The Path to AGI and The Business Altitudes of AI

Just what are the five levels of artificial general intelligence that OpenAI recently laid out?

And how do they correspond to the five altitudes of AI in business we’ve talked about on the show many times?

And hey, Sam Altman, thanks for listening to the show.

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 Blake Morgan about the overlap of AI and customer experience.

But first, as AI Knowhow approaches its one year anniversary, let’s take a stroll down memory lane with Pete Buer.

Pete, I want to do a little foreshadowing here.

Next week is our 50th episode.

It’s hard to believe.

That means that we’ve been recording AI Knowhow for a year now.

I wanted to ask you, as you look back over the last year, are there any guest interviews that really resonated to you?

Which are the ones that stood out?

because I think outside of a couple, you interviewed every single guest, and there have been some tremendous ones.

So who sticks out?

Do I have to name names?

Can I talk about the experience instead?

Yes, to both.

I’ll tell you, I think it has been one of my greatest sources of learning across the entire year, in a role where I need to try to be as smart about the stuff as I possibly can.

And I read the books and I watch the YouTube videos, but having the ability to talk to experts in various dimensions of the profession, real time, and see where the conversation takes us, has hands down been my best learning input for the entire run.

And it’s been on a continuum, right?

There’s the technical set of questions, there’s the people set of questions, there’s the philosophy set of questions.

I feel somehow well-rounded from having had the experience of, I guess, 50 conversations over time.

It’s been awesome.

Okay.

I am going to make you name names though, because I am curious, like, is there anything that stands out?

We’ll just go ahead and say, blanket across the board, such a wonderful guest.

So that way you can just name names now.

I can play along if you make me.

I’d say Christian Motzberg was probably my most impactful, because when we met Christian and we had the chance to talk to him about the distinctive role of humans in the human machine mix, and what it means to be human and what’s at stake around AI.

Number one, it was early in our conversations.

A lot of the conversations weren’t about the human factor.

And I was personally concerned that companies weren’t spending enough time thinking about that angle.

And he was so dead center committed to the question, and so gifted in articulating his views.

And so very clearly passionate about it.

I was really struck and he reinforced for me my commitment to keeping track of the people part.

And I’m grateful to him for that conversation.

I feel lucky that we were able to have it.

Such a great episode.

I couldn’t agree more.

Anybody else or any moments or themes that kind of changed over the last year?

Well, I’ll tell you, a theme that’s been fun to watch has been the hype versus reality spectrum of views on AI.

In like early days, I think we ourselves were a little, I don’t want to say we were hypey, but we weren’t as grounded as we could have been, in the reality of what was to come.

And the conversations kind of followed a pattern, like the Gartner hype cycle at some level, where the realities of company investing and the hard work of transitioning structures and people.

And we’re learning now the skills gap around identifying opportunities to transform.

At the same time as progress feels like it has, I don’t know if it’s slowed, but it just doesn’t have the jet fuel momentum that it once had.

I actually feel like the horizon is getting farther out, like the possibilities.

The more we learn, the more we realize how much harder it is, the more we also realize how great the possibilities ultimately are.

But the bottom line is it’s been a trip watching the evolution of perspectives on the potential for AI and the threats of AI, as we humans have sorted through it across time.

Well, Pete, thank you as always, and here’s to the next 50 interviews.

Thank you, Courtney.

I love it.

That was good.

Best one yet.

How do OpenAI’s five steps to AGI relate to our five altitudes of AI in business?

I talked to David and Mohan recently to dive into this question and the potential impact on businesses like yours.

We’ve talked a few times on the show about the five different altitudes of AI in business, really the five altitudes of business and how AI applies to them.

OpenAI recently announced that they have a five-level evaluation process to determine if or more likely really when we’ll reach artificial general intelligence.

On the show today, David and Mohan, I wanted to see if we could talk through those five different altitudes of business, again, layered with this new five levels coming from OpenAI and see how we expect the application to the different levels in business to change as we go up the levels that OpenAI just recently elaborated on.

So you’re suggesting that OpenAI has taken our framework and they’ve developed their theory of AGI based on how we’re positioning work in business, is that right?

They saw our model and we’re like, oh my gosh, let’s make our model.

If I were them, I would have done that as well.

It’s funny because this framework actually has been core to our evolution as a business and how we’ve been thinking.

You’re right.

We talked about it early on.

We haven’t talked about it as much lately.

But I think that’s because it’s really embedded in the way we think about AI and how it’s going to just roll out through business.

And the framework is really simple.

For listeners that haven’t heard it before, there are five altitudes of work in business.

At the very lowest level, we have manual labor that takes place.

Manual labor, we’ve been talking about for years in terms of robotics, in terms of automating that and leveraging AI to do it.

Then, right above that, you have the base level of knowledge work, which is the execution of discreet tasks of knowledge work.

This is finding discreet use cases that are typically oriented around personal productivity of knowledge work.

Things like writing code, things like drafts of blogs, is where the AI really starts to play out and show that it can assist us with doing the knowledge work that is kind of the rudimentary knowledge work in business.

Use cases like helping with accounting and automating aspects of that, that type of thing.

The third level is all about how you take those unique work streams in business and weave them together to produce a business result.

A blog by itself or code by itself or accounting by itself do not produce business results.

But when you weave them together and you orchestrate a business, then you begin to drive business results.

Okay.

So if we have manual labor at the base, we then have execution work.

Second tier.

Third tier is the operational work of the business.

The fourth tier is the strategic work of the business.

This is all about making trade off decisions in designing your operations.

Those trade off decisions are the next level of knowledge work that we do.

That strategic level is that fourth altitude.

The fifth altitude is all about the ideology of the business.

These are things like defining your purpose.

Why do you exist?

Your values, how do you create a culture of human beings that believe the same things and act in a certain way?

Those are the five altitudes.

Manual labor, execution of knowledge work, operations or the orchestration of pulling those discrete execution points together, the strategic trade-off decisions, and then finally, the ideological work of defining the culture of your organization and the basis of your organization.

We believe that those first two, manual labor and the execution of knowledge work, is exactly where we are with AI right now.

But the frontier is the operational and the strategic, and that is where the highest impact is going to come from, and that is where the most innovative firms are working right now.

Yeah, and I think that’s where some people that maybe aren’t as deep in AI are confused.

They’re like, this is it, ChatGPT, and we’re like, no, no, no, no, no.

So, Mohan, would you break down what OpenAI released with their five levels of where they are towards artificial general intelligence?

Yeah, you know, they have a framework that’s similar to ours, but I think it’s a different paradigm.

The way they build this up is, you know, where we are right now, which is at step number one, clearly, which is around conversational AI, your ChatGPT, chatbots, so on and so forth.

You can make a case that we are now at the precipice of step two, which is more reasoning AI.

You can see that AI agents can do work comparable to some of the high-level tasks that humans do, right?

So it’s able to reason.

So that’s step number two.

Three steps after that, above that, are autonomous AI.

So this is AI agents being able to operate autonomously, autonomously of humans, independently of humans.

That’s step number three.

Step number four is a self-learning component to it, which is innovating AI, meaning it’s constantly innovating and doing things better.

You can see that as something that is inherent in AI with data and machine learning, how it can get better over time.

The fifth and the last step is an organizational AI, where AGI can provide the work of entire teams, of entire organizations.

The five steps, the way they define it from bottom to top are conversational AI, reasoning AI, autonomous AI.

Number four, innovating AI and then organizational AI.

Thank you.

I think for this conversation, what’s really interesting to me is to take the OpenAI model of where we are towards AGI, and go back to our model and look at what’s the next step.

You’re saying, hey, we’re probably on the beginning of reasoning AI, and then after that, autonomous AI, and go back and look at our framework and say, okay, how is this going to change when we get there?

What’s going to change at the operations level?

At the strategy level, where did these things kind of sit parallel?

And what will it look like in business as we go up OpenAI’s model?

Courtney, I think that the second levels is the first intersection that you should take a look at.

So reasoning AI.

OpenAI defines that as AI being able to perform tasks that are comparable to a human at the doctorate level.

When you say the doctorate level, you’re typically talking about an area of knowledge or expertise is what is implied by that.

And I love the fact that they use the word task because whenever we’re talking about execution based work, we’re talking about discrete use cases, completing a task, doing something, but doing it in a vacuum.

Right?

And I think in a lot of ways, we’re talking about the same thing there.

And so it doesn’t surprise me to hear OpenAI say, they think we’re on the precipice of this.

I think in many ways, in many use cases, we’ve already arrived.

I think when they add in at the doctoral level, they’re adding in certain levels of things that probably in an organization, you don’t have the need for a lot of academic type research and kind of the extreme theory that that implies.

But I think at the pragmatic level, the work that’s actually being done absolutely kind of describes an overlap of these two realities and I think is where we exist today.

So if you start there as the baseline, saying there’s a certain amount of knowledge work that’s being done, that is use case and task-oriented, absolutely we are right there where the AI can reason through and complete that knowledge work.

OpenAI then defines that next level, which they call Autonomous AI and I think it’s interesting.

They talk about agents being able to operate independently for several days without guidance.

I don’t know about you, but I know a lot of humans that struggle with this.

Kids, for example, and all sorts of employees.

The question that comes to my mind here is, to what ends?

Yes, can an agent or a human being for multiple days operate autonomously?

Yes, without direction, but to what ends are they operating is the question.

This to me is where the idea of operations comes into place.

Is operations is all about building those systems, those processes, those guidelines, those principles that provide the guardrails to ensure we’re not just executing autonomously, but we’re actually executing autonomously towards an end that will produce the desired result.

And I think that’s a really critical aspect that it’s important that we dive into and build as we are seeking this quote, autonomous AI, what we call that operational altitude of work that we’re trying to automate.

You know, you can see a much faster path along the lines when you scope it down to a domain, right?

So if you think about health care and you think about radiology or you think about education and personalized learning, you can see a path to going up these steps fairly rapidly, not all the way to the end, but fairly rapidly.

But the fundamental definition of AGI is that it should be able to do this in multiple domains, multiple problems, understanding the context, comparable to a high-functioning human that’s going to be able to do it.

So this is a really complex problem.

So that’s why I think it’s fair to say we are at the doorstep of step number two here, which is reasoning AI.

But when you scope it down to a domain, you can see in the next few years, getting to step number three and step number four.

That’s really interesting to scope it down.

So you’re saying if you look at these just as a specific use case, these steps may happen more rapidly, but when you look at them more globally, it may be much, much longer.

That’s right, because the way AGI, the textbook definition of it is be able to understand, learn, apply knowledge across broad range of domains and broad range of problems compared to human activity.

That is a very broad range.

The way AGI is defined as, hey, you’re a digital neighbor who lives around and kind of does everything else that you do or works with you.

So just multiple domains is a little far-fetched in my view.

But when you scope it down, you can see a much narrower but faster path.

And I think not only far-fetched, Mohan, but it is beyond what any human being is, right?

It is the rare individual that can actually be a Ph.D.

level academic, while at the same time as driving pragmatic business decisions, while at the same time as building a culture, while at the same time of doing that across seven different industries, right?

You know, like we’re just creating a true unicorn.

Yeah, this is what I’ve been thinking the whole time is like, I would like to see this human because it doesn’t feel like this human right here.

And I think the exciting thing for me here is this is why when we talk about applied AI, not just the theoretical technological AI, but applied AI, I think you’re in the midst of seeing the launch of several vertical AI businesses, right?

because I think that scoping down that domain and solving the domain problem, both vertically and horizontally and a mixture of both, is going to be essential for really tapping into the benefits that we can get from this technology.

And it’s not about a single LLM in the sky that is the master that can do all things for all people.

Yeah, you can definitely see narrow but deep.

You can see broad but shallow.

But the promise of what people normally imagine is broad and deep, right?

So that’s going to take a while.

But the reason why people get excited about it is the immense potential of something like this, way down the road.

And obviously, there are tons of unlocks needed.

There are risks to be managed.

So it’s a long time away.

But the promise of this immense potential is what excites people.

Now, Mohan, I know you said, you know, it’s a ways away.

But I keep feeling like people are saying things are a ways away and then like all of a sudden it’s happening.

You know, like there’s a press release that, oh my gosh, we made it.

We said it would be five years and it was actually four months.

That’s an exaggeration, but you get the gist here.

You legitimately think we are quite a ways away from really scaling OpenAI’s model that they’ve laid out here.

Yeah, I think so.

You know, when we’re talking about AGI here as opposed to LLMs and those types of AI models, right?

So when you talk about AGI, even sort of the narrow lane that we painted a picture of, it’s several years away.

It’s five to 10 years away in terms of going through those five steps that we defined.

And then when it comes to much larger applications, it’s further more than that.

So we’re talking a generation away at a minimum.

Yeah.

But Mohan, I will say, if you go back to the original frame we have around applying AI to business, I think you are de-scoping there.

You’re moving it to a business domain that then, if we start to look industry by industry, you can absolutely see the orchestration of business and that operational layer come to be within a year or two.

And those types of things within certain domains.

And I think that’s the value of that framework.

It’s thinking about the types of work we do within this domain and scoping it down to a place where we can actually begin to see every single one of those altitudes become automated and focused on the different areas that will drive the biggest impact to our businesses.

Yeah, absolutely.

I think what you need from a business perspective is really around, you really want humans in the loop, right?

So for all the high value activities and decisions that are taken in a business, right?

So I don’t think any one of us would advocate for a human not to be in the loop for high value.

But for anything else, it could be autonomous, right?

So we have a much more different way of thinking about it.

We think of these systems as co-pilots initially, then a thought partner, and then you get into intelligent agent architecture.

So those are steps that we are actively taking right now.

But in the future, it could be an intelligent operating system as well.

So for us, we can see all of this happening in the next couple of years, all of the steps that are described, and I can see a path for us to get there.

David, Mohan, it’s really great stuff.

Thank you as always.

Yeah.

Feel free to let us know anytime OpenAI wants to build on our framework.

I appreciate that.

Awesome.

Thanks, David, Courtney.

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.

Blake Morgan is a customer-experienced futurist who’s been called the Queen of CX by Metta.

She recently chatted with Pete Buer about how she sees AI transforming customer interactions and what businesses need to know and do to adapt.

Blake, so great to have you on the podcast.

Welcome.

So good to be here.

Thank you for having me.

So I’d like to start with the concept of a customer-experienced futurist here at Knownwell.

We consider ourselves at least to be in the flock, if not exactly birds of a feather.

How do you spend your time in that capacity?

Well, I’ve been studying customer experience, which is a topic I find endlessly fascinating.

I’ve been studying it for over 20 years where we didn’t call it customer experience, we called it CRM.

Since then, I’ve held roles at Intel and different kinds of companies.

But the thing that I’m most passionate about is inspiring people to really show up and be the best version of themselves.

Customer experience is the only way to differentiate your brand.

When you’re a customer-experienced futurist, you’re looking at how is society changing?

How are technologies impacting customer behavior?

How are you getting out ahead of customer expectations?

I hope that all people will begin to think of themselves as customer-experienced futurists because this is really the only way to compete is to be thinking, what does my customer need?

How can I provide it for them before they even know they need it?

The futurist is always taking care of today and thinking about tomorrow.

I have a new book out called The Eight Laws of Customer-Focused Leadership.

In the research, what I learned is that the biggest hindering block to creating incredible customer experiences is actually the ability for leaders to hold two opposing ideas in their mind at the same time.

How can I take care of everything that’s on my plate today?

How can I put out all the fires that we have in customer service today?

Well, also building for tomorrow, creating a vision for tomorrow and focusing on innovation and growth.

Most companies just can’t do this.

They’re only able to focus on today.

The futurist angle is, okay, we’re going to take care of everything today.

We’re going to go above and beyond for our customers.

The futurist is, but where’s our vision?

What direction are we headed in one year, two years, five years, 10 years?

May I ask about the book a little?

The subtitle, New Rules for Building a Business Around Today’s Customer.

What’s changing with today’s customer and why do we need new rules to serve them?

Well, generations are changing.

Now, we have Gen Z, the TikTok generation that every generation has different ways they want to interact with brands, what they expect from you.

We have to continue to keep it fresh because even I’m a millennial and I don’t even like TikTok.

I find it overwhelming but that doesn’t mean a 20-year-old is going to feel the same way I do.

We have to continue to stay on top of trends, what’s happening, what do our customers expect, how do they want to interact with us, how do they want to shop, what is their favorite channel for customer service?

That changes all the time.

In the book, I do have some generational research to say, hey, Gen Z expects something way different than baby boomers, here’s what it is.

But no matter what channel you show up on for your customers, you have to ensure that you do a great job.

It doesn’t matter what age the customer is, you have to go above and beyond in every channel.

For example, I remember about 10 years ago, I worked at Intel, the chip maker, and I was hired to make the company more social as far as customer service is concerned.

That was really tough.

It took a year to just get us on Twitter, to responding to customers on Twitter, to create Intel support, which I launched.

It’s funny, now social media, it’s like, well, who cares?

What social media?

That’s so boring.

But if you remember when that came out, it was such a big deal.

Now it’s old hat, it’s boring.

Not that it’s boring, but now we’re on to the next thing.

You can never stop being hungry and interested and curious for what’s happening with your customers.

Our focus is on AI in customer management, customer experience.

In your book, does AI play a role?

Absolutely.

Obviously, in 2023, ChatGPT really just blew the lid off of every company.

Well, now what do we do?

Now, how is this going to change the contact center?

I actually had the opportunity to recently record a course with LinkedIn on change management for AI in the contact center, because change management is going to be a really big piece of how we bring AI into our companies.

I think that understanding that these technologies have big implications for how we work, what our employees are going through, and I’ll give you an example.

A contact center agent in the past would do their summary of their call for each interaction.

And now with AI, the AI can summarize each interaction.

So that agent in the past had maybe two minutes to fill out these notes about the customer interaction.

Well, in the past, it didn’t take the agent two minutes.

Maybe it took 30 seconds to type a few notes.

So they’re getting a 90-second break in between calls.

So now with AI, the agent might say, I don’t like this AI, I don’t like this technology.

Well, they don’t like it because it took away their fake break that they had.

So it’s also about understanding what works well, what doesn’t work well, what do agents need in the agent experience.

So understanding what’s happening with the agent experience is really, really important.

As you describe customer experience as the only remaining mechanism of differentiation, and how every interaction that the humans, I guess, and the machines, and the hybrids all need to show up and do their very best together.

It makes me think about the culture of an organization that ties all the moving parts together and teaches people how to operate in the absence of rules or whatever.

Is culture becoming more important, less important?

Is it a big driver of getting customer experience management right?

Yeah.

If you think about it, New York City just implemented a new rule about trash cans.

They’re now providing trash bins.

If you’ve ever lived in New York City and I did for five years, in the summers that heat, the smell of the trash bags out on the curbs, I mean, it’s unignorable, it’s unavoidable, it’s potent, not in a good way.

They just said now, I think the mayor of New York City said, okay, we’re going to have trash bins.

Culture is almost like how many times does an employee walk by an obviously foul-smelling, you could say trash bag, but it could be anything, some imperfection, some problem and ignore it.

A customer-centric culture is when employees walk by that smelly trash and go, someone should do something, this is terrible, I don’t want our customers to see this.

Do you have a culture of that extreme ownership?

Do people feel accountability over their work?

Do they care about every little detail?

And how many times do employees walk by an obvious imperfection or problem and ignore it?

To create that culture, that’s why I wrote this book on customer-centric leadership, because I believe only the leaders can really set that tone and create that culture or vibe where employees feel proud to work there and they rise to the occasion.

They know their purpose.

When they get up in the morning, they feel directly connected to the work they do.

They feel, they care, they care about doing a good job for their colleagues and their customers.

And back to AI, is AI an enabler, a threat to creating a successful culture as you’re describing?

So if you have leaders making these decisions around the technology that are not thoughtful, that don’t care about customer experience, this is a problem.

And I’ll give you an example, which is Air Canada.

When they released this chatbot and a customer wanted to attend a funeral, someone in their family died, they wanted to attend a funeral, but the cost of the ticket was going to be very expensive, like $7,800.

So this customer asked Air Canada’s chatbot about the bereavement policy, and the chatbot said, yes, you can fly and then get reimbursed later with the bereavement policy.

So when the customer attended the funeral, they came back, they called the contact center and reached an agent, a person.

The agent working for the airline said, oh, I’m so sorry, that chatbot, that’s incorrect.

You have to do the reimbursement before you fly.

Well, they also, airline said, that chatbot, that’s not us, that’s our chatbot.

That’s a third-party technology.

We’re not responsible.

So the customer took the airline to court and ultimately the customer won.

But that is so embarrassing for Air Canada that, hey, this is your technology on your website.

You should be accountable.

If you released a chatbot that hallucinated, that’s your problem that you should absolutely reimburse the customer and give them some credits.

So this is just not the kind of culture and mindset that you want.

Technology should not be released until it is thoroughly QA’d and absolutely perfect.

In the same way that culture guides how employees show up, it should also guide the standards around which technology shows up to serve the customer.

Right.

The technology is a reflection of the culture and the people that work there.

If your technology is hallucinating and isn’t professional or presentable, cut that thing, get it out.

It’s dangerous.

You clearly know this space incredibly well, and we’re always interested to understand what’s happening at the edge of practice.

Do you have any favorite AI-powered tools or systems you’ve got your eye on or you’ve been playing with?

Yeah, I think Copilot Agent Assist is really interesting because agents are often very overwhelmed.

And Copilot or Agent Assist is almost like Waze, the driving app for agents, where now they’re getting guided by a technology that just makes their life easier.

And so now they can focus more on the emotional side of being an advisor to the customer instead of just, oh gosh, I have to sort through 40 different desktops and all this text where the technology is doing the heavy lifting and they can focus more on the relationship with the customer.

Well, I’m watching the clock, so I will wind us down.

But I’d like to say thank you very much for being here and sharing your thoughts.

And best of luck with the new book.

Thank you so much.

I’m sure it will fly off the shelves, yeah.

Thank you very much.

Thanks as always for listening and watching.

Don’t forget to leave us a rating and a review.

It really helps more people find the show.

At the end of every episode, we like to ask one of the large language models what they think about the topic at hand.

Hey, ChatGBT, what’s happening?

This episode, we’re talking about the five steps of AGI from OpenAI, and how they relate to or don’t our five altitudes of AI in business.

So what do you think?

I think layering OpenAI’s AGI levels onto the business altitudes you all developed is a cool way to see where AI might shift from simple tasks to big strategic decisions.

It’s exciting to explore how these frameworks align and where they might take us in business.

Now, you’re in the know.

Thanks as always for listening or watching.

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