Gen AI and the Evolving Workplace

Tom Soderstrom (06:18):
Welcome to “Conversations with Leaders”. My name is Tom Soderstrom and I'm an AWS Enterprise strategist and I'm delighted to be joined by my peers, Matthias Patzak and Arvind Mathur. And we're going to talk about generative AI. But we're going to talk about the workforce parts. No technology has ever been successful without the workforce adopting it. So how do we make that happen? We talked to... How many customers did we talk to last year? Matthias?

Matthias Patzak (06:45):
1600.

Tom Soderstrom (06:45):
1600 executive teams. So we want to share those lessons learned. Do you want to introduce yourself?

Arvind Mathur (06:52):
Yeah, sure. Hey everyone. I'm Arvind. I've joined the enterprise strategy team, but recently this year I've been a CIO and digital leader at companies like Kellogg's, Prudential Insurance and Procter and Gamble. I’ve driven change and transformation in diverse companies, but the same kind of challenges I've seen and now hoping to help our customers with that.

Tom Soderstrom (07:13):
Fantastic! Matthias.

Matthias Patzak (07:15):
So I'm Matthias, based out of Munich, Germany, four years now with AWS and I was CTO of AutoScout24. It's European largest internet site where people can buy and sell cars.

Tom Soderstrom (07:25):
Great. And I did a transformation for Jet Propulsion Laboratory in NASA into cloud.

So customers are asking us a lot today. Tell me something that happens, something real, a real use case with generative AI. How about Matthias? Can you give us an example?

Matthias Patzak (07:43):
Yeah. When people think about generative AI use cases, they think about a chat window and they ask a question, they get an answer or they get a picture with a person with four, five or six fingers. But actually what excites me are generative AI use cases at scale at enterprises. And so I just said, I'm out of Munich, Germany. And what really inspires me is BMW, their headquarters in Munich and they used to be a car manufacturer, but in the last decade they also became a tech company.

And so they have built a large tech organization for onboard IT on the car, but also offboard. And in the offboard domain, they have more than 450—not software developers—but 450 DevOps teams. And this architecture in this organization is creating around 14 billion requests a day and they're creating 145 terabyte of data a day.

Matthias Patzak (08:43):
So large organization, large architecture, complex system. And what they do and what they did, they used Amazon Bedrock in the AWS console to help their engineers to optimize, manage and observe performance, availability of the system. So an engineer can ask a question, "What's this going on in the system?" They can see the raw data, they can dive deeper and the system is going to give advice on what's going on. And this is efficiency increases and productivity increases at scale, and not “just create a picture with a person on a beach”.

Tom Soderstrom (09:18):
That's a great use case. Arvind?

Arvind Mathur (09:20):
Yeah. I'm going to show an example from one of my previous experiences, another traditional organization. It wasn't gen AI back then, but was how what you would call traditional AI. And now the potential with gen AI is even bigger. And this is a situation where a traditional insurance company, starting to feel the heat from insurtechs and all these digital natives coming into the industry, were sitting around and saying, “What would the customer experience be of a digital native coming into this business disrupting us?” And the key really was to make the experience the journey for them, whether they're buying insurance or claiming to be dramatically more frictionless than it is today. So when we were benchmarked, we were doing anywhere from two to six weeks to underwrite a new policy or to fulfill a claim. We said, “What would it take to do this in minutes?”

And the advantage with a company like insurance is there's a lot of data historical. And so we leveraged that, built models that could understand risk in real time and be able to underwrite that and also to judge whether a claim being made is appropriate or not. And indeed, for a majority of our customers, we were able to cut these times down from weeks and months to minutes.

Tom Soderstrom (10:38):
Amazing.

Arvind Mathur (10:39):
And that was just a tremendous experience improvement. And now as I talk with some of our customers who are exploring things like this in insurance, with gen AI you can do this dramatically more simply and more powerfully than it was much harder to do this six, seven years ago. So much easier with gen AI now.

Tom Soderstrom (10:58):
Great use cases. I would say one of the things that excite me the most being an ex-programmer is that we've seen over time, due to cloud, the cost of networking has come down, cost of storage, cost to compute, cost of software has gone up. So the capability of having a coding buddy, and I'm talking now about Amazon developed Amazon Q Developer is huge.

We did a test where there was actually a scientific test, which I liked from a science background, that compared people who did use it and who didn't use it. And the ones who used this coding buddy was 57% faster and 27% more likely to go into production. That's the big deal. It's amazing. At Jet Propulsion Laboratory, we had the most senior developers, the best ones be teaching the younger ones. Now you could have the code developer, Amazon Q Developer be that coach and insert the security policies and all that so you can be compliant much, much faster. So really good use cases. So when we talk about efficiency, we talk about scale, we talk about all this. Is AI going to take their jobs? What do you think?

Arvind Mathur (12:16):
Yeah, I think I do believe AI and gen AI in particular is a very big deal. It's extremely transformative. Anything which is transformative is also disruptive. I do think this will change work in a very significant way. It will displace some people. But I'm also an optimist on this. I do believe it's such a powerful capability that it will spawn new businesses. It'll create new kind of jobs that we can't even imagine today. And overall, I believe that it will drive business outcomes and overall GDP growth in a way that it will be a positive force for humanity. So I'm very optimistic about it.

Tom Soderstrom (12:56):
Okay.

Arvind Mathur (12:56):
It will displace individuals and I think that's where the opportunity is for us to lean into this and learn ourselves and help our people learn this capability so they can constantly be reinventing, because in the future, reinvention is the key.

Tom Soderstrom (13:10):
Good. So you're optimistic.

Arvind Mathur (13:12):
I'm optimistic.

Tom Soderstrom (13:12):
Matthias.

Matthias Patzak (13:13):
I'm optimistic as well. But yes. So we saw it with typewriters, personal computers, web technology, mobile, and now we see it with generative AI. Some of the jobs will be displaced over time, especially on the long term. So jobs are going to change, but a lot of new jobs are going to emerge. But the main thing, what's going to happen is, jobs are going to change.

So I used to be a software developer. Right now I use a lot of generative AI myself day by day. And so all of the knowledge workers who do creativity work will use generative AI. And leaders need to support their staff in upskilling, and reskilling, and gaining these skills and staying relevant on the job market. And what we will see on a global scale in more countries earlier than in others, a skill gap. And I'm out of Germany and we already see skill gap emerge, and generative AI is going to help those countries to close this skill gap, especially for knowledge workers.

Tom Soderstrom (14:22):
You also mentioned something we talked about earlier about elderly workers that I thought was fascinating. Would you mind sharing that story?

Matthias Patzak (14:28):
Yeah. Recently, I attended an event, Hamburger IT-Strategietage. It is the German largest conference for IT leaders. And I attended a round table with people from the German manufacturing industry. So old industries, traditional organizations, and they have older workforce and usually when they introduce new technologies, they get resistance from the older people in the workforce. With generative AI, and this was one of the anecdotes that one of the leaders shared and others agreed, that the older people in the workforce are welcoming and appreciating generative AI because they feel this technology is the first technology that gives them an advantage and helps them to keep up with the young digital talent in translating, in having access to information, doing analytics. And so this is the first technology where the older workforce feel they can get an advantage.

Arvind Mathur (15:28):
It's a level of reform.

Tom Soderstrom (15:29):
It's great.

Matthias Patzak (15:30):
This is awesome.

Tom Soderstrom (15:30):
Yeah. And you mentioned you're from Germany, Singapore, Sweden, and the US and I recently visited Sweden. And a lot of the older people are being left behind. I think because they don't have the technology skills, so they speak into the phone, now they can get answers. So I'm very, very happy about that.

So I think is it going to take jobs? I don't think so. Amazon added 750,000 robots and 1 million jobs. So it's going to displace some. And I think you can... Actually, let me ask you, how should leaders, as leaders, how do we address this? And I'll give my opinion. What do you think? How do we address this skills gap and making sure it doesn't take our jobs?

Matthias Patzak (16:30):
So, actually we have seen this before. We have seen new technology being introduced in organization and it's still going on. So it was when internet technology came. Then lately cloud technology and still a lot of organizations are still on the way to digitalizing their internal processes, their business models. And the same is true with cloud technology. So people and organizations are on change initiatives now all the time. And it's accelerating, and all the strategies and tactics we had for introducing digital technologies cloud technologies, it's the same when introducing generative AI.

And basically, you need to double down on change management. You need to double down on communication. Why introduce certain technologies? You have to provide psychological safety. You provide career development, career opportunities re- and upskilling. And then you need to provide different ways of learning because every individual learns differently, different times and different ways. And organization needs to provide these learning opportunities for their staff.

Arvind Mathur (17:45):
Yeah, I want to share, agree with you, Matthias, on this one. I want to share a story also from that same experience. We applied a lot of AI into all kinds of business processes and one of those was customer support, which is a classic use case for AI and gen AI now.

But this piece I want to share is when we started doing this work to improve customer service experiences, it was some of our agents who used to sit at the customer service desk who had the experience and deep understanding of what conversations do our customers want to have and how to close those, what information to seek and how to close those in the best possible manner, were the ones that leaned into this and became what we ended up calling bot trainers. And it wasn't just a one-time job, we trained them to, on a weekly basis, look at what conversations are happening through this AI bot and understand where the bot's not performing well enough, how to retrain it, what data to extract from historical data to improve the training so that more and more of the conversations could be closed satisfactorily.

And this became a great story. It was in the news for us, and it became an inspiration for everyone else to say, "I want to lean in and learn this new thing because this puts me on a new S-curve." So I think as leaders to your question, what we need to do is help our folks learn this new capability, but then turn some of those into success stories and talk about it. That encourages everyone else to also do the same.

Tom Soderstrom (19:20):
I love it. I think one thing hasn't changed, leaders need to inspire, and we have more to inspire about now. I agree with everything you said. And I think when we think about how do we make sure that our employees don't get replaced by AI skills, there's reskilling. It's not necessarily hiring new people. It's reskilling the people who are there that already know how the business works.

And how do you do that? Let me start with how not to do it in my opinion, which is send everybody off for training for six months and come back and don't do anything with it. Instead on-the-job-training, being able to learn. And I love the fact that you have the people who have the problem train the bots in what to do. We did that with great success at Jet Propulsion Laboratory, used Alexa, Alexa for work and was able to automate some of the more boring jobs.

Tom Soderstrom (20:13):
So we talked about what are some good use cases. AI is not going to take our jobs, and here's how we address it as leaders. So if we're going to reskill, what are some of those skills that you see coming that's going to be the important thing for our workers and leaders to focus on?

I'll start. I think that as you think about generative AI, the internet made information accessible, democratized it. I think generative AI will democratize creativity. So the creative skills, the soft skills are going to be ever more important because everybody can participate in this journey of discovery because the next language is natural language. So everybody can participate. And in doing that, there's a corollary piece, which is data. We have to teach and train our workforce how to use data and what insights they can get out of data. Because without data, generative AI is not very helpful.

Matthias Patzak (21:18):
So with generative AI, we will see internal efficiency increases plus new business models going to emerge. With internal efficiency increases, so we are going to save 18 minutes, 80 minutes of a knowledge worker. But what are we going to do with this 80 minutes? More meetings, more emails, more PowerPoint? I hope this is not the solution.

Tom Soderstrom:
Hear, hear.

Matthias Patzak (21:41):
So one of an essential skill is that we give the workers the freedom to fill out this time, this creativity work as you said, in thinking about how to leverage current products, current customers, and current capabilities of an organization together with new technologies to help traditional organizations to develop generative AI-native business models. Because we are going to see generative AI native-challengers emerge. And so we need to really help customers to work on generative AI-native business models. And to do so, you need to master as a knowledge worker, prompt engineering. But I doubt that many organizations are going to invest in proper prompt engineering trainings. Why? Most of the office workers, probably 99% of office workers have Excel installed on their PC and on their laptop, but many people I know, they just know how to sum up a column or change the color of a cell. But Excel is such a powerful tool. And the same is true with generative AI. It's really, really powerful, but we need to train the people to make the most out of it.

Tom Soderstrom (22:52):
Good.

Matthias Patzak (22:53):
I'm curious.

Tom Soderstrom (22:53):
So soft skills, prompt engineering...

Matthias Patzak (22:55):
Creativity.

Tom Soderstrom (22:56):
And creativity.

Arvind Mathur (22:58):
Creativity. So I completely agree with you guys and I think it's important to get going and I'll share the story again. Sometimes to create a culture of learning and growth, especially when it comes to really advanced technologies which are indistinguishable from magic. Sometimes actually, a lot of our folks do think this is magical and they can never be able to do this.

So take an example, you mentioned Alexa earlier. In one of my previous experiences where I was trying to transform the organization to be more of a learning culture, and this is like eight years ago almost. We got a few Alexas in and we showed them how to make a very simple conversational tool. The beauty of it was people who had never done coding were able to build this Alexa app and suddenly this thing which was black magic to them was demystified.

Tom Soderstrom (23:52):
That's right.

Arvind Mathur (23:53):
And it created this energy, "Hey, I can learn this thing."

Matthias Patzak (23:56):
Yeah.

Arvind Mathur (23:57):
And it's important to do experiments like that that demystify this really complex thing. It's not that difficult. And with tools like Bedrock, et cetera, we are making this so simple to apply that if you get people to play with it, they will then develop this urge to learn and this culture of learning.

Tom Soderstrom (24:16):
That's a great point. I'm going to riff on that for a minute. So this is Conversation with Leaders. So how do we make the leaders use generative AI to become better leaders? Can they, and what would they do?

So I'll start with I tell you what not to do, which I made a mistake early on. Don't create the center of excellence because people will think they're excellent and everybody else will hate them in the organization. Instead, create a center of engagement, still a COE. But as leaders, we now engage everybody to adopt this new culture, learn, experiment, try things at very low risk, and very low cost. And I think if we do that, then we are... Who do you go to for your help with technology? Probably your teenagers. We all need to become technology teenagers. We need to learn this, get hands-on. Try Party Rock for instance. It's a great way to try it. And then learn enough where we can put it in context. So that's what I would think.

Matthias, what do you think? How can we use generative AI to make us better leaders?

Matthias Patzak (25:25):
Yeah.

Tom Soderstrom (25:25):
Can we?

Matthias Patzak (25:26):
Yeah, yeah, so my first impression when you asked the question was seeing a leader sitting before the laptop and then just chatting with ChatGPT or with any other generative AI tool. But this is not what leaders should do.

Leaders should go out of the office, talk to the people, inspire the people, co-create the strategy and drive business value. But to establish generative AI in an organization. So what is the secret sauce in modern organization?

It's autonomy and having loosely coupled small independent teams. But to be independent, they need to have the right data and they need to have decision-making authority. And generative AI as a tooling when being customized with your business processes, knowledge about your products and your customers, it can enable those independent teams making even much faster decisions with a lower blast radius when an experiment fails. And this is for me from an organizational perspective, the power of generative AI.

Tom Soderstrom (26:30):
That's great.

Arvind Mathur (26:32):
I completely agree with you both.

Tom Soderstrom (26:34):
I think I liked your earlier examples of giving you something to try, bring in the people who are doing it. So coming up with use cases would be my addition to that. Give people something to try.

Do you have any bold predictions and perhaps hopes for how generative AI will improve our workforce, make them even better, so to speak?

Arvind Mathur (27:03):
Yeah, I'm personally super excited. I think this is extremely transformative, huge potential. And there are many areas.

I'll talk about one which I personally feel very excited about is the area of education and talent development. My bold prediction is that anyone entering, any kid entering the education system now will have a completely different experience.

Now tell me something. We've all had those experiences where, "Hey, you know what? My fifth grade math teacher was so good, he or she took interest in me, understood me, and created this excitement for mathematics in me, et cetera." But we only think about that about very specific individual teachers. Imagine if every grade, every subject teacher was like that. And I think that's the potential generative AI has. I predict that kids who are entering the education system now will very soon have a super teacher for every subject in every grade who deeply understand their individual skills, and talent, and potential, and shape that in a way that over their education, not just one time but ongoing, every individual will be able to achieve their full potential based on their skills, interests, and talents.

Tom Soderstrom (28:22):
That's a beautiful addition.

Arvind Mathur (28:23):
I think that's… just imagine what the world would be if that becomes reality.

Tom Soderstrom (28:27):
I think all of us immediately, mine went to my old history teacher, who's that person. Matthias, I'm sure you have somebody. And all the listeners have. What if we could all have that? That's a beautiful vision.

Arvind Mathur (28:38):
That's awesome.

Tom Soderstrom (28:39):
Okay Matthias. Bold prediction.

Matthias Patzak (28:41):
So when the generation of my parents realized those productivity gains with different technologies, they were able to come from a 50-hour work week to a 40-hour work week. So I'm looking forward to a 30-hour work week.

Arvind Mathur (29:00):
Yes.

Tom Soderstrom (29:00):
Now we get more time to spend with our children.

Matthias Patzak (29:02):
Yep.

Tom Soderstrom (29:04):
So mine is going to be perhaps pretty wild and bold. One of my passions and hopefully all of our passions is sustainability. We got to save our planet. And how do you get people to care about it? To really care is to understand and interact with it. So using generative AI to ask questions about where's the next flood going to be? Where's the next fire? How much insurance is right for this environment? And then access the world's databases using generative AI to rewrite the query so that you can access it. Of course, assuming you had permission. That could help save our planet.

If you were going to recommend one thing that the listeners should do, and we should do, what would that be? Arvind?

Arvind Mathur (30:01):
I would say we've all heard that gen AI was the fastest growing thing. A hundred million people going on this within a few weeks, but there are still hundreds of millions who haven't.

Tom Soderstrom (30:11):
Yeah.

Matthias Patzak (30:12):
Yeah.

Arvind Mathur (30:12):
If you've not started using this daily, start now. Figure out your favorite app, download that and use it for your personal lives. This is transformational, like I said earlier. And the more we get familiar with it, understand its potential, its capabilities, and its limitations, the more opportunities we will have to become masters of this and make it work for us.

So get going now.

Tom Soderstrom (30:37):
Awesome. Matthias?

Matthias Patzak (30:40):
The people who are get going are probably using just a single tool and my recommendation would be go out and try out 3, 4, 5 new different tools, a different search engine, different tools for photo recording or whatever you can do. Go out, build, try out more tools, and experiment with the technology and not just with the single tool.

Tom Soderstrom (31:01):
So learn and be curious.

Matthias Patzak (31:02):
Learn and be curious.

Tom Soderstrom (31:04):
I would say, looking at the data, we're seeing that companies, individuals think and companies think that generative AI will do a lot of big things. But a much smaller fraction actually has a use case. So I would say come up with a use case that you can try very quickly, I'd say two weeks, and with very few people, two or three at the most. And then tell the story of that to success breed success. So I would say learn and be curious, check out new tools, and then start with a use case now.

So thank you very much, Matthias, Arvind.

Matthias Patzak (31:41):
Thank you.

Arvind Mathur (31:42):
This was good fun.

Listen now

Listen now

Listen now