Working with AI: What It Means to Be AI Driven

Does AI represent an extinction level event for your company?

If so, how can you make sure your company becomes the proverbial birds instead of the pterodactyls?

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

This week, we have a discussion with Wolf Ruzicka about his book, AI Driven, and what it means to apply an AI-first mindset in your business.

But first, put on those hiking boots and high waters because it’s time for another installment of AI in the Wild.

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

Hey, Pete, how are you?

I’m good, Courtney.

How are you doing?

Doing good.

First up, according to The Verge, T-Mobile’s parent company is making an AI phone with Perplexity Assistant.

Pete, this feels very B2C to me.

To be honest, I kind of have a little bit of like rabbit twitch.

What’s the takeaway for business leaders here if they’re reading between the lines?

Yeah.

So T-Mobile’s parent company is Deutsche Telekom, and they’re behind this notion of developing an app-less device or as you said, as they refer to it, an AI phone.

So basically, think of your current device taking voice commands and working independently across all the apps on your phone without you having to dig in your pockets and go through them manually to carry out tasks like ordering that new video bird feeder that I’m dying to get on Amazon.

If successful, this model would redefine our relationships with our devices, right?

Moving from manual text-based inputs to voice-driven commanding.

Other companies headed down this path as well.

Just in the news this past week, Microsoft is working on something similar in the healthcare space.

It’s called the Dragon Co-Pilot, meant to simplify their frenetic lives of medical clinicians.

As you say, faithful listeners might in this moment be brought to mind of the good old Rabbit R1 device.

Same promising concept the Rabbit was, though a total failure on delivery.

First off, it was the second device who wants to carry that around.

And then secondly, more importantly, it wasn’t successfully doing the job of creeping across all your apps and seamlessly engaging them when you gave it a voice command.

And so this will be the thing to watch with the Deutsche Telekom device.

Can they get the user experience to live up to the promise?

And if so, it’ll be a monumental innovation.

Yep, very interesting.

And I don’t know if you’ve put in your pre-order or not as you’re waiting for your…

What were the name of those shoes?

Oh, right, the Reebok.

Yeah.

Yeah, as you’re waiting for your scintillates and your defunct rabbit, hopefully your pre-order for this can come fast as well.

Pete, thank you as always.

Thank you, Courtney.

If you are listening to this today, on the release day of this episode one, thank you.

But two, be sure to join us tomorrow for a LinkedIn Live where we’ll be showcasing the latest and greatest of the Knownwell platform.

If you are listening sometime after April 15th, visit knownwell.com/showcase to watch the replay.

If you’re someone in a professional service firm or leading a professional service firm and are really looking for how to move from a subjective view of the health of your customers to a real objective view to drive your business from, we definitely want you to check out the Knownwell platform and see what the new era of AI intelligence in professional service looks like.

We can’t wait to see you there.

Wolf Ruzicka is the SVP of Strategic Advisory at Solvd, the author of AI Driven and a board member of a number of technology companies.

He sat down with Pete Buer recently to talk about his book and why AI is the key to staying alive in the age of digital Darwinism.

Wolf, thanks so much for being here with us today.

I’d like to start off by asking about your book, AI Driven, Staying Alive in the Age of Digital Darwinism.

Supremely apt that a book about survival of the fittest would be co-authored by a fellow with the name Wolf, like the way that works out.

Help us out.

What’s the premise?

So the premise is it all happened.

So first of all, thank you for having Repeat and the Knownwell team.

I really respect everything that you and particularly David are doing.

So I’m very happy to be here.

We’re very happy.

So the story of the book is that about three years ago, a publisher called me and it was during lockdown.

And they said, you know, we read everything you’ve written about AI.

We’ve listened to your interviews that you’ve done.

And particularly the interview with the Chinese TV station, nine or ten years ago, when Google beat the reigning Go Grandmaster in South Korea, and you know, you’re opinion, you seem to be very opinionated about AI.

Would you mind writing a book about AI for enterprise CXOs?

So chief executive officers, chief marketing officers and so on.

And so I had nothing better to do and neither did Victor, the CTO at the time who wrote the book with me.

So we started writing very casually.

The sliver of the population that was aware of these superpowers was very small.

And then of course, as Microsoft made this big move with OpenAI, and OpenAI made this genius stroke of putting a chat interface on top of existing GenAI models, everyone started talking about it.

So we rushed it out over Christmas, New Year, Easter, every weekend we could.

Our families did not like us anymore, but then it hit the market.

And so the premise is that if you don’t adopt this now, this is an extremely urgent situation for any type of enterprise, company, government, nation, state, school, nonprofit, you name it, to become versed in it.

You don’t have to become an expert.

That’s what you have people like me for.

But at least understand the gravity of this emerging technology, then you will be left behind, just like the dinosaurs after the meteor strike and the mammals will emerge that are AI natives.

Now, the good news is, as we explain in this book, there is a journey to be taken, which is five steps, only five steps.

And these five steps will bring you from dinosaur level extinction threat to being the equivalent of an AI native mammal.

Well, of course, I have to ask what the five steps are.

So the first step is to get ready, prepare yourself for these new technologies.

You don’t want to just plug generative AI, which is an extremely expensive technology, if used incorrectly, into legacy type of system.

So do you have to, in chapter one, we explain how a very large company started understanding what is the core of what they do from a technical perspective and start owning it.

You don’t want to give this to a software vendor that you are then beholden to.

And essentially you’re giving your digital livelihood to somebody else.

This is the digital age.

This is the AI age.

So the core of what you do, you need to start owning it.

And this is not a flip the switch type of thing.

It is a known methodology in software engineering that we use for wrapping these systems into something you own.

So you can turn them off and you are the owner of your own digital destiny.

And then chapter two teaches with the real life example of a very large city in the United States of how to embrace AI in the simplest possible way.

The simplest possible AI sits on top of structured data.

Think of spreadsheets, tables of data, numerical data most often.

And AI is a prediction technology.

These are probabilistic technologies.

So start predicting, say, arrival times of something, or the weather of tomorrow, or any kind of situation that may come at you.

And start predicting the future based on numerical structured data.

Chapter three is then the next level up, which is 80% of the data that we have is unstructured data.

Think in PDF documents, very hard to decipher type of data structures, images, you name it.

And start thinking through how to make this unstructured data actionable.

AIs look for patterns.

And so find ways to create patterns out of unstructured data so that you can free up 80% of your data.

On average, it’s 80%.

Then chapter number four, we explain all the different best practices on how to apply AI in preparation for generative AI.

For this, we used, so for the previous chapter, we used a legal technology company.

Legal documents are famously unstructured data.

How do you deal with unstructured data in a legal environment to predict the outcome of court cases or things like that?

And then the following chapter four, we simply list all the different best practices in order to get ready for generative AI, which is chapter number five.

For this, we use a very large financial conglomerate out of Europe where we help them think through the various applications of AI in a multi-country, multi-regulation, multi-privacy data laws, protection laws environment, which means we had to use literally every best practice in the book.

And then generative AI has two superpowers and we often hear of their matching capabilities.

So we did not pick matching as a superpower of generative AI.

We use process acceleration where the results can be absolutely dramatic.

And here the example is the Department of Defense.

We didn’t work directly with the Department of Defense, but with an intermediary and help that intermediary accelerate a process that used to take the Pentagon 19 months to complete.

From day one until completion of process because of the complexities and the many people that are involved and the many moving parts.

So you don’t just throw a generative AI engine at any kind of process.

You try to go one level deeper.

A process is an accumulation and a series of tasks.

And some tasks lend themselves to generative AI, others don’t.

When you have a hammer, everything looks like a nail.

So now today we see generative AI be thrown at any kind of process.

And it costs a lot of money.

You don’t need probabilistic technologies at every step of the way.

But when you do it intelligently, the outcome can be just like what we did.

The process that previously took 19 months to complete is now being completed every single week.

As you’ve looked at these case examples and come up with your five step process for how to advance on the journey to AI readiness, what are you finding are the barriers?

What are companies, what’s getting in the way of companies making progress?

So I call it the AI chasm.

And maybe you remember back in the day, during the internet days, there was this great book, Crossing the Chasm, it was a Bible.

The same thing is happening today in AI.

So I see this perpetual recycling of AI proof of concepts.

And what I encounter most is once they reach that chasm, these perpetual POCs don’t make it into production systems at scale.

And so crossing this chasm is one of the big goals for my company to show particularly larger enterprises how to cross the chasm without opening Pandora’s Box.

And how about executive teams?

What’s the state of readiness?

Are, you know, is there a big education hurdle at the top even to get started on this work?

Absolutely.

So I spend my days almost every day in front of executive teams or board of directors where I have to explain what AI is and what AI is not.

The most important thing is to understand what AI is not.

And so I remind typically these executives, it’s a, you know, in a nutshell, the gist is AI was meant to be called many decades ago, predictions based on advanced data processing.

And that makes it very, you know, self-explanatory.

You’re dealing with probabilistic technology.

So predicting something is not always useful.

Sometimes you need certainty.

You need 100% deterministic results.

So you already start understanding it can’t be applied everywhere.

You’re dealing with data, and you’re dealing with advanced data processing, meaning you typically have to, you have no choice, but to go into one of the hyperscalers, because you don’t want to replicate these data center networks that we have in, you know, Northern Virginia.

Take advantage of these, you know, of these vast infrastructures that are here for you without, you know, getting stuck in there.

So make it, build it in a way that you can move from hyperscaler to hyperscaler, that some of the things may still be operating on your own servers that you may still have and haven’t depreciated yet.

And so these kinds of education sessions oftentimes help executives to really think through within their subject matter, within their industry, within their department on how to apply AI, how not to apply AI.

And the second, I think, most important statement in all of this is that an AI is nothing else than three things, only three.

It’s mathematics, it’s compute power and it’s data.

And so once you start combining, you know, where you own your own data, where you can acquire data, which compute power you might leverage and then mathematics, which model or set of models might make the most sense, then you very quickly get a list of use cases out of an executive meeting.

Yeah, nice.

And so I assume some of your teaching is around reminding executive teams that this actually is not a technology project, but time to solve business problems and figure out how to do it.

And you know, sometimes I’m lucky.

In one of the previous ones, they came prepared and so they had 12 use cases that we discussed.

First one, no.

Second one, can be solved much cheaper, easier, faster using traditional technologies.

And so we whittled it down from 12 to 3, where Gen.

AI can really have incredible results.

I think that’s a chapter to be added after the fact or a white paper to be written about, you know, not chasing what’s exciting or where the technology leads you, but instead go back to first principles, focus on business problems and work from there.

Correct.

Couple questions about the book, if I may.

There’s a reference to digital marines, which sounds awfully cool.

What are digital marines and why do we want them on the team?

So not every engineer is like any other engineer.

There are some engineers that have just been around the block so many times that they do live and breathe first principles, that they do live and breathe architectures.

The key in all of this is always an architecture.

There’s three types of architectures.

There’s enterprise architecture for large companies like a capital one.

Let’s just pick one.

There’s system architecture that is made up of various software architectures.

So layering correct architectures into an organization is not everyone’s forte.

Funny enough, it is the forte of our company.

We’re famous for software architectures, system architectures and enterprise architectures, chapter one of the book, that allow you to innovate, that free you from being beholden by some other vendors that you may need to replace over time.

And so it was actually coined by our customers that our team seemed to be the digital marines.

And our job is not to replace their engineer.

So famous example is one of the largest companies in the United States.

They have 6,000 engineers.

It’s not a tech company.

And we came in as the digital marines, as the CIO called us, and we created the beachhead in Normandy so that the army of 6,000 engineers doesn’t get replaced, but can storm the beaches safely.

Nice.

That’s a wonderful image.

Very clear.

Also in the book, there’s reference to this notion that not all hallucinations are bad hallucinations.

I feel like we’ve just done the job of educating leadership teams what hallucinations are and why they should be feared.

But now apparently there are some good ones.

So can you tell us a little more?

Of course.

I like to say that hallucinations are not a bug of a probabilistic technology.

Probabilistic technologies mean that there is a certain error range and it will always be there.

It’s not a bug, it’s a feature.

It comes with the nature of the beast.

And so again, one of the reasons why generative AI is not the hammer that will fix every single, should be used on every single nail.

So what do hallucinations provide you in certain tasks that you may have to complete?

So let’s assume you are a teacher and you want to create a multiple choice questionnaire for your students, individualized for every single student, so that Pete gets a different set of multiple choice question or answer possibilities than Wolf, based on my level of no education versus yours, based on your experience versus mine.

So a teacher would have to create such an answer set for every single individual student.

Now, what is the nature of a multiple choice answer set?

There’s only one right answer, deterministically, 100 percent correct.

All the other ones sound correct to a certain level of a certain degree.

And why don’t we have an AI hallucinate them on our behalf, so that we get a suggestion from the AI.

We already know the correct answer.

So, we don’t need it for that.

But what if we ask the AI to create five potential answers for Pete and five potential answers for Wolf, based on our various characteristics that make you you and make me me?

We need them to be wrong.

Therefore, a hallucination is our best friend in this scenario.

It’s just one example.

Yeah, that’s awesome.

Thank you.

I’m going to wrap up with a question about another hat that you wear, which is chairman and you have other roles too on boards, quite a number of them as I’ve just learned.

You mentioned a conversation that you had recently with a company and raised the frame in paradigm versus out of paradigm.

Can you tell us a little bit more and what you’re seeing there that’s interesting?

Yeah.

Most non-experts in AI, my mom, as one example, now thinks that an AI is a chatbot and is an intelligent chatbot.

I personally have chatbot fatigue.

If someone gives me yet another chatbot, I’m just going to jump over the ledge.

I’ll make it a very short ledge or a very low ledge so I don’t get hurt.

When you have the power of generative AI at your fingertips and all you can think of is creating yet another chat interface, then you’re missing the boat.

Chat is in paradigm at this point, and other applications of generative AI are out of paradigm.

One example I have where this is very clear to me is, I’m on the board of advisors of a company called Stand Up AI, and Stand Up AI focuses on a very interesting problem statement.

And I’ll simplify it, it is much more, much deeper than this.

But imagine the process of, for the federal government, to create an RFP or an RFI, stands for Request for Proposal or Request for Information, or a Grant Application.

And then think about the thousands and thousands of potential bidders on an RFP.

Now, in paradigm would be that the federal government now uses a GENAI to help them write the RFP.

Let’s just stick with an RFP.

And that a bidder uses a generative AI to help them write a response to this RFP.

So fast forward a few months, and we will have a GENAI compete against a GENAI and trying to trick each other into winning this RFP or losing the RFP makes no sense whatsoever.

Feel free to do it.

But I think it completely misses what a superpower of GENAI is.

And I mentioned one earlier, it is matching.

So have a GENAI, stand up AI uses a generative AI, all kinds of generative AIs under the hood, to read and understand every single RFP, RFI, grant, you name it, of the federal government.

And again, the use case is bigger than this, but I’ll just stick with that particular slice.

And then it understands every potential bidder for any potential RFI, RFP and grant in the United States or even beyond.

And it matches them.

So there is so many, why does the federal government issue an RFP?

Because they don’t want yet another one of the, we call them the beltway bandits here in Washington DC to win yet another contract.

They would like for smaller, innovative, faster moving, cutting edge type of mom and pop shops, startups, midsize companies to be aware of the perfect RFP for these particularly niche skills.

And the mom and pop shop would love to know which of the RFPs they have a much higher probability of winning.

And so stand up AI matches the appropriate mom and pop shops, but yes, also the beltway bandits for the appropriate RFPs, where they have, say, a 97% chance, probably, to win the RFP based on the following strengths that the gen AI has identified inside of the bidder and based on the weaknesses inside of the potential bidder so that the bidder can really make sure that they respond correctly and have a higher chance of winning and essentially have an automated way of watching these opportunities that are out there moving left and right, coming in, coming out every single day, probably tens of thousands of them and take advantage of only the ones that make the most sense to the benefit of the federal government and therefore to the benefit of the citizenship.

So out of paradigm, an unusual application of one of the superpowers of gen AI, I think a wonderful use case because it has essentially the triple bottom line.

It has a positive impact on the country.

It has a positive impact on the government and it helps companies that would normally not be competing in certain areas for, you know, so.

Wonderful.

I’m gonna resist the urge to make a comment about improvements to government efficiency that could come out of a process like that.

And instead, thank you for the terrific example to understand the difference between in paradigm and out of paradigm.

And I think that’s incredibly important to listeners to actually get an appreciation for the fact that it’s been some time now.

And the examples that we’ve been so fascinated by in the last 12 months are kind of getting to be yesterday’s news.

And we need to be leaning farther forward in the innovation cycle.

Absolutely.

Well, thank you.

It’s been a wonderful, wonderful conversation.

I feel like we only just scratched the surface.

I’m hoping we could get together again.

Of course.

Anytime.

And thank you to Knownwell and to Pete and Dave for having me.

David.

David DeWolf.

Awesome.

Thank you.

Thank you, Wolf.

Pleasure.

Thanks as always for listening and watching.

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

And we’d really appreciate it if you can leave a review or share this podcast with someone you know would enjoy it.

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

So, hey, Gemini, this episode, we’re talking about how organizations can become more AI driven.

So, what do you think?

To become more AI driven, businesses should invest in training across all departments, so people understand AI’s potential and how to work with it.

They also need to build a data-centric culture, where decisions are informed by insights generated through AI analysis, making it a fundamental part of their operations.

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.