Voiceover (00:03):
Welcome to the FutureWork Playbook. This podcast series features founders, investors and legal talent that will help you embrace technology and transform your organization for a better future. This series is hosted by Natalie Pierce, the chair of Gunderson Dettmer's Labor & Employment law practice. Natalie and her guests are committed to helping you develop new playbooks to elevate your game.
Natalie Pierce (00:29):
Hi, this is Natalie and welcome to this episode of the FutureWork Playbook. This season, we've been focused on generative AI, its legal implications and how to prepare workers for the future. Today, we welcome world-renowned New York Times bestselling author and transformational leader Charlene Li. Charlene is an expert on disruptive transformation in the future of work. She's advised hundreds of companies ranging from Adobe to Southwest Airlines, including, by the way, 14 companies on the Dow 30 chart, she's advised them on how they can develop winning strategies for disruptive growth.
(01:16):
She received her undergraduate degree and her MBA from Harvard. I recently listened to her course on generative AI and workforce transformation, listeners, it's fantastic and available on LinkedIn Learning. And her Leading Disruption newsletter is a must follow to keep your competitive edge with this profound technology. Charlene, thank you for joining us.
Charlene Li (01:46):
Thank you so much for having me.
Natalie Pierce (01:48):
So, diving into our discussion today, as I mentioned, you've been helping organizations navigate transformative technologies for many, many years now. What is it that makes generative AI different?
Charlene Li (02:06):
Well, first of all, it's really simple to use. Anybody with a web browser, on their computer, on their phone, can use it and that just makes it so democratized. You don't need to be a company to be able to get access to the software and the power of it so anybody can pick it up and use it. It's simple, super easy to use. The second thing is it is incredibly powerful. So, it has the power to give us, as individuals, this superpower to do things we could never have done before. I can now create coding and applications, I can write things, I can do analysis that could never have done on my own. I can do it so much faster and easier so it just gives us so many more capabilities.
(02:58):
And the transformative nature of this is that, as an organization now, it's coming in from all sides across every part of your organization and people are beginning to get these superpowers and it's impacting your workplace. And the most interesting thing is you don't control it because you can try to shut it off but, if people can see that these tools can give them the ability to do their work so much better, they're going to want to use them. So, that puts a lot of pressure on organizations to say what will we do with this technology. People have experienced it and it's really hard to put that genie back into the bottle.
Natalie Pierce (03:36):
I think you said it perfectly, the genie is definitely out of the bottle. And if you think that you, as an organization, can take away people's superpowers when superpowers, as you said, have now been democratized and, really, given it's easy to use nature, we are seeing generations, all different generations in the workforce really taking advantage, all different skillset. And like it or not, as you said, the genie out of the bottle so it's really a matter now of helping employers navigate. Let me ask you this. When did you first get a look? You've been following transformative technologies for decades and really helping with the incorporation of transformative technologies into the workplace but when did you first get a look at generative AI and what were some of your first impressions?
Charlene Li (04:34):
As soon as it came out in November of last year, I was playing with it, as you can imagine, what is this new thing? Even before then, when Dolly and some of the other imaging tools were available, I was just playing with the power to be able to create. I'm not an artist, so to be able to create art, to create beautiful things with words seemed like an incredible thing. So, when ChatGPT became widely available in November of last year, the ability to play with it, to have a do things and then, especially when ChatGPT four came out, again, it's just so powerful as a tool. I could see that a good chunk of the work that I do, 70% of the work that I do is being impacted by these new tools.
(05:19):
So, it was hugely transformational for me right from the very beginning and I just started playing with it and doing things with it to the point now where I even have my own private GPT against my own content so I can talk to myself, brainstorm with myself, write with myself as a collaborator by using these tools.
Natalie Pierce (05:38):
Yeah, I loved that. Can you share with our listeners a little bit more? The last time we spoke, you talked about a chatbot that you've been developing with your own content.
Charlene Li (05:48):
Right. I call it CharGPT, just to be [inaudible 00:05:51]-
Natalie Pierce (05:51):
I love that.
Charlene Li (05:52):
... ChatGPT and I have it using ChatGPT 3.5 and also GPT four, and it's an off-the-shelf piece of software that I have uploaded my last book, my most recent columns, especially the ones that talk about disruptive transformation and generative AI. So, I know the content is really well formatted, it's really clean, it's fairly structured and then I can query myself against it. It's private so it's not feeding the general GPT engines and I can literally have it summarize things, I can ask myself questions and see how I would answer it. So, it's a shortcut to being able to talk to myself and I even use it to create some first drafts, just some initial first drafts to get an idea to give myself a shortcut which I then can edit.
(06:43):
But now I'm starting to upload, because I'm writing a book about generative AI, again, into a private database that only I have access to, myself and my co-author. And we're putting in our notes, we're putting in our interview notes so it's a lot looser but we can go and query and find all the examples of generative AI councils and summarize the top approaches. And it uses the power of GPT against my own bespoke content to create some summaries that would've taken us hours to pull over all these interview notes to create the same thing.
Natalie Pierce (07:22):
I love that. And the thing we keep, I think, on a daily basis, getting surprised by what generative AI is capable of doing. And so, for you to be able to feed your own knowledge and insights, I'm sure you're seeing some correlations being made that maybe you hadn't thought of before and that is the power of being able to train it with your ... To feed into this creation of this model thoughts. Are you finding, are you and your co-author looking at new correlations that you perhaps hadn't seen or thought of before?
Charlene Li (08:04):
Well, the thing that it keeps ... Again, it's very dependent on the prompts that you put into it and we're just beginning to play with these things. But I think the real lesson to be learned from this is that the real power of these tools isn't the general knowledge base. Again, it's very powerful for that but it's in the very tailored pointing and the use of these technologies against a very specific set of data. So, if you are a company and you have a ton of information inside your own company, of the expertise of your employees, of the ways that customers, individual customers have interacted with you and what has worked and what hasn't worked, being able to tap into that data and use it in new and different ways to support your employees for them to find and collaborate with each other, just even internally, nothing even externally facing.
(08:59):
Because one of the biggest challenges inside of organizations is I know somebody in my organization knows something about this but I don't know where. Does the left hand or the right hand know what each other are doing? How do we stay on top and aligned with each other? And to use these tools to just tap into the internal data is such unexplored areas of opportunity, I think. There's all the customer facing information that you have that's unique to you and your customers at the individual level that could create new customer experiences and new products and services, again, on behalf of your customers that are just beginning to explore.
Natalie Pierce (09:37):
Charlene, I think you and I both agree that, at this point in time, embracing change is not only a choice but a necessity for survival and progress and that the shift toward the widespread adoption of generative AI really is inevitable. And I know that's why you've been in such high demand, particularly, probably, in this last year but you work with so many boards and executives across industries. I'm wondering, what are the most common questions that you're hearing about generative AI?
Charlene Li (10:21):
I want to emphasize, it is still really early and most of the questions I get are still what can this do that can help my company. So, I get an idea of what it is at a high level, I play with it a little bit but what should we be focusing on, how do we prioritize what we do, what are the best use cases or applications inside my company and that's a very valid question at this point. And the overwhelming reaction I get from people is that this is a lot, how do I get my hands around this, how do I even take the first steps to figure this out because it feels like it's hard, it's a technology space, I don't understand it.
(11:05):
And what I encourage people to do is step back away from all the use cases and instead ask a very simple question whenever somebody comes to you and says, "Hey, we should look at this technology, whatever," ask them a very simple question and it's a strategic one. How is this going to help us achieve our strategy? How will this potentially put our strategy that we have defined at risk? What is reinforced and remains even more true now about our strategy? What things are challenged and no longer true about our strategy and we may need to take it in a different direction? That is the question for the top executives and for the board to be asking. Not what the technology does but a deeper question is how does this technology help or hurt us.
Natalie Pierce (11:51):
I like that. And one thing I've heard you say before, Charlene, I think it's so true, is that it's very important right now to just begin to know that very few people know what's really going on. We don't have a lot of answers but, as you've advised, you've got to be looking at what the questions are and that there's a lot of interesting questions out there and having that experimental mindset is really what's required to be successful in this space.
Charlene Li (12:24):
Right. And it's hard because, when we think about strategy, we think about things being etched in stone. This is the path we're going on and we're 100% sure this is going to be the right direction so go on it. And in the generative AI space, we're not even sure what the technology can do, why it does it, what it's going to look like tomorrow. My experience every morning as I wake up, I look at a couple of my feeds, I get a couple emails and I feel like my hair is on fire because so many things have yet again changed. So, how do you create a strategy in the face of so much change and uncertainty? And to your point, you have to get started.
(13:05):
So, you start with the things that are not going to change. You start with your values, you start with your strategy, those are the firm foundations from which you look at all of these technologies. And I'm a big believer in starting with those things, especially the more convoluted, confusing and uncertain the spaces, be really clear about what your strategy does and what it doesn't do because what strategy always is. And then ask the questions of how is this going to help our strategy or hurt it. I keep coming back to that and be very focused and directed on that because, if something is interesting but it's tangential, it's more tactical, then you can go either it's not important or it's a no-brainer, just do it. If it falls into those two categories, address it and then move on to the bigger ones that require your attention.
Natalie Pierce (13:54):
Yeah, [inaudible 00:13:55].
Charlene Li (13:55):
So, can we write better copy and improve our marketing with these tools? Yeah. Is that a strategic issue that's going to make or break what we do? No. Then just go ahead and do it. If it's something that is strategic, I know one organization said that customer experience is our number one competitive advantage so they're going to take a close look at how copywriting can be taken to the next level and, not just something that's a more tactical thing, but it is something strategic to them because it adds to that experience. So, they're going to look at that very closely, invest in it, experiment with it and very carefully move into that because something that could make or break their strategy.
Natalie Pierce (14:35):
I would love if we could stick on that point for a moment, just about this company that you're working with and their space is really customer experience. Can you give any examples of how that company or any other client you've worked with has really seen the benefits of embracing generative AI in the workplace and really aiming toward that strategy?
Charlene Li (15:07):
Sure. Again, something that's been in the works for quite a while is, when you call into a call center, now, they can see from the initial prompts how fast you're typing, your tone and inflection of your voice, how irate or urgent something is and they can cue up the scripts for the agent that identifies that this person's upset so you may want to ask these questions or-
Natalie Pierce (15:35):
Oh, wow.
Charlene Li (15:36):
... try to calm them down. So, that's already been out there for a while. Now what happens is, with generative AI, it'll not only create the scripts but they'll pull up, for example, and, again, this is all conjecture and still very much in the works, they recognize that person from their phone number and pull up their entire history and then write the scripts in real time for the agent to use because that what generative AI does, it writes actual scripts based on a set of data or information in the context. And if they can understand the emotions of that context and, in real time, cue the responses and generate more text for the agent as suggested, again, the agent can use their own judgment, but they're creating options and texts and pulling up customer service records or responses or ...
(16:31):
Again, especially for support calls, it's an entire database of hundreds of thousands of documents that will say, "Oh, yeah, here's a possible solution or ask these questions to narrow it down even more." So, you're not dependent on the knowledge of that individual customer service representative but they have the power of the entire organization and the history with that client, with that customer, to be drawn into real-time scripts and that will improve your customer experience, for example, in a significant way.
Natalie Pierce (17:05):
That's a perfect example. To be able to help pull up that history, get the personality, the prior exchanges, I think that is an excellent example of human machine teaming, this augmented collaboration. We can and should learn to combine people and technology in a way that really leverages the respective strengths of both. And so, to put that kind of information, as you said, for the customer service individual to do what they would like with it but to have that real time feedback to better set an organization for those successful communications that lead to better outcomes, lead to better productivity, profits, workplace environment, et cetera. That's such a great example.
Charlene Li (18:01):
Right. And there was recently a study done by nine leading researchers with a consultant from BCG where they followed the tasks that people were given and granted people who they were able to do more tasks and less time with better quality with generative AI, in particular, OpenAI and ChatGPT, than without it. But what was really interesting and this is what I'm seeing play out in a couple of other places too, the most impactful error were for the lowest performers, people who were not great at certain things, they became really good at it. So, the quality increased substantially and the speed of which they did things increased also.
(18:44):
And when I was speaking to one organization that had a lot of process and call centers and things, they go, "When we saw this, we saw, not an opportunity to save costs, but to increase the performance of every single person but especially of our entry level, lower performing people to make them top performers. And then give our top performers the ability to go do all the things that we've wanted to do on behalf of our customers but we didn't have the time, people and money." So, I think the real opportunity here isn't that we save costs and take it right down to the bottom line to increase profits but that, if you really want to grow, in the past, if you wanted to grow, you had to add new people and take on new costs.
(19:28):
And I think the real opportunity here is that we're going to have this really interesting window of time where we can reverse the growth equation, where you can actually grow substantially with the same people, without spending any more money, without any more time then you can grow.
Natalie Pierce (19:48):
I couldn't agree more. I think, so many people, they are apprehensive about the impact that generative AI can have on businesses, jobs, the economy and many see this revolution instead as a real opportunity to shape our future for the better. And even in your example of taking an entry level position, somebody who might not have as high of a skillset, and giving them this tool to make them be able to effectively do their job better. And historically, we really have seen that, while job destruction certainly is a real effect of automation, we've seen that historically but what we've also seen every single time is the efficiency and productivity gains from automation helping to create jobs.
(20:46):
And Charlene, you mentioned the ... I don't know, you probably did see this recent Wharton Business School study that came out, I forget the name of the professor but-
Charlene Li (20:58):
Ethan Mollick, mm-hmm.
Natalie Pierce (21:00):
Oh, okay. Yes, of course. But I was blown away to see this experiment with two groups of Boston consultants. If you're going to get to BCG, you're going to be top of your game to begin with, you even know who the professors. You want to share with the listeners that recent idea? And I have the stats in case you [inaudible 00:21:26].
Charlene Li (21:26):
Yeah, I think it was 21% increase in productivity.
Natalie Pierce (21:30):
25%.
Charlene Li (21:32):
25%, right, right.
Natalie Pierce (21:34):
Yes, yes.
Charlene Li (21:34):
And this is a very informal tool. They segmented a couple hundred consultants and gave them a fake case to work on, basically, I think it was [inaudible 00:21:44].
Natalie Pierce (21:43):
Yeah.
Charlene Li (21:44):
And they measured how quickly they were able to do things and how well. And they've had various rankings, as you can imagine, of skills and capabilities of different consultants and they found the lowest level consultants were able to increase by 25%. I think even the highest level consultants were able to increase it but less so, their increase was, I think, only 11%. So, across the board, you see it but the real biggest benefit are the people who are at lower levels of capabilities. So, again, that's the AI that we have today and this is just off the shelf, untuned to your specific company, again, just really generic, very early, relatively primitive tools compared to what we'll have in the future.
(22:34):
So, it's very encouraging and most productivity gains that we see from technology are usually in the single digits. And this, again, without any tuning, is giving us 25% productivity gains, we've never seen anything like this.
Natalie Pierce (22:48):
[inaudible 00:22:49]
Charlene Li (22:49):
And depending on the [inaudible 00:22:50] you're seeing sometimes, 50, 100, 2X, 5X gains in productivity. What do you do with that?
Natalie Pierce (22:59):
It's unbelievable what's happening and just wait until we see you start seeing the next generation of LLMs that are just right on the horizon. And you talked about your waking up every morning being like, "Okay, what did we find today? What did we learn today?" I'm just blown away by what we're seeing with these impressive multimodel capabilities, having generative AI that can now process and understand multiple types of data inputs.
(23:41):
And then, going back to the BCG consultant study done by the Wharton Business School professor, not only did they see that 25% increase in terms of the productivity of tasks being completed, but also 40% higher quality results than those consultants who were told that they couldn't use artificial intelligence to augment their efforts. So, when you talk about superpower, Charlene, without doubt, is a superpower.
Charlene Li (24:16):
Right. It's less the fear if that AI is going to replace humans, it's much more likely it's humans with AI replacing humans without AI.
Natalie Pierce (24:25):
Would not have said it better, that's exactly, exactly the case. I'm going to move us on a little bit. This whole notion of getting started, this is something you've been so passionate about for such a long time and you help companies with baby steps. But now, there's something about a sense of urgency like, "Okay, ready, go, now." So, even setting companies up to start now with your 90-day plan, don't worry about the fact that, historically, it's been a long-term strategic plan of years, we're going to have to be a little more flexible, maybe be willing to change.
(25:12):
So, I know this is a big question and something that you've dedicated, really, your career to. But in the context of generative AI, what are some first steps that a company can take to encourage the responsible use of generative AI as they start experimenting with it?
Charlene Li (25:32):
Right. And I think the first thing to overcome is that you should be doing something. And the biggest fear is, well, is it too early because I don't want to be one of those early adopters that falls on their face. And I go, "It's not too early. There are immediate gains that you can use and you should absolutely, at the minimum, knowing what you can and can't do with it and then strategically decide you're not going to do something because it's not the right time, the technology isn't there, whatever, but make it an active decision rather than a default one.
(26:06):
And at a minimum, every company needs to at least play defense. You have to have a policy, guidelines, a code of conduct of something that tells your people what they can and cannot do with generative AI because they have access to it, as we've talked about before, they are probably using it in their personal lives and they're wondering, "Well, can I use this at work or can't I?" One company said, "Well, we should take a survey and really understand is this something that our employees are interested in using?" I go, "You can dispense with the survey, I can guarantee you, they're using it. They're aware of it, they're probably using it already." They go, "Do you think?" I'm like, "Oh, yeah. Go check your logs to see how many times ChatGPT is being actually used and accessed." And I go, "I guarantee you, people are using it."
(26:59):
And so, have a point of view of what they can use it for, what's good and safe, what do they need permissions, what training you need to have. So, you have to at least have the defense in place to protect your employees, protect the company, protect your customer data, make sure that there's data governance in place. And I feel like, especially for boards, it's a fiduciary responsibility that you make sure your company has these things in place at a minimum and then you can go and decide what else you want to do.
Natalie Pierce (27:30):
Absolutely. And speaking of the board's responsibilities, and I'm really glad that you brought that up, but what we've certainly seen is that adoption is slowed and competitive. Edge is lost when executive leadership is not charged with and really held accountable to present the board calculated options to make smart decisions around things like generative AI adoption. And I think we both agree that that really is part of the board's duty of oversight, making the right decisions around strategic plans. And I'd like to ask you, I know you advise a number of boards, but how might we tie oversight responsibilities of the board of directors to the strategic plan of the company when it comes to generative AI?
Charlene Li (28:29):
Well, I think the board of directors, at a minimum, has a conversation with the executive team and CEO and say we want to understand how much of an opportunity or threat is generative AI to our strategy and come back. Even in this conversation, where does our strategy, where is it vulnerable, where are the opportunities? And even at the high level, if you know just a little bit of what generative AI can do, you don't need a lot and it doesn't take a lot of imagination to think, okay, here are the broad strokes of my strategy and here are my initial gut feelings of where it could be. And I will say, having worked with people just for a couple hours, just working through their actual strategic plan and looking where generative AI could help or not or hurt it, you have a pretty good idea after just an hour or two of just playing with it that you know what the big buckets are for exploration and you probably have an idea where it's not.
(29:31):
So, again, every company is different, every industry is different and, depending on how good your data set is, how advanced you are, you can start with an initial list of big buckets with which to explore.
Natalie Pierce (29:48):
That's great.
Charlene Li (29:48):
But you should know what those buckets are because you know your strategy really well. You don't need to know the technology really well, you need to know your strategy really well. And any company worth its weight, it should know its strategy really well.
Natalie Pierce (30:02):
Absolutely. And one thing I've heard you say that I absolutely love and I think it's true is that you have to be willing to change your relationship with risk. It can't just be the safe wait and see and, like you said, just getting started with those discussions now, if you haven't already, it's critically important. I think we're going to find a lot of companies whose failure to adopt change, it really will leave them marginalized and they're likely to become casualties rather than beneficiaries of these types of transformative changes that we seek when on generative AI.
Charlene Li (30:47):
It's interesting, the companies I'm finding who have an easier time, not that it's easy ever, but an easier time working with generative AI are the ones who have really gone through a true transformation of the workforce and of their leadership, their culture and their strategy because of things like digital transformation. They didn't just put in place a technology, they actually transformed their organization to be more agile, to be able to deal with uncertainty, to have a right size relationship with failure, not to have all the answers all the time and instead really seek out how to ask really good questions and, therefore, lead in a very different way.
(31:31):
Because generative AI is putting right in front of us the ability, we have to have the ability to be able to be flexible, to be agile, to be resilient in the face of these changes. And I think this whole thing of moving your leadership from having all the answers to asking great questions is especially important when our employees become super powered, become these super humans. They don't need you to tell them how to do their job anymore, they know how to do their job better than you ever did. So, what does that leave you as their leader? What is the guidance that you can provide them? And it becomes much more around shaping the direction of travel, creating the guidelines, making it so that we can use those governance, the nasty word governance, not to slow things down, but to speed things up. It's a very different way of approaching leading and managing.
Natalie Pierce (32:28):
So true. Approaches to organizational effectiveness and change, really, as you said, they become possible when you have cells of agile culture that are building up around successful generative AI implementation. And again, couldn't agree with you more that agility, now probably more than ever, is key. One thing that you've done a great job with is helping organizations understand why it's important to form an AI council. Can you speak with our listeners a little bit more about that?
Charlene Li (33:10):
Sure, yeah. The AI council is, again, usually cross-functional, more than just your IT department which is where AI traditionally has been, it's been treated like a technology. But because generative AI is hitting across your organization and it has strategic and business implications, you need that representation across your organization and especially from your business leaders, your line of business, your commercial representation, your operations people, HR, legal, for example, and you want to keep that council as small as possible but have the most important people on it. And I've seen every type and shape of AI council and the best ones are ones that are, again, derived from their strategy.
(33:58):
If your strategy is very focused on having the best products and you want your creative people, the people who understand the customers, very close to them, the product managers and the people who maybe ... And customer data because you're going to be tapping into that as well as IT. But an organization like the one I was talking about with customer experience is going to want the business heads, the people in operations who actually make it happen and the head of HR, for example. So, it's depending on where you see those big strategic buckets again, you can pull the people from those big areas, those key functions, into the core group and then line up people who will be adjacent to it that you want to consult or be informed with, be informing about decisions that may be coming down.
(34:47):
But the idea of the AI council is everyone who needs to be in the room to make a decision is there. That's the key thing because you want to be able to make decisions and move forward. So, having those guidelines, again, about the defensive guidelines of how far you can go, how fast you can go, a clear direction from your executive team and board about the big buckets to explore to support the strategy and then let this team go at it.
Natalie Pierce (35:12):
I think that's great advice. Have it be a small AI council supported by an extended team so that you can hold onto that necessary agility to pivot and move quickly. Charlene, as we know, some companies are continuing to take this wait and see approach. What are some of the risks that you see if companies fail to embrace generative AI quickly?
Charlene Li (35:44):
Well, the number one thing is that your employees will start losing faith in the organization especially if they feel hamstrung that they can see this thing, they can see it on their computer, they can see it on their phones, they're seeing their friends and colleagues at other companies doing these amazing things like, "You will not believe what I could do at work yesterday." And you're thinking, "Oh, man, I wish my company was worth it enough that I could be allowed to use that but they just keep shutting me down." But guess what happens to that employee? They're going to leave. They're going to go to another company that will support them, allow them to reach their maximum potential. And so, you'll be losing some of your best people.
(36:26):
You won't be able to recruit people. It means that, when you're hiring for somebody, you're not hiring the best people because you're not asking for people. If you're hiring a copywriter, can they write but can they also use these AI tools in the right way? You don't know how to evaluate that even because it's not a skill that you've asked for and developed and understand so you're not hiring the best people who are at the top of their game. It also means that a strategy perspective, with each month that you delay, you're just going to be that much slower in implementing any new opportunities or defensive opportunities and your competitors could be moving ahead and you're, again, you're allowing people who may be even more laggard than you to leapfrog you.
(37:10):
So, I'm not saying that you have to do everything but you need to have an opinion on when is the right time and you can absolutely decide that what we really need is a private LLM to be able to make this happen. And we could do that today but it would be hugely time-consuming, expensive, we just do not have the expertise so we're going to wait a few months for these things to become more integrated into our existing enterprise tools and, when those are available, we'll turn them on. In the meantime, we're going to do these stock gap measures but we're not going to go out and build our own private LLM. Do you see the difference in the terms of-
Natalie Pierce (37:47):
Absolutely, yeah.
Charlene Li (37:48):
... not waiting versus saying we're not going to do anything?
Natalie Pierce (37:52):
Exactly, exactly. If you put up your defenses, I think, employees, they just want to know what are the guardrails. How can I use my superpower responsibly? But if you do not allow employees to meet their potential which these tools do, I think you're absolutely right, Charlene, that you're going to see people go to the companies that will and they'll be viewed as more likely to keep that competitive edge into the future. But I think it's misguided to think that employers can just wait and build their own. And there certainly are wonderful enterprise solutions available that will allow you to keep your data private but, yeah, couldn't agree more this thinking that we can somehow control what employees who are so hungry for knowledge and to advance in their careers and improve their skills and knowing how easily available the generative AI tools are telling them no is not the right answer.
(39:07):
Charlene, for early adopters, where do you see generative AI heading and what do you expect to see in the future? So, where is generative AI heading?
Charlene Li (39:20):
Right. Again, we talked a little bit about the private LLMs but I think we've been hearing a lot about these co-pilots, these agents, so to speak, that can just act on your behalf and be your collaborator. So, initially, they'll be co-pilots, they'll be sitting there next to you ready and waiting for your commands to just be your assistant. You think about having an amazing assistant who's capable of doing so many things and they're right there. The real way that these things become powerful is that you granting them more and more access to your information.
(39:56):
So, to process my emails, to write really good emails, they need access to my emails in the past history of how I've written them and also my communications with that particular person or that particular team. And I have some now that are built into the email tools that I use and it is quite helpful to summarize a chain of 20 different emails and to be able to just write a quick summary and then do whatever I want so I can quickly write more complex emails in that manner. I don't use it for everyone but having that co-pilot there using my voice, super effective and helpful. But in the future, I think these co-pilots become a lot more autonomous. I'll give it an example.
(40:40):
When I book travel now, I don't use an agent anymore, I basically have to go and find my ticket and do everything by hand. And I'm like, "Wow, I remember the days when I could just call somebody and have them do all that for me." And I think the agents will be able to do that but they'll understand I prefer these airlines or, looking at my calendar, these are the times will be best times for me to fly or not and it'll understand all my preferences and do all of that for me in the background because it follows me around and knows what my preferences are for everything. From understanding my travel preferences to how I like certain meetings with certain people to when I need to have a break because I've just been in back-to-back meetings so it'll hold the calendar and not schedule anything at that time.
(41:27):
So, the biggest question here is how much autonomy do we want to give these tools? How comfortable are we sharing that information? How comfortable are we letting these things make decisions on our behalf? And that's a very different type of tools, I think, that we're still going to have to develop our comfort levels at doing that. In the same way, we had to move from using bank tellers to get our money to getting it out of a machine and trust it. Every single time. They do it correctly. So, it's going to take some more time for us to do that. The technology can do a lot of these things already but I don't think we're ready as humans to necessarily seed that kind of control yet.
Natalie Pierce (42:10):
I think that's right. Though some of the examples you gave, I think, yeah, I would take that in a heartbeat today but it's true. I think that we'll continue to see some of these emergent or unanticipated qualities or behaviors and I think you're right that, the more that we can provide our own information, our own communications, past writings, et cetera, we will see with this next generation of LLMs that really are right on that horizon, progress in areas like personalized reasoning, common sense and judgment and that will then call the question of how much of us are we willing to seed. I think those will be very important questions.
(43:09):
You talked a little bit about the fact that you have a book coming out. Can you tell us any more, just something to preview, what you're focused on with your co-author and when we might expect it to come out?
Charlene Li (43:25):
Sure, yes. I'm writing it with a dear friend, Katia Walsh, who's the chief digital officer at Harvard Business School and she has deep, deep experience being a chief data officer and AI officer at very large companies all around the world so invaluable experience that she's pouring into the book. And the tentative title right now is called Mega Transformation because both of us have been in the technology and business space for over 30 years now and we have never seen anything like this The speed, the potential for transformation and how, across the board, it is, again, from an organizational point view but also across society.
(44:05):
So, we think this is something hugely transformational and, as such, we have to approach it in a very different way of setting our strategy. So, the book is called Mega Transformation, how do we create a rigorous generative AI strategy in 90 days. And we lay out week by week all the things you have to do to address inside of that strategy and how to approach it in a rational way that's starting with your strategy and your values and who you are as a company. Rather than developing thousands of use cases of what AI can do and then narrowing it down, you're starting with something you already know that's true which are these foundational elements of your company.
Natalie Pierce (44:44):
So, now that everyone has heard this, everyone is going to want to get a copy of Mega Transformation. What is anticipated date of publication? Do you think perhaps this year, next year?
Charlene Li (44:58):
I wish this year but the way publishing works is not working in our paper. We're still writing it and so we're finishing up the manuscript this fall and we'll hope to publish it in January.
Natalie Pierce (45:08):
Wonderful, wonderful. So, Charlene, listen, we're just about out of time but I'd love to wrap up with a quick fire challenge if you're willing to participate.
Charlene Li (45:19):
Sure, go for it.
Natalie Pierce (45:21):
Great. Listen, I heard that you make your own kombucha, have you picked up any other new skills recently?
Charlene Li (45:31):
Well, I don't know if it's still recent but I really enjoy teaching my pet, my cat how to do tricks. And so, we're always learning new tricks and I am visiting my brother, he has a brand new kitten so, of course, what did I do last night? I'm teaching this cat how to sit and training. And people think, "Wait a minute, this is not a dog. This is a cat, right?" Yeah, I don't teach dogs, old dog new tricks, I do what seemingly is impossible. I go and teach cats how to do tricks and they do it, they will absolutely do tricks.
Natalie Pierce (46:04):
You got the kitten to sit?
Charlene Li (46:06):
Yeah, yeah. They do this natural thing when you take your hand and sweep up your, they say, love fingers, they will follow your finger motion up. And when they actually look up, they automatically will sit and then you give them lots of praise or treats every time they [inaudible 00:46:20] a command set and they learn how to sit.
Natalie Pierce (46:23):
Charlene Li, kitty whisperer. That's amazing, I really didn't know that that was possible. One more question. If you could go to dinner with an AI avatar, who would it be?
Charlene Li (46:38):
Well, if I could sit down with Satya Nadella who's the CEO of Microsoft. I have tremendous admiration for how humanly he leads the company with empathy and he just always, from the very beginning, has said it's all about empathy for each other, for our customers, for ourselves and using that. Having a leader that's so values and humanistically oriented is so rare, you just don't hear many CEOs talking like that. So, just to have dinner with him and just understand how he thinks, I've read pretty much everything he puts out there. So, I know he's not necessarily available but I would just love to be able to sit there and have conversation with Satya's avatar and just ask him tons of questions.
Natalie Pierce (47:28):
Yeah, he's truly a remarkable leader who really came up through the organization. And as we've had founders as guests on the podcast, his name has certainly come up on multiple occasions so I love that answer. So, we are out of time. Charlene, it's been so great talking to you. Your work is so important and you're truly an inspiration. Thanks for sharing your perspectives on generative AI and your vision on the future of work.
Charlene Li (48:08):
Thank you so much again for having me.
Natalie Pierce (48:09):
And thank you, our audience, for listening.
Voiceover (48:13):
You've just listened to the FutureWork Playbook. This podcast series is brought to you by Gunderson Dettmer, the world's number one law firm representing venture capital funds and high growth companies. Join Natalie Pierce on our next episode as she and her guests help prepare your organization for the future. Please subscribe to the FutureWork Playbook.
Welcome to the FutureWork Playbook. This podcast series features founders, investors and legal talent that will help you embrace technology and transform your organization for a better future. This series is hosted by Natalie Pierce, the chair of Gunderson Dettmer's Labor & Employment law practice. Natalie and her guests are committed to helping you develop new playbooks to elevate your game.
Natalie Pierce (00:29):
Hi, this is Natalie and welcome to this episode of the FutureWork Playbook. This season, we've been focused on generative AI, its legal implications and how to prepare workers for the future. Today, we welcome world-renowned New York Times bestselling author and transformational leader Charlene Li. Charlene is an expert on disruptive transformation in the future of work. She's advised hundreds of companies ranging from Adobe to Southwest Airlines, including, by the way, 14 companies on the Dow 30 chart, she's advised them on how they can develop winning strategies for disruptive growth.
(01:16):
She received her undergraduate degree and her MBA from Harvard. I recently listened to her course on generative AI and workforce transformation, listeners, it's fantastic and available on LinkedIn Learning. And her Leading Disruption newsletter is a must follow to keep your competitive edge with this profound technology. Charlene, thank you for joining us.
Charlene Li (01:46):
Thank you so much for having me.
Natalie Pierce (01:48):
So, diving into our discussion today, as I mentioned, you've been helping organizations navigate transformative technologies for many, many years now. What is it that makes generative AI different?
Charlene Li (02:06):
Well, first of all, it's really simple to use. Anybody with a web browser, on their computer, on their phone, can use it and that just makes it so democratized. You don't need to be a company to be able to get access to the software and the power of it so anybody can pick it up and use it. It's simple, super easy to use. The second thing is it is incredibly powerful. So, it has the power to give us, as individuals, this superpower to do things we could never have done before. I can now create coding and applications, I can write things, I can do analysis that could never have done on my own. I can do it so much faster and easier so it just gives us so many more capabilities.
(02:58):
And the transformative nature of this is that, as an organization now, it's coming in from all sides across every part of your organization and people are beginning to get these superpowers and it's impacting your workplace. And the most interesting thing is you don't control it because you can try to shut it off but, if people can see that these tools can give them the ability to do their work so much better, they're going to want to use them. So, that puts a lot of pressure on organizations to say what will we do with this technology. People have experienced it and it's really hard to put that genie back into the bottle.
Natalie Pierce (03:36):
I think you said it perfectly, the genie is definitely out of the bottle. And if you think that you, as an organization, can take away people's superpowers when superpowers, as you said, have now been democratized and, really, given it's easy to use nature, we are seeing generations, all different generations in the workforce really taking advantage, all different skillset. And like it or not, as you said, the genie out of the bottle so it's really a matter now of helping employers navigate. Let me ask you this. When did you first get a look? You've been following transformative technologies for decades and really helping with the incorporation of transformative technologies into the workplace but when did you first get a look at generative AI and what were some of your first impressions?
Charlene Li (04:34):
As soon as it came out in November of last year, I was playing with it, as you can imagine, what is this new thing? Even before then, when Dolly and some of the other imaging tools were available, I was just playing with the power to be able to create. I'm not an artist, so to be able to create art, to create beautiful things with words seemed like an incredible thing. So, when ChatGPT became widely available in November of last year, the ability to play with it, to have a do things and then, especially when ChatGPT four came out, again, it's just so powerful as a tool. I could see that a good chunk of the work that I do, 70% of the work that I do is being impacted by these new tools.
(05:19):
So, it was hugely transformational for me right from the very beginning and I just started playing with it and doing things with it to the point now where I even have my own private GPT against my own content so I can talk to myself, brainstorm with myself, write with myself as a collaborator by using these tools.
Natalie Pierce (05:38):
Yeah, I loved that. Can you share with our listeners a little bit more? The last time we spoke, you talked about a chatbot that you've been developing with your own content.
Charlene Li (05:48):
Right. I call it CharGPT, just to be [inaudible 00:05:51]-
Natalie Pierce (05:51):
I love that.
Charlene Li (05:52):
... ChatGPT and I have it using ChatGPT 3.5 and also GPT four, and it's an off-the-shelf piece of software that I have uploaded my last book, my most recent columns, especially the ones that talk about disruptive transformation and generative AI. So, I know the content is really well formatted, it's really clean, it's fairly structured and then I can query myself against it. It's private so it's not feeding the general GPT engines and I can literally have it summarize things, I can ask myself questions and see how I would answer it. So, it's a shortcut to being able to talk to myself and I even use it to create some first drafts, just some initial first drafts to get an idea to give myself a shortcut which I then can edit.
(06:43):
But now I'm starting to upload, because I'm writing a book about generative AI, again, into a private database that only I have access to, myself and my co-author. And we're putting in our notes, we're putting in our interview notes so it's a lot looser but we can go and query and find all the examples of generative AI councils and summarize the top approaches. And it uses the power of GPT against my own bespoke content to create some summaries that would've taken us hours to pull over all these interview notes to create the same thing.
Natalie Pierce (07:22):
I love that. And the thing we keep, I think, on a daily basis, getting surprised by what generative AI is capable of doing. And so, for you to be able to feed your own knowledge and insights, I'm sure you're seeing some correlations being made that maybe you hadn't thought of before and that is the power of being able to train it with your ... To feed into this creation of this model thoughts. Are you finding, are you and your co-author looking at new correlations that you perhaps hadn't seen or thought of before?
Charlene Li (08:04):
Well, the thing that it keeps ... Again, it's very dependent on the prompts that you put into it and we're just beginning to play with these things. But I think the real lesson to be learned from this is that the real power of these tools isn't the general knowledge base. Again, it's very powerful for that but it's in the very tailored pointing and the use of these technologies against a very specific set of data. So, if you are a company and you have a ton of information inside your own company, of the expertise of your employees, of the ways that customers, individual customers have interacted with you and what has worked and what hasn't worked, being able to tap into that data and use it in new and different ways to support your employees for them to find and collaborate with each other, just even internally, nothing even externally facing.
(08:59):
Because one of the biggest challenges inside of organizations is I know somebody in my organization knows something about this but I don't know where. Does the left hand or the right hand know what each other are doing? How do we stay on top and aligned with each other? And to use these tools to just tap into the internal data is such unexplored areas of opportunity, I think. There's all the customer facing information that you have that's unique to you and your customers at the individual level that could create new customer experiences and new products and services, again, on behalf of your customers that are just beginning to explore.
Natalie Pierce (09:37):
Charlene, I think you and I both agree that, at this point in time, embracing change is not only a choice but a necessity for survival and progress and that the shift toward the widespread adoption of generative AI really is inevitable. And I know that's why you've been in such high demand, particularly, probably, in this last year but you work with so many boards and executives across industries. I'm wondering, what are the most common questions that you're hearing about generative AI?
Charlene Li (10:21):
I want to emphasize, it is still really early and most of the questions I get are still what can this do that can help my company. So, I get an idea of what it is at a high level, I play with it a little bit but what should we be focusing on, how do we prioritize what we do, what are the best use cases or applications inside my company and that's a very valid question at this point. And the overwhelming reaction I get from people is that this is a lot, how do I get my hands around this, how do I even take the first steps to figure this out because it feels like it's hard, it's a technology space, I don't understand it.
(11:05):
And what I encourage people to do is step back away from all the use cases and instead ask a very simple question whenever somebody comes to you and says, "Hey, we should look at this technology, whatever," ask them a very simple question and it's a strategic one. How is this going to help us achieve our strategy? How will this potentially put our strategy that we have defined at risk? What is reinforced and remains even more true now about our strategy? What things are challenged and no longer true about our strategy and we may need to take it in a different direction? That is the question for the top executives and for the board to be asking. Not what the technology does but a deeper question is how does this technology help or hurt us.
Natalie Pierce (11:51):
I like that. And one thing I've heard you say before, Charlene, I think it's so true, is that it's very important right now to just begin to know that very few people know what's really going on. We don't have a lot of answers but, as you've advised, you've got to be looking at what the questions are and that there's a lot of interesting questions out there and having that experimental mindset is really what's required to be successful in this space.
Charlene Li (12:24):
Right. And it's hard because, when we think about strategy, we think about things being etched in stone. This is the path we're going on and we're 100% sure this is going to be the right direction so go on it. And in the generative AI space, we're not even sure what the technology can do, why it does it, what it's going to look like tomorrow. My experience every morning as I wake up, I look at a couple of my feeds, I get a couple emails and I feel like my hair is on fire because so many things have yet again changed. So, how do you create a strategy in the face of so much change and uncertainty? And to your point, you have to get started.
(13:05):
So, you start with the things that are not going to change. You start with your values, you start with your strategy, those are the firm foundations from which you look at all of these technologies. And I'm a big believer in starting with those things, especially the more convoluted, confusing and uncertain the spaces, be really clear about what your strategy does and what it doesn't do because what strategy always is. And then ask the questions of how is this going to help our strategy or hurt it. I keep coming back to that and be very focused and directed on that because, if something is interesting but it's tangential, it's more tactical, then you can go either it's not important or it's a no-brainer, just do it. If it falls into those two categories, address it and then move on to the bigger ones that require your attention.
Natalie Pierce (13:54):
Yeah, [inaudible 00:13:55].
Charlene Li (13:55):
So, can we write better copy and improve our marketing with these tools? Yeah. Is that a strategic issue that's going to make or break what we do? No. Then just go ahead and do it. If it's something that is strategic, I know one organization said that customer experience is our number one competitive advantage so they're going to take a close look at how copywriting can be taken to the next level and, not just something that's a more tactical thing, but it is something strategic to them because it adds to that experience. So, they're going to look at that very closely, invest in it, experiment with it and very carefully move into that because something that could make or break their strategy.
Natalie Pierce (14:35):
I would love if we could stick on that point for a moment, just about this company that you're working with and their space is really customer experience. Can you give any examples of how that company or any other client you've worked with has really seen the benefits of embracing generative AI in the workplace and really aiming toward that strategy?
Charlene Li (15:07):
Sure. Again, something that's been in the works for quite a while is, when you call into a call center, now, they can see from the initial prompts how fast you're typing, your tone and inflection of your voice, how irate or urgent something is and they can cue up the scripts for the agent that identifies that this person's upset so you may want to ask these questions or-
Natalie Pierce (15:35):
Oh, wow.
Charlene Li (15:36):
... try to calm them down. So, that's already been out there for a while. Now what happens is, with generative AI, it'll not only create the scripts but they'll pull up, for example, and, again, this is all conjecture and still very much in the works, they recognize that person from their phone number and pull up their entire history and then write the scripts in real time for the agent to use because that what generative AI does, it writes actual scripts based on a set of data or information in the context. And if they can understand the emotions of that context and, in real time, cue the responses and generate more text for the agent as suggested, again, the agent can use their own judgment, but they're creating options and texts and pulling up customer service records or responses or ...
(16:31):
Again, especially for support calls, it's an entire database of hundreds of thousands of documents that will say, "Oh, yeah, here's a possible solution or ask these questions to narrow it down even more." So, you're not dependent on the knowledge of that individual customer service representative but they have the power of the entire organization and the history with that client, with that customer, to be drawn into real-time scripts and that will improve your customer experience, for example, in a significant way.
Natalie Pierce (17:05):
That's a perfect example. To be able to help pull up that history, get the personality, the prior exchanges, I think that is an excellent example of human machine teaming, this augmented collaboration. We can and should learn to combine people and technology in a way that really leverages the respective strengths of both. And so, to put that kind of information, as you said, for the customer service individual to do what they would like with it but to have that real time feedback to better set an organization for those successful communications that lead to better outcomes, lead to better productivity, profits, workplace environment, et cetera. That's such a great example.
Charlene Li (18:01):
Right. And there was recently a study done by nine leading researchers with a consultant from BCG where they followed the tasks that people were given and granted people who they were able to do more tasks and less time with better quality with generative AI, in particular, OpenAI and ChatGPT, than without it. But what was really interesting and this is what I'm seeing play out in a couple of other places too, the most impactful error were for the lowest performers, people who were not great at certain things, they became really good at it. So, the quality increased substantially and the speed of which they did things increased also.
(18:44):
And when I was speaking to one organization that had a lot of process and call centers and things, they go, "When we saw this, we saw, not an opportunity to save costs, but to increase the performance of every single person but especially of our entry level, lower performing people to make them top performers. And then give our top performers the ability to go do all the things that we've wanted to do on behalf of our customers but we didn't have the time, people and money." So, I think the real opportunity here isn't that we save costs and take it right down to the bottom line to increase profits but that, if you really want to grow, in the past, if you wanted to grow, you had to add new people and take on new costs.
(19:28):
And I think the real opportunity here is that we're going to have this really interesting window of time where we can reverse the growth equation, where you can actually grow substantially with the same people, without spending any more money, without any more time then you can grow.
Natalie Pierce (19:48):
I couldn't agree more. I think, so many people, they are apprehensive about the impact that generative AI can have on businesses, jobs, the economy and many see this revolution instead as a real opportunity to shape our future for the better. And even in your example of taking an entry level position, somebody who might not have as high of a skillset, and giving them this tool to make them be able to effectively do their job better. And historically, we really have seen that, while job destruction certainly is a real effect of automation, we've seen that historically but what we've also seen every single time is the efficiency and productivity gains from automation helping to create jobs.
(20:46):
And Charlene, you mentioned the ... I don't know, you probably did see this recent Wharton Business School study that came out, I forget the name of the professor but-
Charlene Li (20:58):
Ethan Mollick, mm-hmm.
Natalie Pierce (21:00):
Oh, okay. Yes, of course. But I was blown away to see this experiment with two groups of Boston consultants. If you're going to get to BCG, you're going to be top of your game to begin with, you even know who the professors. You want to share with the listeners that recent idea? And I have the stats in case you [inaudible 00:21:26].
Charlene Li (21:26):
Yeah, I think it was 21% increase in productivity.
Natalie Pierce (21:30):
25%.
Charlene Li (21:32):
25%, right, right.
Natalie Pierce (21:34):
Yes, yes.
Charlene Li (21:34):
And this is a very informal tool. They segmented a couple hundred consultants and gave them a fake case to work on, basically, I think it was [inaudible 00:21:44].
Natalie Pierce (21:43):
Yeah.
Charlene Li (21:44):
And they measured how quickly they were able to do things and how well. And they've had various rankings, as you can imagine, of skills and capabilities of different consultants and they found the lowest level consultants were able to increase by 25%. I think even the highest level consultants were able to increase it but less so, their increase was, I think, only 11%. So, across the board, you see it but the real biggest benefit are the people who are at lower levels of capabilities. So, again, that's the AI that we have today and this is just off the shelf, untuned to your specific company, again, just really generic, very early, relatively primitive tools compared to what we'll have in the future.
(22:34):
So, it's very encouraging and most productivity gains that we see from technology are usually in the single digits. And this, again, without any tuning, is giving us 25% productivity gains, we've never seen anything like this.
Natalie Pierce (22:48):
[inaudible 00:22:49]
Charlene Li (22:49):
And depending on the [inaudible 00:22:50] you're seeing sometimes, 50, 100, 2X, 5X gains in productivity. What do you do with that?
Natalie Pierce (22:59):
It's unbelievable what's happening and just wait until we see you start seeing the next generation of LLMs that are just right on the horizon. And you talked about your waking up every morning being like, "Okay, what did we find today? What did we learn today?" I'm just blown away by what we're seeing with these impressive multimodel capabilities, having generative AI that can now process and understand multiple types of data inputs.
(23:41):
And then, going back to the BCG consultant study done by the Wharton Business School professor, not only did they see that 25% increase in terms of the productivity of tasks being completed, but also 40% higher quality results than those consultants who were told that they couldn't use artificial intelligence to augment their efforts. So, when you talk about superpower, Charlene, without doubt, is a superpower.
Charlene Li (24:16):
Right. It's less the fear if that AI is going to replace humans, it's much more likely it's humans with AI replacing humans without AI.
Natalie Pierce (24:25):
Would not have said it better, that's exactly, exactly the case. I'm going to move us on a little bit. This whole notion of getting started, this is something you've been so passionate about for such a long time and you help companies with baby steps. But now, there's something about a sense of urgency like, "Okay, ready, go, now." So, even setting companies up to start now with your 90-day plan, don't worry about the fact that, historically, it's been a long-term strategic plan of years, we're going to have to be a little more flexible, maybe be willing to change.
(25:12):
So, I know this is a big question and something that you've dedicated, really, your career to. But in the context of generative AI, what are some first steps that a company can take to encourage the responsible use of generative AI as they start experimenting with it?
Charlene Li (25:32):
Right. And I think the first thing to overcome is that you should be doing something. And the biggest fear is, well, is it too early because I don't want to be one of those early adopters that falls on their face. And I go, "It's not too early. There are immediate gains that you can use and you should absolutely, at the minimum, knowing what you can and can't do with it and then strategically decide you're not going to do something because it's not the right time, the technology isn't there, whatever, but make it an active decision rather than a default one.
(26:06):
And at a minimum, every company needs to at least play defense. You have to have a policy, guidelines, a code of conduct of something that tells your people what they can and cannot do with generative AI because they have access to it, as we've talked about before, they are probably using it in their personal lives and they're wondering, "Well, can I use this at work or can't I?" One company said, "Well, we should take a survey and really understand is this something that our employees are interested in using?" I go, "You can dispense with the survey, I can guarantee you, they're using it. They're aware of it, they're probably using it already." They go, "Do you think?" I'm like, "Oh, yeah. Go check your logs to see how many times ChatGPT is being actually used and accessed." And I go, "I guarantee you, people are using it."
(26:59):
And so, have a point of view of what they can use it for, what's good and safe, what do they need permissions, what training you need to have. So, you have to at least have the defense in place to protect your employees, protect the company, protect your customer data, make sure that there's data governance in place. And I feel like, especially for boards, it's a fiduciary responsibility that you make sure your company has these things in place at a minimum and then you can go and decide what else you want to do.
Natalie Pierce (27:30):
Absolutely. And speaking of the board's responsibilities, and I'm really glad that you brought that up, but what we've certainly seen is that adoption is slowed and competitive. Edge is lost when executive leadership is not charged with and really held accountable to present the board calculated options to make smart decisions around things like generative AI adoption. And I think we both agree that that really is part of the board's duty of oversight, making the right decisions around strategic plans. And I'd like to ask you, I know you advise a number of boards, but how might we tie oversight responsibilities of the board of directors to the strategic plan of the company when it comes to generative AI?
Charlene Li (28:29):
Well, I think the board of directors, at a minimum, has a conversation with the executive team and CEO and say we want to understand how much of an opportunity or threat is generative AI to our strategy and come back. Even in this conversation, where does our strategy, where is it vulnerable, where are the opportunities? And even at the high level, if you know just a little bit of what generative AI can do, you don't need a lot and it doesn't take a lot of imagination to think, okay, here are the broad strokes of my strategy and here are my initial gut feelings of where it could be. And I will say, having worked with people just for a couple hours, just working through their actual strategic plan and looking where generative AI could help or not or hurt it, you have a pretty good idea after just an hour or two of just playing with it that you know what the big buckets are for exploration and you probably have an idea where it's not.
(29:31):
So, again, every company is different, every industry is different and, depending on how good your data set is, how advanced you are, you can start with an initial list of big buckets with which to explore.
Natalie Pierce (29:48):
That's great.
Charlene Li (29:48):
But you should know what those buckets are because you know your strategy really well. You don't need to know the technology really well, you need to know your strategy really well. And any company worth its weight, it should know its strategy really well.
Natalie Pierce (30:02):
Absolutely. And one thing I've heard you say that I absolutely love and I think it's true is that you have to be willing to change your relationship with risk. It can't just be the safe wait and see and, like you said, just getting started with those discussions now, if you haven't already, it's critically important. I think we're going to find a lot of companies whose failure to adopt change, it really will leave them marginalized and they're likely to become casualties rather than beneficiaries of these types of transformative changes that we seek when on generative AI.
Charlene Li (30:47):
It's interesting, the companies I'm finding who have an easier time, not that it's easy ever, but an easier time working with generative AI are the ones who have really gone through a true transformation of the workforce and of their leadership, their culture and their strategy because of things like digital transformation. They didn't just put in place a technology, they actually transformed their organization to be more agile, to be able to deal with uncertainty, to have a right size relationship with failure, not to have all the answers all the time and instead really seek out how to ask really good questions and, therefore, lead in a very different way.
(31:31):
Because generative AI is putting right in front of us the ability, we have to have the ability to be able to be flexible, to be agile, to be resilient in the face of these changes. And I think this whole thing of moving your leadership from having all the answers to asking great questions is especially important when our employees become super powered, become these super humans. They don't need you to tell them how to do their job anymore, they know how to do their job better than you ever did. So, what does that leave you as their leader? What is the guidance that you can provide them? And it becomes much more around shaping the direction of travel, creating the guidelines, making it so that we can use those governance, the nasty word governance, not to slow things down, but to speed things up. It's a very different way of approaching leading and managing.
Natalie Pierce (32:28):
So true. Approaches to organizational effectiveness and change, really, as you said, they become possible when you have cells of agile culture that are building up around successful generative AI implementation. And again, couldn't agree with you more that agility, now probably more than ever, is key. One thing that you've done a great job with is helping organizations understand why it's important to form an AI council. Can you speak with our listeners a little bit more about that?
Charlene Li (33:10):
Sure, yeah. The AI council is, again, usually cross-functional, more than just your IT department which is where AI traditionally has been, it's been treated like a technology. But because generative AI is hitting across your organization and it has strategic and business implications, you need that representation across your organization and especially from your business leaders, your line of business, your commercial representation, your operations people, HR, legal, for example, and you want to keep that council as small as possible but have the most important people on it. And I've seen every type and shape of AI council and the best ones are ones that are, again, derived from their strategy.
(33:58):
If your strategy is very focused on having the best products and you want your creative people, the people who understand the customers, very close to them, the product managers and the people who maybe ... And customer data because you're going to be tapping into that as well as IT. But an organization like the one I was talking about with customer experience is going to want the business heads, the people in operations who actually make it happen and the head of HR, for example. So, it's depending on where you see those big strategic buckets again, you can pull the people from those big areas, those key functions, into the core group and then line up people who will be adjacent to it that you want to consult or be informed with, be informing about decisions that may be coming down.
(34:47):
But the idea of the AI council is everyone who needs to be in the room to make a decision is there. That's the key thing because you want to be able to make decisions and move forward. So, having those guidelines, again, about the defensive guidelines of how far you can go, how fast you can go, a clear direction from your executive team and board about the big buckets to explore to support the strategy and then let this team go at it.
Natalie Pierce (35:12):
I think that's great advice. Have it be a small AI council supported by an extended team so that you can hold onto that necessary agility to pivot and move quickly. Charlene, as we know, some companies are continuing to take this wait and see approach. What are some of the risks that you see if companies fail to embrace generative AI quickly?
Charlene Li (35:44):
Well, the number one thing is that your employees will start losing faith in the organization especially if they feel hamstrung that they can see this thing, they can see it on their computer, they can see it on their phones, they're seeing their friends and colleagues at other companies doing these amazing things like, "You will not believe what I could do at work yesterday." And you're thinking, "Oh, man, I wish my company was worth it enough that I could be allowed to use that but they just keep shutting me down." But guess what happens to that employee? They're going to leave. They're going to go to another company that will support them, allow them to reach their maximum potential. And so, you'll be losing some of your best people.
(36:26):
You won't be able to recruit people. It means that, when you're hiring for somebody, you're not hiring the best people because you're not asking for people. If you're hiring a copywriter, can they write but can they also use these AI tools in the right way? You don't know how to evaluate that even because it's not a skill that you've asked for and developed and understand so you're not hiring the best people who are at the top of their game. It also means that a strategy perspective, with each month that you delay, you're just going to be that much slower in implementing any new opportunities or defensive opportunities and your competitors could be moving ahead and you're, again, you're allowing people who may be even more laggard than you to leapfrog you.
(37:10):
So, I'm not saying that you have to do everything but you need to have an opinion on when is the right time and you can absolutely decide that what we really need is a private LLM to be able to make this happen. And we could do that today but it would be hugely time-consuming, expensive, we just do not have the expertise so we're going to wait a few months for these things to become more integrated into our existing enterprise tools and, when those are available, we'll turn them on. In the meantime, we're going to do these stock gap measures but we're not going to go out and build our own private LLM. Do you see the difference in the terms of-
Natalie Pierce (37:47):
Absolutely, yeah.
Charlene Li (37:48):
... not waiting versus saying we're not going to do anything?
Natalie Pierce (37:52):
Exactly, exactly. If you put up your defenses, I think, employees, they just want to know what are the guardrails. How can I use my superpower responsibly? But if you do not allow employees to meet their potential which these tools do, I think you're absolutely right, Charlene, that you're going to see people go to the companies that will and they'll be viewed as more likely to keep that competitive edge into the future. But I think it's misguided to think that employers can just wait and build their own. And there certainly are wonderful enterprise solutions available that will allow you to keep your data private but, yeah, couldn't agree more this thinking that we can somehow control what employees who are so hungry for knowledge and to advance in their careers and improve their skills and knowing how easily available the generative AI tools are telling them no is not the right answer.
(39:07):
Charlene, for early adopters, where do you see generative AI heading and what do you expect to see in the future? So, where is generative AI heading?
Charlene Li (39:20):
Right. Again, we talked a little bit about the private LLMs but I think we've been hearing a lot about these co-pilots, these agents, so to speak, that can just act on your behalf and be your collaborator. So, initially, they'll be co-pilots, they'll be sitting there next to you ready and waiting for your commands to just be your assistant. You think about having an amazing assistant who's capable of doing so many things and they're right there. The real way that these things become powerful is that you granting them more and more access to your information.
(39:56):
So, to process my emails, to write really good emails, they need access to my emails in the past history of how I've written them and also my communications with that particular person or that particular team. And I have some now that are built into the email tools that I use and it is quite helpful to summarize a chain of 20 different emails and to be able to just write a quick summary and then do whatever I want so I can quickly write more complex emails in that manner. I don't use it for everyone but having that co-pilot there using my voice, super effective and helpful. But in the future, I think these co-pilots become a lot more autonomous. I'll give it an example.
(40:40):
When I book travel now, I don't use an agent anymore, I basically have to go and find my ticket and do everything by hand. And I'm like, "Wow, I remember the days when I could just call somebody and have them do all that for me." And I think the agents will be able to do that but they'll understand I prefer these airlines or, looking at my calendar, these are the times will be best times for me to fly or not and it'll understand all my preferences and do all of that for me in the background because it follows me around and knows what my preferences are for everything. From understanding my travel preferences to how I like certain meetings with certain people to when I need to have a break because I've just been in back-to-back meetings so it'll hold the calendar and not schedule anything at that time.
(41:27):
So, the biggest question here is how much autonomy do we want to give these tools? How comfortable are we sharing that information? How comfortable are we letting these things make decisions on our behalf? And that's a very different type of tools, I think, that we're still going to have to develop our comfort levels at doing that. In the same way, we had to move from using bank tellers to get our money to getting it out of a machine and trust it. Every single time. They do it correctly. So, it's going to take some more time for us to do that. The technology can do a lot of these things already but I don't think we're ready as humans to necessarily seed that kind of control yet.
Natalie Pierce (42:10):
I think that's right. Though some of the examples you gave, I think, yeah, I would take that in a heartbeat today but it's true. I think that we'll continue to see some of these emergent or unanticipated qualities or behaviors and I think you're right that, the more that we can provide our own information, our own communications, past writings, et cetera, we will see with this next generation of LLMs that really are right on that horizon, progress in areas like personalized reasoning, common sense and judgment and that will then call the question of how much of us are we willing to seed. I think those will be very important questions.
(43:09):
You talked a little bit about the fact that you have a book coming out. Can you tell us any more, just something to preview, what you're focused on with your co-author and when we might expect it to come out?
Charlene Li (43:25):
Sure, yes. I'm writing it with a dear friend, Katia Walsh, who's the chief digital officer at Harvard Business School and she has deep, deep experience being a chief data officer and AI officer at very large companies all around the world so invaluable experience that she's pouring into the book. And the tentative title right now is called Mega Transformation because both of us have been in the technology and business space for over 30 years now and we have never seen anything like this The speed, the potential for transformation and how, across the board, it is, again, from an organizational point view but also across society.
(44:05):
So, we think this is something hugely transformational and, as such, we have to approach it in a very different way of setting our strategy. So, the book is called Mega Transformation, how do we create a rigorous generative AI strategy in 90 days. And we lay out week by week all the things you have to do to address inside of that strategy and how to approach it in a rational way that's starting with your strategy and your values and who you are as a company. Rather than developing thousands of use cases of what AI can do and then narrowing it down, you're starting with something you already know that's true which are these foundational elements of your company.
Natalie Pierce (44:44):
So, now that everyone has heard this, everyone is going to want to get a copy of Mega Transformation. What is anticipated date of publication? Do you think perhaps this year, next year?
Charlene Li (44:58):
I wish this year but the way publishing works is not working in our paper. We're still writing it and so we're finishing up the manuscript this fall and we'll hope to publish it in January.
Natalie Pierce (45:08):
Wonderful, wonderful. So, Charlene, listen, we're just about out of time but I'd love to wrap up with a quick fire challenge if you're willing to participate.
Charlene Li (45:19):
Sure, go for it.
Natalie Pierce (45:21):
Great. Listen, I heard that you make your own kombucha, have you picked up any other new skills recently?
Charlene Li (45:31):
Well, I don't know if it's still recent but I really enjoy teaching my pet, my cat how to do tricks. And so, we're always learning new tricks and I am visiting my brother, he has a brand new kitten so, of course, what did I do last night? I'm teaching this cat how to sit and training. And people think, "Wait a minute, this is not a dog. This is a cat, right?" Yeah, I don't teach dogs, old dog new tricks, I do what seemingly is impossible. I go and teach cats how to do tricks and they do it, they will absolutely do tricks.
Natalie Pierce (46:04):
You got the kitten to sit?
Charlene Li (46:06):
Yeah, yeah. They do this natural thing when you take your hand and sweep up your, they say, love fingers, they will follow your finger motion up. And when they actually look up, they automatically will sit and then you give them lots of praise or treats every time they [inaudible 00:46:20] a command set and they learn how to sit.
Natalie Pierce (46:23):
Charlene Li, kitty whisperer. That's amazing, I really didn't know that that was possible. One more question. If you could go to dinner with an AI avatar, who would it be?
Charlene Li (46:38):
Well, if I could sit down with Satya Nadella who's the CEO of Microsoft. I have tremendous admiration for how humanly he leads the company with empathy and he just always, from the very beginning, has said it's all about empathy for each other, for our customers, for ourselves and using that. Having a leader that's so values and humanistically oriented is so rare, you just don't hear many CEOs talking like that. So, just to have dinner with him and just understand how he thinks, I've read pretty much everything he puts out there. So, I know he's not necessarily available but I would just love to be able to sit there and have conversation with Satya's avatar and just ask him tons of questions.
Natalie Pierce (47:28):
Yeah, he's truly a remarkable leader who really came up through the organization. And as we've had founders as guests on the podcast, his name has certainly come up on multiple occasions so I love that answer. So, we are out of time. Charlene, it's been so great talking to you. Your work is so important and you're truly an inspiration. Thanks for sharing your perspectives on generative AI and your vision on the future of work.
Charlene Li (48:08):
Thank you so much again for having me.
Natalie Pierce (48:09):
And thank you, our audience, for listening.
Voiceover (48:13):
You've just listened to the FutureWork Playbook. This podcast series is brought to you by Gunderson Dettmer, the world's number one law firm representing venture capital funds and high growth companies. Join Natalie Pierce on our next episode as she and her guests help prepare your organization for the future. Please subscribe to the FutureWork Playbook.
