Finding the Next Competitive Advantage When Everyone Has AI

We’ve all heard the analogy that AI is like electricity, a revolutionary new invention that will dramatically change the way we live and work.

But what does competitive advantage look like when everyone has access to the same utilities?

How can you establish competitive motes in the age of AI?

And can we come up with a more current description for competitive advantage than motes, me lady?

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 CEO, David DeWolf, Chief Product and Technology Officer, Mohan Rao, and NordLite CEO, Pete Buer.

We also have a discussion with Philippe Wellens, CEO of Kleio AI, about how to ensure your company transforms with AI.

But first, Pete Buer joins us as always to break down the business impact of some of the latest and greatest in AI news.

Hey, Pete, how are you?

I’m good, Courtney.

How are you doing?

Doing good.

Pete, the Financial Times recently reported that smaller AI native consulting firms are starting to outpace the big four in agility and innovation.

What’s the takeaway here?

Thanks, Courtney.

According to the FT legacy consulting giants, like Deloitte et al, are finding themselves way down by scale and bureaucracy as they seek to compete and making it harder on them to take advantage of the benefits of AI.

At the same time, as you referenced, we’re seeing boutique players in the consulting space like Quantum Rise, who we’ve had on the show, of course, and Intellics Core, positioning themselves as nimble AI-first, AI-native partners for consulting.

The takeaways, towering strengths, we get towering weaknesses.

We’ve seen this in so many different ways.

Of course, the big four firms have brand power, global reach, deep client relationships, endless resourcing, but those strengths can double as a friction when agility, speed, innovation is required.

And the smaller firms that are born AI-native leverage the strength of partnerships to access the capabilities that they don’t have and as a result have significant competitive speed advantage.

They can pivot faster, they can take more risks, they can get working prototypes into clients’ hands without a 12-month steering committee process.

Second, maybe more importantly, I don’t know, but at least the lesson for business leaders on the call, it’s not just structure, but culture that differentiates competitively.

AI-native boutiques have built their culture and everything they do in their day-to-day operations around the notion of driving impact, driving experimentation, iterating, being agile and moving fast.

So final thought, I think the lesson here goes beyond whether this is about the consulting industry or even the structures of big versus small firms to the notion of mindset.

Even when you’re working within the bounds and the limitations of enterprise scale, how as a leader can you drive the methods and thinking of an agile, native AI innovator?

It does feel like a big challenge to those firms.

Frankly, doesn’t sound like a very fun job.

It just with the speed of AI and how quickly it’s changing and kind of turning that huge cruise ship feels quite daunting from this seat right here.

So good luck to all those behemoths trying to keep up with this ever evolving technology.

Pete, thanks as always.

Thank you, Courtney.

If AI is the new electricity and everyone has access to the same utility provider, where should companies look to establish their competitive advantage?

I sat down with David and Mohan recently to get their thoughts.

Today, what I would like to talk about is everybody’s getting AI, everybody is figuring out how to use it.

And there’s a point where there’s no longer a differentiator.

In many industries, competitors are using the same models, the same APIs and similar features that are being commoditized.

So here’s my question.

When AI becomes like electricity, we just all have it.

We all know how to use it.

We’re pros, and we all have equal access.

What’s the next advantage?

Courtney, I actually love your analogy, because when you say electricity, where I go is, electricity is the commodity, right?

What matters is how do you use it, right?

And you can use your electricity to power a lamp, or you can use it to power a machine, or you can use it to power your, I don’t know, Peloton, right?

And I think that is what really matters, is that organizations, it’s not enough to have it anymore, like just have the AI, being able to, but where do you use it to actually impact the way you are operating your business?

That’s the fundamental question, right?

Mohan, real quick.

Did you notice that he said he liked my analogy?

He didn’t reframe it, didn’t add anything to it?

I did.

This is episode 99 today.

It took this long.

I finally figured it out after 98 episodes.

I know, right?

I am feeling all the jams today.

I’m taking this as a win.

And man, I’m proud of myself.

I actually didn’t write it, so I can’t really take it.

But let it be noted on this podcast, it happened one time.

Okay.

So Mohan, what do you think?

If we got electricity, we’ve all got it.

Some of us are using it for pelotons and some of us are just using it for lighting the room up.

Yeah.

I really think of it as everyone has a new baseline, which essentially means everybody has the same foundational models.

The question is, the differentiators are going to be, what are you going to feed into it?

Do you have proprietary data?

Do you have a real moat around it?

Or do you have what everybody else has?

So it really always comes down to data and pipelines in AI.

Then the question is, okay, how are you going to shape the output?

And what form factors will it get into?

And it’s a boring word, but I think it’s really going to make a comeback.

It’s about workflows.

Can AI live in all of the workflows?

Essentially, when it is part of a workflow, there are compounding advantages, as opposed to it being something discreet, right?

So that’s the second point.

And the third, I think, is around the speed of learning.

Because you can run so many experiments, can an organization learn faster and faster and faster?

And I think these three things are going to define the new differentiation when everybody has the same foundational models.

I mean, ultimately, it comes down to bending it to your proprietary and unique advantages that nobody else has.

It’s always been the game.

Mohan, I think there might be two things, as you play that out.

There will be certain things that are just table stakes, right?

Every organization uses artificial intelligence in order to, I don’t know, write, copy, like, make that up, right?

And those are the things that drive operational efficiency, right?

It’s just pure efficiency.

It’s about doing it better, faster, stronger.

And there will be table stakes that we all have to live up to.

And you have to be ready to adopt those.

But that’s not where the innovation comes from.

And that’s not where, as you’re describing, true differentiation for us to be able to compete come from.

And I think that’s where you’re saying that we truly stand out.

It’s not just plugging it.

Everybody plugs in a lamp to be able to see in their office, to be able to work longer hours than just sun, daylight.

Well, where do we differentiate?

We differentiate in the areas that are tailor-made, that have a defensible moat, all of the things that you described.

Yeah, you know, in a way AI gives you the power.

But these proprietary advantages, whether it’s data or how you learn from it, gives you the advantages.

So the power is there.

The question is, how do you take the advantage?

That’s what it really comes down to.

Do you think there’s anything to, you know, for so long, it was like specialization, specialization, specialization.

And now there’s kind of been a shift back.

Like the generalists are kind of, especially in the marketing world, for example, like kind of the superpower because now you have AI agents that you can go deploy in all these different ways.

Do you foresee when it comes to differentiation that there might be a winner and a loser there with being a generalist versus very specialized in a certain skill set?

I think, I think it’s going to be, you know, the generalists are going to stage a comeback here.

Overall, I would kind of, it’s a nuanced type of generalist.

I think you need to be a generalist overall, but you need to be specialized enough to dive into the details, right?

So of each of these things, right?

So to be able to check if the outputs you’re getting are the right outputs or not, be able to troubleshoot it and it’s scale, how do you deploy it and make sure that it’s not wrong, right?

So I think it’s not going to be individual knowledge worker who’s very specialized in one or two or three things, because AI will be able to do that faster and also be able to learn better.

It’s the generalist, but having specialized skills.

I actually like the example that you used to of a marketer, because a generalist marketer is different from just a generalist generalist.

And I do think there is something to domain expertise that matters, that a human being can bring to the table and connect dots in.

And so I see it almost as if it’s a spectrum of true just business generalists to deep specialists.

It’s actually somewhere in the middle in my mind.

It’s somebody that can bring context to the table, domain expertise, experience to the table, that’s relevant to what we’re trying to accomplish, that is focused and can deploy the right agents.

But that individual does not have to have incredible depth in a single discipline within that practice area.

I mean, it simply comes down to, can you out learn faster than anybody else?

Right?

So whatever it is that you’re doing with the help of this new power that we all have, can you out learn and can you, because with marketing, it’s trial and error often.

You gotta try five different things to get on the right answer for you and the right answer is the right answer in the moment and may not work three months from now.

Right?

So it is all about learning and be able to out learn everybody else.

That’s your competition.

And what is that skill set?

Yeah, it’s probably some generalists, some specialized, some combination, probably right in the middle, sounds like the right answer to me.

It’s almost like you are a marker, Mohan.

That sounded so right on that I feel so seen right now.

Courtney is having a good day today.

I love marketing because how dynamic it is.

It’s, yeah.

We talked about being a generalist, but there’s also a certain, you said the details, but there’s also a certain amount of, at least right now, technical ability with some of these workflow, agentic setups that you’ve kind of got to be, you do have to kind of have a little bit, I’m kind of hoping that changes as we progress, but it does seem like you still have to have some technical capabilities, especially when you set up a workflow and then it’s constantly being dumped because a new version of whichever platform you’re using upgrades.

And so it’s just like very tedious in a way right now as well, which is not everybody’s skill set.

Yeah, I mean, you know, having proprietary data and being able to organize it into clean pipelines is the sustainable advantage here, right?

And it is not that you do it once and then you’re good forever, like you alluded, I mean, you know, we face this every day in our business, the APIs change, they don’t tell you when it changes, right?

Something breaks, you got to keep things going and as clean as possible.

That’s where you get the best outputs, right?

So it is the raw materials that goes into the factory.

Essentially, that is the data.

And that and the more proprietary data that you can build, whether it is based on user interactions or whatever it is, the more proprietary data that you can have is when you can build something where it cannot be imitated by anybody else.

I think a key piece of that, Mohan, is that every organization has proprietary knowledge that they have created from their business, right?

It’s just the exhaust of doing that business, they have special know-how.

Historically, that proprietary data has been thought of as very structured data, not necessarily know-how, right?

And I think there is a massive opportunity for businesses that haven’t historically thought of themselves as having proprietary data to have proprietary data by leveraging this knowledge asset that exists in their business.

And you can capture that now in very, very easy ways, right?

Record a video, just record a memo, all right?

Take all of your written documentation, right?

You can find ways to harness the power of that and organizations now can tap that intellectual knowledge that exists and is unique to their business in new, really exciting ways.

Yeah, that’s really well said.

I mean, you know, sometimes it’s in structure data, but most of the times, 80%, I would say, is in people’s heads, right?

So it’s about getting that out, like I said, whether it’s in a video form or whatever makes it the easiest to analyze.

And the game here is getting 10% better every whatever week, quarter, whatever is your scale, right?

So be able to iterate on that into a better place is what you got to do.

The best way to get data, of course, is to have a digital representation of everything that you do, which is creating these data factories and pipelines.

And it can be done in every business, right?

So whether you’re in manufacturing or whatever, there is a representation of digital data of what you do.

It just has to be instrumented as you run your business over time.

Otherwise, what you suggest is a great idea, which is just get it out of people’s heads in whatever form, and just feed that into the pipeline.

David, Mohan, I feel like this is a really interesting topic and kind of thought-provoking on where is the knowledge that gives us a competitive advantage, and how do we get at that knowledge.

So for everybody listening, giving you a little homework to think through.

David, Mohan, thank you as always.

Thanks, Courtney.

Thanks, Mohan.

Thanks, David.

The new era of commercial intelligence, well, it’s here.

If you’re interested in having real-time, objective intelligence on the health of your commercial relationships, you might be interested in trying out Knownwell.

So stop flying blind and start sprinting ahead.

Go to knownwell.com/experience to set up a time to speak with the Knownwell team.

Philippe Wellens is the CEO of Kleio AI, and the former head of International Data Science for C3 AI.

He and Pete Buer recently chatted about how you can make sure your company is prepared to transform with AI.

Philippe, welcome.

So nice to have you on the program.

Glad to be here.

Thank you for having me, Pete.

Yes.

To give our audience a little context for the conversation, can you tell us about Kleio and your role, please?

Yeah.

So at Kleio, we offer a conversational AI platform for B2B and B2C industries and companies that essentially have very complex products that are hard to sell, where you need human guidance.

And so we offer AI agents that really provide this guidance.

When we say complex, it’s typically products with many options, many variants.

So think, for example, in the manufacturing space, it could be a hydraulic pump that needs to be sold, it can be in the travel space or in real estate or in the energy.

We are active across all these industries, and that’s where we see tremendous value being created with AI.

Conversational AI, is that something new?

Like, would the majority of our listeners are in professional services, middle market companies, not all, but lion’s share.

Is that found a home in that space yet?

Absolutely.

Look, I mean, we work more with large enterprises, but it’s also something that a lot of smaller and SMBs can actually work with.

So, conversational AI, it’s essentially an AI agent that is going to conduct the conversation with a human, so with us.

And so, it’s going to do it in such a way that the quality of the conversation is close to the one of a human.

So, it makes it very, you know, it’s like you had chatbots for decades that actually never really worked, where today you have conversational AI that actually works.

And it can actually do much more than just have conversations.

It can also do some tasks, for example, for you on your behalf because it is able to have the same, like a bunch of agents essentially working together to do stuff for you.

Yeah.

Awesome.

And in professional services, you know, the traditional historical, I guess, limitation to scale is that so much of the delivery of value has been through the vehicle of the human.

And I assume conversational AI is a tool for scaling that and changing the economics of the business.

It’s massive.

You know, like the return on investment that you can get with conversational AI and more broadly with AI agents is actually very, very large.

Nice.

And the reason for that is that, you know, so far we always have been talking about or until like two, three years ago, it was about AI.

We used to call it machine learning.

And these were like small models that were actually connecting to data or to tools sometimes, but they were just connecting to that and providing insights.

Now you have AI agents that are LLMs, so foundational models, AI, connected to data and tools that can take action on your behalf.

They have some level of autonomy.

They can actually decide on what is the next course of actions that I will want to take.

And so the impact on the business can be very high because of that.

They can actually really, I wouldn’t say replace humans, but they can augment what they do by essentially freeing time from, let’s say, the low value, the time consuming task and allow them to essentially work on what they do best and where their impact is going to be the highest.

In just 20 seconds of listening to you, my mind is exploding with possible use cases, all the different ways conversational AI agents could be launched into the business.

And I think that’s a challenge for business leaders, figuring out where to start.

How do you frame that?

How do you think about that?

Well, it is indeed probably one of the hardest, the toughest questions is, and it is where to start, right?

And the reason for that is that you have so many different options in front of you that are all of a sudden available because of this new technology.

And typically, organizations and the larger they are, the more complex it is to actually know where to start.

Well, they see many options and they’re like, well, fast forward in five years, I want all of this, all my workflows to be fully automated with AI.

Okay, but where do we start today?

And what we see is that actually in the statistics, I was reading a report from MIT yesterday or two days ago, it was saying that 95% of all AI pilots that were started in the last year are failing.

I was going to bring that stat up today too.

Awesome.

You go.

And it’s really reflective of what we see, right?

In the sense that a lot of these companies have actually failed for probably two reasons.

One is they started with the wrong problem.

The second one is that they did it the wrong way, meaning with the wrong people or the wrong organization.

So if you look at where to start, there was a framework that I’ve been using for many years.

What I didn’t say in the introduction is that I also used to work for C3AI for many years together with my co-founders of Kleio.

And we used to work with the largest Fortune 5,000 companies in the world doing AI and optimizing their processes.

So we’ve been used to identifying really the sweet spot where AI is actually very useful and relevant.

And so the way I would use in a very simple manner here for the listeners is, well, where do we have, one, a business problem that is actually with a potentially high return on investment that is initially narrow enough, but once you expand it and you scale it out, it’s very large and the impact of value can be massive.

The second one is what is feasible for AI based on the technology that you see today in 2025 in the time of the recording.

And so you need to think about data.

Where do you have data?

Where do you have systems in place?

Where do you have business logic that is properly described, that you can now codify into AI agents?

And then you also need the right level of C level sponsorship.

So what type of problem will have the right executive, who will be a capital allocator, who will also be desiloing the different teams, making sure everyone collaborates together, and making sure that things get done, and the right things get done as well.

So this is only where to start, you know, on like what to start with.

And the second part is, well, how do you do it, right?

And so again, you could see that in the same report from MIT, they were saying, well, if you actually work with an AI native pure player, you have three times higher chances of success than if you build it internally.

Why is that the case?

It’s essentially because of two reasons, or that is that I see.

One is talent.

It’s very hard to find the right talent to recruit them, to retain them.

They also can be very expensive depending on the market where you are.

The second one is, well, the infrastructure you need to actually have AI agents running.

We like to say that actually 95 percent of the codes is the supporting infrastructure, and only 5 percent is the AI agent.

If you know this, it means that you actually, when you do an AI project, it’s not just about building a few AIs and putting them together.

A flashy demo can be done in a few days.

It’s actually thinking about the supporting infrastructure and doing this properly.

If you combine that where to start, like the what to start and the how to start, and you take them to consideration, I think a lot of companies could already improve a bit the odds of succeeding with AI.

You quickly painted a vision of what execution looks like in the future.

Workflows predominantly managed by AI agents.

And I guess I understand from a little bit of the background reading, overseen by an orchestrator.

How does that system work?

An orchestrator overseeing a bunch of agents, you press a button, they’re built, they run, you forget about them or can all help break loose?

It depends.

What I would say is if you are, you really have two modes, right?

You have the fully autonomous mode where the AI is going to do its own job on your behalf, and you get it to run completely on its own, and it gets back to you with the results.

The other option is that you have a human in the loop that will be checking every now and then and validating the next course of actions.

What we do, for example, at Kleio, is that we have typically for one deployment, one customer, let’s say 30 something AI agents working together with an orchestrator sitting on top of it and managing these resources.

You can really think of it as the orchestrator is the manager of the employees of the team members, and each team member has its own set of responsibilities, a job to fulfill, essentially, to complete.

And so, it’s going to be, in our example, for a sales conversation, you have one agent that is going to be about asking the right clarification questions because the need is not well defined, it’s a bit vague.

Another one is going to go and tap into the product catalog to be able to recommend the right product for the need.

Another one is going to reformulate the wording.

Another one is going to connect to systems, la la la.

And so the orchestrator makes sure that when a query comes in from the user, that query is reformulated and then dispatched to one or multiple agents to come, to work together and come back to the orchestrator with the response.

And that’s how it works.

Now, it’s already hard enough to get them to work together sometimes.

Honestly, the technology is great, but it requires really some controlling of bios.

And so the anarchy is not today already happening.

I mean, they’re not intelligent enough, I would say.

Now, that said, we still need to have the right, let’s say, pre-

and post-processing layers.

You need to have groundways in place.

The reason for that is that it might be that a user is saying something that is off topic, is not allowed, is not as part of the perimeter.

It’s maybe unethical or whatsoever.

It might also be an attack simply, like the user might be a hacker trying to hack into your system.

So you need to have the right protection for that.

And it might also be that the LLM, the AI agent is actually hallucinating, like we say, meaning it’s making up some of the information.

So we have a lot of these groundways in place to make sure that none of that happens.

And this is how we can then make sure that our customers are happy because they have these very professional conversations where their prospects, without them being in the loop, as you have these AI agents running on their own and doing all of that.

I’m imagining all these agents working in a coordinated fashion behind the scenes and decision criteria that’s moving the user from one to the next.

What’s the experience like on the user side?

Is it sort of just seamless and, oh yeah, look, I’m getting all my questions answered?

Yeah, yeah, yeah, look, I mean, we’ve been also interviewing hundreds of people following all these conversations and they were in different ranges of age and they all said, it looked very natural, like no surprise.

It’s not like as if they were, well, maybe they were expecting something more, but they came out of that saying, it felt like I was talking to someone, got what I need, but this is already to us a great achievement.

The next, let’s say, step we are taking is that we realize we need to redefine the way the humans, us, talk with the machines, the AI agents.

And so this human-machine interface needs to completely be reinvented.

And the funny thing with that is that our customers today are asking us to replace their existing website with a fully interactive, ephemeral conversational interface that is being morphed during the conversation to be hyper-personalized to the user.

That means that when you go to a website with a lot of today complex products, typically what happens is you scroll, you browse, you put your keywords, you go through static filters, then you get lost and you abandon.

And maybe you come back again and you try again.

But now imagine you have an interface that is just creating and pushing exactly the right amount of information that you need at this point in time, based on what you’ve provided as information so far.

Well, the conversation would be so much better, right?

And so it all goes about, it’s about trust, but it’s also about like, how can you really guide by the hand in a super personalized fashion, the user in a way that not even us today, using our words can actually achieve that.

Because an interface can achieve much more if you have text plus images, plus video, plus audio.

All of that combined into one interface and it’s amazing the experience is so much better.

So it goes, you know, it’s aligned with the ChatGPT and the perplexity and the Google Gemini of this world that are all doing this consumer personal AI agent.

We’re doing this for companies, for enterprises that want to also have their own ambassador, their own AI agent, talking to their customers, talking to their employees to augment them.

If I’m running this piece of the business, how is my job as a leader different?

As I think about building new value for customers, managing risk, there’s some role that trust plays in all of this.

Is my job going to change running a system, a business that’s made up of humans and systems?

Well, you want to choose the right teams definitely, and the right vendors to put this in place because you want to make sure, given the non-deterministic nature of the models, that most of the time, that is 99.999% of the time, the conversations are well conducted and perform as expected.

And so you need to have the right systems in place for that.

The second thing is that so far, we were a lot of people, all the organizations were exposing data to humans.

Now they are exposing their infrastructure, their data, their systems to AI agents, that as we were saying in the beginning, they can actually take decisions and act on the behalf of someone.

They are autonomous.

So with that in mind, you need to think a bit more carefully about, well, how do I essentially rethink my architecture and what I authorize?

And so you see a lot of new frameworks, protocols also being pushed by the market leaders in this space, the model providers as well, to essentially enable this communication between the organization, the enterprise data and systems and the AI agents.

But I think the most important to me is really think about long-term transformation.

If you are a leader, it’s important to come up with beyond all these technical security, etc.

considerations.

It’s more about, well, this is going to happen, right?

If I don’t take action as a leader in my organization, then I’m at the risk of disappearing as an organization in a few years from now because my competitors will be doing it.

How can you think about that?

How can you integrate with AI in your organization in the fastest, most effective way that is also the most impactful way?

That’s really what should be top of the agenda for many of these leaders today.

We have time for one more question, and I was going to spring the 95% fail rate stat on you, but you pulled a judo move and used it on me first.

So rather than ask your interpretation of it, which you’ve shared, thank you very much.

I wonder if there’s a question around how ROI or how success is being measured that’s worth talking about.

Like are these experiments or are these use cases deserving of a different way of thinking about ROI?

Yes, that’s a good, very important question.

It’s about return on investment and it’s about time to value.

And the great thing with AI agents is that you can now finally have the attribution to it.

In the sense that because they perform actions on your behalf, you can see the job being done in some way, right?

And so you can measure that.

It’s observable.

So the attribution can come to the AI agents.

And so with attribution comes return on investment.

So for example, to give you some examples, because of course all of the return on investment KPIs will always be very different from deployment to deployment.

In sales, when we do lead generation, well, of course, we count how many more leads have we generated.

When we do online transactions, how many more transactions and associated revenues have we generated?

When we augment sales, well, how much extra revenue per sales employee have we generated?

And we see very good numbers, right?

Like we’ve been able, for example, add just a bit to time value as well.

Within one month of deployment, we were able to triple the amount of lead generated, right?

So these type of KPIs, when you follow them with AI agents, the impact sometimes can be really massive.

The second one is time to value.

So time to value, well, how quickly can you get these results?

Of course, everyone wants them yesterday.

And so for that, I would like, oh, it’s important to realize that this is, it takes time.

So it’s super easy.

I would say really it’s a matter of a few days to have a working prototype, a demo that is very flashy and looks good.

To put this in production live, it takes typically months.

Something really powerful that is strategic.

And so you need a multi-year transformation roadmap.

It’s not something that is going to happen like that.

So time to value, ROI, these two come, go hand in hand.

And they are very important for agent AI.

Philippe, I just want to say thank you so much for spending time with us today.

It’s absolutely fascinating.

It was a great conversation.

Thank you.

Thank you, Pete.

All the best.

Thanks as always for listening and watching.

Don’t forget to give us a rating on your podcast player of choice.

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

Hey, Perplexity, today we’re talking about establishing competitive advantage when everyone has access to the same AI tools.

I really like your deep research tool.

So can you give us any recommendations?

Stand out by combining AI research with your own creative insights.

Just using the tools isn’t enough.

Perplexity’s deep research gets powerful when it fuels ideas the competition hasn’t seen coming.

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.