

But all VC investing is always assuming a one or two or three out of 10 hit.

So there’s going to be losers here.

But I think it’s a lot more thoughtful than that first wave that we saw in the early days of 2023 that we then saw the pullback from.

I think that part of the trend matters a whole lot in this conversation.

Yeah, and extending that thought further along, you know, for businesses who are not in AI, but do other productive things, right?

So you can think of it as how is the buy landscape evolving and where is your build landscape, right?

So you got to look at your business and say, here are things that I got to build because nobody can build this or will build this good enough.

But here are the places where there is investment going and likely I will be able to buy and able to complete my portfolio that way.

So it’s important from knowing the trends, as David said, and what it means for companies, what it also means for everybody else in terms of future build buy decisions that are coming.

The other piece of that that I would say in terms of build buy, I love that comment, Mohan, because I think we’ve talked a few times on this show before about being careful not to build something that you can buy six months from now.

And I think there’s probably a lot of risk right now for enterprises who are looking at their investments and saying, okay, I want to use this technology, let me build it.

You called out specifically, there’s a third of the money going to tooling.

So that’s empowering the folks building things.

It’s selling the shovels to the gold diggers.

You then have horizontal AI and vertical AI.

I think that vertical AI specifically is going to be one of the areas where you see more buy than build, and you’re going to see a next generation of platforms, a next generation of products, a next generation of tooling that is really providing very viable options for companies to buy, to be able to leverage AI in solving real business problems versus having to build it from the ground up.

David, I know you have been out there talking to a lot of VCs as you’ve been pursuing our own funding rounds for Knownwell.

What have been the takeaways for you as you’ve been just talking to people out in the community?

I think the big one, it’s actually a little bit shocking when these numbers were released.

The talk track in the market right now is that it’s still a brutal fundraising environment.

It’s really hard to raise.

I think what’s interesting about this data is it shows that with the right investments in AI, you can raise capital in an otherwise capital constraint, not constrained because there’s plenty of market out there, but just hesitant market because of what’s going on.

But I do think that what VCs are really looking for is first and foremost, true differentiation, right?

There are a lot of people, a lot of entrepreneurs, a lot of companies that even got funded in 2023 around a thin wrapper around ChatGPT, right?

It’s basically a user experience that will be commoditized very, very quickly.

I think really looking for what is the sustainable competitive advantage that this company has.

I think domain expertise matters a lot.

I think true intellectual property development matters a lot.

I think really not just using an LLM to do something cute matters a lot.

And then I think there is starting to be this grappling with what is a true AI first company versus a traditional SaaS business that happens to have functions and features using AI around the edges.

Right.

And so those are some of the different talk tracks that I’m seeing.

And then in all markets, it matters.

I think it really matters right now because we’re so early on in the AI that all of those things are de-risked when you have a team that has been there, done that before.

And so where I’m seeing funding happening is where there’s a truly differentiated use case that solves very real problems that has buying signals and there is a team that is driving that business that has navigated other cycles before.

And to me, that’s where the winners are.

And I’d be willing to bet if we dove into that $24 billion, we would find a preponderance of that.

Of course, $6 billion of that is Elon Musk and it kind of proves it out, right?

Audacious problems and a proven entrepreneur.

So those are some of the themes I’m seeing.

Mohan, what do you think this means for executives when we are thinking about ROI?

Does this mean there’s some light at the end of the tunnel when we’re thinking about getting ROI from these new AI tools and platforms?

Yeah, it’s increasingly possible to dream about making ROI calculations in the space, right?

So because you’ve got tools that are coming, you can see the horizontal AI tools that are out there, whether it’s the note taker that we all use, those are maturing very, very, very quickly.

There are horizontal tools in marketing, in software development, so on and so forth.

You can see that.

You can really derive ROI from the usage of those.

The vertical apps that are purpose-built for specific problems are also emerging, and that is the dominant category of where funding is going.

So the executives should look at it because even in our own experience, every six months or so is a new era, it seems, where you have to reinvent what it is that you’re doing.

So for a company that’s focused on other core use cases, you can’t be reinventing your AI platform every six months.

So you should be paying attention to which vendors can provide those.

And then take the sliver of use cases that you really think you should invest in, that are your build decisions and invest your IT team in building those.

So that is increasingly a good set of vertical and horizontal app companies.

The important differentiator here is these are GenAI applications as opposed to GenAI checks.

Mohan, there is a key word that you said in there that I’m really curious about, because when I think about VCs, you’ve hit the broad categories, but you use the word category, and there are different categories of vertical software, for example.

Do you think AI is going to drive a brand new set of category creation, or do you think it will more reinvent existing categories?

I think it’s going to reinvent existing categories, but reimagined in a totally new way.

Reimagined as an AI native application, where data and algorithms are at the core of what’s been built, and built in a very specific way for the customers to solve very, very specific problems.

So I think it’s going to be, you know, when you think about vertical categories, you think of healthcare, you think of, you know, these are, I don’t think there’s going to be new products as much, there could be, but vast majority of them are going to be re-invented, like never before, that could not have been imagined before.

So it’s going to be the same category, but entirely different new level.

What do you think?

I think you’re probably right.

You know, I don’t know, I had never thought about that question until you said the word.

And I think there is probably something to step back and learn from history here.

Creating a new category just from a company perspective is really, really hard.

I do think there are waves of technology that are so big that they tend to have the collateral damage of creating new categories and destroying categories.

And it strikes me that AI is big enough for that to happen without a doubt, right?

I think it’s one of the most transformative technologies of our time and probably just like the Internet created new categories, so too will AI.

But I think you’re probably right that there will be more categories reinvented, and it will be the minority that are brand new.

I think this is a really interesting topic, and I think for all the executives listening, really important to kind of keep some semblance of what’s happening in this world, because it really informs what is happening on the horizon and what is coming your way.

So hopefully you found this discussion helpful.

Mohan, will you just hit again those three different areas that you’re seeing funding go to, because I think that’s a helpful take away here.

Yeah.

You know, if you remove the general purpose AI models out there, whether they’re open source or they’re run by one or the other companies, if you take them away, they’re broadly in three categories.

There is vertical AI applications, there is horizontal AI applications, and then there is AI infrastructure and development tools.

Right?

And broadly from what we’ve seen in the last quarter, it’s about a 40, 30, 30 spread of investment.

Yeah.

It’ll be really curious to see if that continues, and we’ll keep you up to date on this podcast.

David, Mohan, thank you.

Sure thing.

Acquiring new customers is more time consuming, costly and resource intensive than keeping the customers you already have.

That’s why we’re building an AI platform that warns you about at-risk clients and gives you actionable intelligence on what you can do to prevent surprise churn.

Go to knownwell.com today to learn more and sign up to see what we’re building.

DeBuer is back with another installment of AI in the Wild.

Hey, Pete.

Hey, Courtney.

How are you?

I’m doing good.

This week, I wanted to highlight a press release from Accenture and Google Cloud.

The release shares the details of a strategic alliance between Accenture and Google Cloud, and advancing solutions in generative AI and cybersecurity.

Pete, can you help us break down what’s happening here?

You betcha.

So one of the big takeaways for me is in the subtitle, which goes 45% of joint client projects have moved from gen AI proof of concept to production.

It is so fashionable these days to be skeptical of AI because of protracted timelines to ROI.

I think this article and the reality of the joint work between these two companies and their portfolio should put a little wind back in the sales of the great AI work that’s going on across businesses.

So I know it’s a press release, and so we sort of have to take it with a little bit of a grain of salt.

But if even close to 45% of the joint enterprise customers that are working on proof of concept are committed to taking their ideas to production, that’s a big deal.

It might be helpful for the ROI skeptics too.

Corporate large enterprise CFOs tend not to be loose with the ink when signing off on investments like the ones that we’re talking about here.

The press release is helpful in that it offers examples of work that the two companies are doing that span applications.

So everything from supply chain to customer management and sales to cyber security and industries and they even give a couple of case examples.

Brazilian bank that’s reinventing their CX, yielding communications that are 80 percent faster and 100 times more personalized.

I don’t know if the CFO is going to sign off on that as a hard metric, but that was their offer.

And then Radisson Hotel Group boosting advertising team productivity by more than 50 percent and increasing revenue more than 20 percent.

Now, that’s something if I was CFO, I would sign off on.

So anyway, indications to me that as we’ve been saying all along, the ROI on AI isn’t a question of whether but more one of when.

Pete, this is really encouraging news and I think for everybody listening, hopefully it does put a little wind in your sails for the projects that you may be pursuing.

Thank you again, Pete.

You bet.

James O’Brien is a co-founder of Ducky AI, a investment-backed AI startup, helping companies revolutionize their customer support and knowledge management functions.

I sat down with him recently to learn more about Ducky AI and the problem they’re solving.

James, great to have you on AI Knowhow.

Thanks for joining us.

Yeah, thank you so much for having me.

It’s great to be here.

Yeah, always good.

I have to give a little shout out because I’m based in Nashville.

You’re based in Nashville, so shout out to being in Nashville.

I mean, it’s the little things that sometimes rise to the top.

I hear it said that it’s the new AI Hub.

That is what I have been hearing as well, so I’m glad we could get that out there.

Yeah.

Everyone, Nashville, it’s the place to be for AI.

So glad we got that out of the way.

So can you give us a little introduction to Deci AI and your role?

Yeah, for sure.

I’ll start by saying that I kind of fell into this by way of a very circuitous entrepreneurial background.

Like so many people, I moved down to Nashville, Tennessee, singing in a band, had an idea that I couldn’t get out of my head, turned it into a company, failed at that company, got lucky enough to join early at another company that kind of went not kind of went super up and right, and then did a couple other startups along the way.

And then with Ducky, I actually, I’m grateful to say, found my co-founder before we found an idea.

I was in crypto for a long time.

I never really thought that I was going to get into AI, but there’s really two things that matter to me.

It’s working with people that I like and trust and working on interesting things.

And AI seems to me, and again, anybody feel free to disagree, but the most substantial technological shift of our lifetime, most likely, and if that’s not interesting, I don’t know what is.

So anyway, when we were looking around for an idea before Duckie was Duckie, we ended up talking to just a ton, a ton of people, and everybody pretty much said knowledge is a huge problem.

Knowledge is all over the place, it’s fragmented, people are not on the same page.

How do we make this better?

How do we make this easier?

We have all these tools that are supposed to make our life better, but in fact, we just end up context switching between a million different browser tabs.

We took this idea, which I’m sure as you can imagine, is incredibly broad, and we looked for better founder market fit.

I’ve done a lot of customer support in my life for my first company, for the second company I was a part of.

It’s just a very logical and useful application of the technology.

Hong and I created Ducky specifically for customer support.

Hong is my co-founder, is a machine learning developer, so he runs everything that is technical and product.

I’m grateful to say now with a small but mighty team, and I run everything that is not technical or product, so sales, BD, marketing, operations, all that slack morale, all that good stuff.

I would love to talk a little bit more about what you’re building.

How did you discover?

I know you said you were in customer support.

How did you go about really honing in on the problem and starting building the platform?

Yeah, excellent question.

So the problem really just smacked us in the face, quite honestly.

We interviewed hundreds of people within the world of business looking for an idea.

So Hong and I, my co-founder Hong and I, have always been really fascinated with this concept of the human-computer interface, which is like we have all the software and now things are changing a little bit since AI has hit the mainstream and these AI wearables and things of that nature.

But for the most part, you’re either dictating to or typing into a product, you’re clicking around and we were like, there’s got to be a better way for humans and software to interact.

So that was one of our core theses always.

A big part of that, again, as we started to, and I kid you not Courtney, like 9.5, I know this isn’t how people work, but 9.5 people out of 10.

When we interviewed everybody, it was like knowledge is a problem.

It’s all over the place, it’s fragmented, like I don’t know what to find, it’s gone.

I’d write it in Slack, it’s gone even though it’s super useful.

We were just laser focused on this concept that if something has been mentioned before, if a problem has been solved before, why do we have to go search for the information to solve it again, or take up somebody else’s time to ask about the problem and fix it again?

Why can’t we have something that delivers the answer to us that we don’t have to ask?

I’m not talking about a chat pod here where I’m like, hey, Ducky, what’s the deal with this?

Don’t get me wrong, talking to your spreadsheets, that concept is really cool and it’s incredibly powerful.

But what we thought was even more powerful was in the case of Ducky, like I’m working on a customer support ticket, right?

Like somebody e-mailed me.

Instead of me having to go ask one of my coworkers or even ask a chat pod, why can’t a software solution just tell me, hey, based off of everything I know about your business, and I probably have a better recall on your business’s information than you do, this is the information you need to solve this thing.

That’s really the initial idea that we started working on with Ducky, and it’s really the core architecture of everything that we do today.

Right now, it’s a co-pilot, so it sits as a Chrome extension next to any support ticketing system that you use because we, I suppose, were reasonable enough to think that we’re not a multi-billion dollar company.

As an early stage company, if I went to you and was like, hey, I know your whole business runs on Zendesk, but you should rip it out and you should replace it with us.

We just didn’t think that was super reasonable, so we wanted to build something that was very universally useful to any team, really no matter the software stack that they were using that could deliver the right information at the right time to cut down on that whole research process and hopefully instill alignment for teams across the board.

That’s amazing.

You’re saying eventually, would you consider that you are moving to an ambient technology?

That’s a really great word that I have never thought of in this regard before.

Yeah, I think of, I usually talk about it more like something like living in the ether.

In fact, we’re about to roll out a new version of the software that internally we call Invisible Ducky because it’s always there and ready to be used or ready to give you something interesting, but it’s not always in your line of sight, so to speak.

Yeah, I love the way you said that.

That’s very much in line with what we’re thinking.

I think the problem there is that we’re all hard-coded to think in terms of like, what can I click to do the thing I want or what UI can I interact with in order to do the thing I want.

Maybe the learning curve will be a little bit tougher, but hopefully it can just become more and more intuitive over time.

I would love to talk a little bit about what some of the mechanics behind how Ducky learns and how it decides, like does it give you options on what tone it uses or which flavor of support that it gives in different situations?

How is it mechanically built?

That is a wonderful question and something that we wrestle with a fair bit.

Right now, and this is a quote that I’ll steal from my co-founder, Hong loves to say Ducky is very opinionated.

That means that Ducky does something that it believes is in line with the way that your company does something.

One of the first things that we built to speak specifically to one of your questions is what we call a tone engine.

By looking at all of your previous customer support interactions, Ducky creates a tone profile and that’s anything from warmth, professionality, length of message, even emojis.

I never thought we’d build an emoji engine and it’s one of the first things we built because it’s one of the first things that one of our customers cared about.

Ducky makes that choice for you that is in line with what it believes to be the best examples of support that your team has done in the past.

Moving forward, the things that we would like to do to make that better are firstly, we just rolled out something internally that we refer to as a grading system.

Essentially, every response that Ducky sends out, it then cross compares to the actual message that gets sent.

It’s like Ducky generates a message it thinks you can send, but then it cross compares it to the message that the human being actually sends.

Then it learns from the differential essentially between those two things.

To go back to your other question though, and this is one we talk about a lot because people, as I’m sure you can imagine, request it a lot, which is they’re like, hey, I want a little toggle that says, hey, make it more professional or make it more cheeky or make it shorter or make it longer.

That’s all well and good.

Our ears are open to that request.

What we’ve found is that it’s not, it’s one of those examples where I feel as though people very often are asking for a feature that’s not actually going to solve the pain that they’re hoping it will solve.

They’re hoping for like more malleability and control, but in reality, like if it could just deliver like a top notch response that was in line with the way they wanted to see it every time, there would be no need for those like individual toggles for like specific categories of response types.

So our goal is to try to do the whole thing better instead of give people a whole lot more to interact with and then taking more time to send out like an individual response as an example.

Yeah, I love that.

I feel like sometimes that is what is missing is you want it to be closer to the end result.

And so many times you have to, you know, with, let’s just use ChatGPT, you have to continue to like prompt and prompt and prompt, and then you’ve got to edit.

And at the end, you’re like, well, that wasn’t really, you know, did that really save me that much time?

I don’t know.

So I love that.

It’s really interesting.

So in a recent press release on your pre-seed funding round, congratulations, by the way.

Thank you.

You wrote, Decky is committed to building the world’s best customer support AI to work alongside and augment the amazing humans that hustle all day to help solve customer problems.

Do you see a point in time where the technology progresses to the point that it is replacing those humans?

So one thing that I accidentally said a couple of weeks ago, but it’s come to be one of my favorite phrases, is AI is coming for your tasks.

It’s not coming for your job.

So I think that in certain categories and e-commerce workflows are an excellent example of this.

Certain workflows are already very replaceable by AI.

Like there’s a company called Sienna that does awesome, awesome AI for e-commerce applications, right?

Where it’s like somebody can go in, they can talk to the AI, they can say, I want a refund, I want a replacement.

And the AI is not only communicating with them, but it’s also pinging various APIs internally to cue up that refund, to cue up the logistics, the shipping label, to put something back in inventory, right?

Like those very like cut and dried workflows, a lot of those are already very, very replaceable via AI.

But then I think there’s a lot of, then we kind of have to break the next category down into two things.

There’s a lot of things that I don’t think AI is yet capable of replacing, and some of them I don’t think AI will ever be capable of replacing.

And one is the whole like strategy and empathy side of things, which is that if we can take like the rote and monotonous work away from people and let them focus on, what are our customers saying?

What does that mean for our business?

And like, how do I communicate this properly to the right teams such that we can all be steering in the right direction for something that like actually makes more of a higher level difference for our business and our customers?

I think that’s one of the best uses of AI right now is just like unlock human brain power for the kind of stuff we’re really good at.

And then the second category is a lot of people just want to talk to human beings.

Like ironically, more people want to talk to human beings now, given that AI chatbots have become super ubiquitous.

They just want to feel seen.

They want to feel heard.

They want to know that there’s actually somebody taking care of their problem.

And, you know, I don’t, I can’t forecast like 10, 20, 40 years down the road, but I don’t think that’s going to go away in any given time.

I think it will go away for, again, like a lot of transactional stuff.

But when it comes to certain industries and certain problems, I still think people will want to talk to people.

And that I think is a beautiful thing.

And, you know, I’m, I’m, I’m an extrovert, so I’m biased.

But I couldn’t agree more.

It’s, it’s helping us be even more cognizant of what makes us human.

And relationship is certainly that.

Well, I love we talk a lot on this podcast about knowledge management and how it’s going to change business the way that we know it.

And I definitely see Deci AI as leading the way in that.

So thank you so much for joining us.

Thank you for having me.

This was really fun.

Thanks as always for listening and watching.

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

And hey, we’d love for you to actually subscribe to this channel.

It ensures that you get our latest episodes each week.

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

So hey, perplexity, what’s happening?

This episode, we’re talking about where the VC money is going in AI.

What can you tell us?

Hey there.

It looks like AI is stealing the show in the VC world right now.

According to recent data, AI startups captured a record 28% of global VC funding in Q2 2024, totaling a whopping $18.3 billion, with big players like Elon Musk’s ex-AI raising billions on their own.

And now you’re in the known.

Thanks as always for listening or watching.

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