Leadership Insights: Generative AI

Transcript

Mike Vizard: Hello, and welcome to the latest edition of the Digital CxO Leadership Insights series. I’m your host, Mike Vizard. Today we’re with May Habib who is the CEO for Writer. They are of a platform that is based on one of these generative AI platforms. And it’s going to make us all much more productive than we ever imagined. May, welcome the show.

May Habib: Oh, thanks so much for having me, Mike.

Mike Vizard: I think that, you know, clearly Microsoft and Google and everybody’s gotten the lion’s share of the noise around this whole thing. But you guys have been at this for a while now. And what’s your sense of how is this gonna play out? Are we gonna have multiple types of generative AI platforms for different types of use cases, and people shouldn’t just assume it’s going to be one to rule them all?

May Habib: Oh, absolutely. We’re going to have dozens and dozens of different generative AI platforms, I also think we are going to be dropping the generative, and really realizing that AI, writ large, is transforming all of our knowledge work processes at work. And it is going to go way beyond you know, fun images and writing emails for you really soon. We already today, Mike have customers who are using writers LLMs to do things like extract insights from market movements to communicate the results of insurance claim type of reviews to customers who, you know, are rewriting entire customer funnels by segment for true personalization. So, already, we have gone so much further with the teams who have really leaned in here. And they blend in because you know, the software is there. So like you said, we’ve been at this for years; we’re so excited that ML and NLP is the center of a global conversation now. But there are definitely, you know, there’s a lot to be desired of some of the commercial tools that have been put out by the Microsoft and the Googles. And this is an area where the enterprise adoption has been actually outpacing kind of consumer adoption and such. Yeah, as a result, the tools are, and the platforms are a lot more robust already on the enterprise side.

Mike Vizard: So you guys have already had an enterprise edition of your product. So how is that being used today?

May Habib: What we launched, and what we announced is that for the last few months, we have had customers using our own proprietary foundation model. So the vast majority of the other AI companies that folks will have heard about, depend on an API from Open AI. And Open AI is incredible. They really pushed the field forward – incredible researchers there. And that is an amazing API. But what we realized was in the enterprise, we have to be right all the time, it is four nines of requirements for the accuracy of what comes out of LLM, and just the architecture of large language models makes it so that that is like literally impossible for the vast majority of use cases. So we’ve had to put our own architecture in place for large language models, impossible to do if we were just fine tuning an open source model, or if we were using Open AI as models directly. And we have put a bunch of small brains over the big brain. And so for enterprises, and CIO, CTO CTOs who embrace the solution. You know, they, what they love about Writer is that they are getting both AI that they trust, that is just as great as if they had built it from scratch themselves in terms of data and privacy concerns. It is just as good and often better than the open AI models. And from an end user perspective, they’re getting output that is still much closer to what they would write themselves. It’s that four nines accuracy. And because of the quality of the output, we have customers who have taken humans completely out of the loop for workflows, like SEO for e-commerce for customer communication. And so they are literally using Writer to write thousands of things at once that don’t need to be reviewed – they are that confident of the output. So that’s all a result of us being enterprise from day one. And what we announced was the open sourcing of two of our models, so that customers could actually use those and run them on their own hardware if they want. And then the API for that third model that had you known before today, been available to enterprise customers via the app that they built on top of it, that we announced an API for them to be using it directly.

Mike Vizard: What does it take for the models to learn the nuances of each company’s processes and content? And nomenclature? How does does that happen instantly? Or does that take a little effort?

May Habib: Yeah, it’s a really good question. So there is data that the customer provides us. And then there is data that we crawl for them and adjust for them and include in the model. So you know, the name of the game is to make the model theirs as quickly as possible, and in a secure way where they trust that because of our architecture, and because of our privacy policies, and in our security stance, and the fact that we’re privacy perverse, their data isn’t saved down by Writer. We don’t have it, we don’t save data generations, their content is not owned by us, we don’t even have to reuse it to further train the model or fine tune the model for another customer. You know, very different than the chat GPT is of the world where even Amazon and Microsoft have asked employees to not put sensitive and customer information, their user information, their company information there. So for the most valuable use cases, you need to use your own data to get answers back on it. And Writer provides that to CIOs, CTOs, who really want to get the productivity boost of this latest generation of AI tools, but don’t want to make the compromises necessary on on the privacy part.

Mike Vizard: Right? You don’t want to get a cease and desist letter from somebody who’s going to claim that you copy their stuff in your marketing material, because you used an AI engine that crawled it off of the Internet somewhere.

May Habib: Oh, that is gonna happen a lot.

Mike Vizard: What is your sense of what the impact on productivity is going to be? And I’m asking this question, because for all the investments that we have made in technology for the last two decades, that productivity number in general has not really moved. And some might argue it’s actually declined. So are we on the cusp of something here with general AI, that we’re going to actually see some profound changes in productivity output? And what we see in those types of numbers?

May Habib: Oh, Mike, I mean, it’s nuts. Right? For 20 years, we’re making advances and people like us are still working 12 hour days, so I’m with you on that. I think the top few percent, in any profession are going to continue to pull those kinds of hours because the return on their time is is so highly leverageable to, you know, financial outcome. I do think, though, that the vast majority of companies that employ knowledge workers will need 20 to 30%, fewer of them. So that is a really profound change, I think, to the structure of our economies, developed economies that I don’t think really has gotten its fair share fair share of the limelight of this discussion. Partly because I think we won’t notice as much company cuts, 10% because of the macro, and then cuts another 10% of you know, smaller number, and then maybe another 5% and never replaces those jobs. You know, that’s kind of I think, what we’re going to see over the next few years; certainly people will be hiring, but it’ll be as companies grow. So we are, we are going to be a lot more productive. We already are a lot more productive. You know, for example, we’re a marketing team of four people. We get almost 3 million visits a month. And we use AI for everything. So and every day. Yeah.

Mike Vizard: Everybody talks about how there’s going to be job issues coming of AI. But the other side of this thing is, will we not lower the barrier for people to launch their own companies and their own businesses, and we’ll need some sort of expansion of the number of organizations that may be smaller and arguably more nimble, but we’re not going to have this notion of everybody working for these large organizations when maybe everybody can work for themselves?

May Habib: Oh, I love that vision of the future. Mike, I totally agree. So many of these tools just make it easier for all of us to spin up a web presence, an e-commerce presence, a customer support presence, so I’m all for it. And you know so much of the happiness research shows that you know, that kind of work brings a lot of joy even if it is very, very stressful to be a small business owner. I grew up in a small business, a small family business and I love the idea of the work that we do to make it easier for people to do that.

Mike Vizard: What do you think will be the impact on processes? And I asked this because so much of what we call digital today is really just somebody slapped the mobile front end on a paper based process and said this was profound. But if we have the kinds of capabilities that you’re talking about, do you think that either the processes themselves may somehow reengineer themselves on their own? Or will we sit down and kind of rethink these processes in a way that makes them much more efficient? Because the machine will wake up one morning and tell us, “Hey, you know, this is not the way to do this?”

May Habib: Yeah, it’s such a good question. You know, I love when people share links in Slack. I’ll tell you a story that illustrates this, and I hate it when people share links, and don’t tell me the takeaway, right? Like, we all were really busy. If you’re sharing a link with me, can you please tell me what the TLDR is. And someone did that a couple days ago. My marketing team is not going to be happy I shared this, but we have a beautiful Slack application that summarizes what went on in a channel, so you don’t have to read and scroll so much. And so I asked Writer to tell me what was in the article, and it did beautifully. And you know, I’m the CEO of a company that made the damn thing. And it took me like four or five weeks to like, start using it that way. And so you know, it’s gonna take time for all of us to adopt this, the DNA and sort of the muscle memory of like, “Oh, I have a really smart junior researcher with me at all times, what could I get done faster, by using that person more?” Like, imagine somebody sitting in front of you, as an intern, a smart intern from college, and you’re sort of under utilizing them all day. You kind of wouldn’t do that to a human, and we’re sort of doing that to AI all day long. It’s why we love the electricity analogy for this latest crop of AI innovation. You know, like, AI is going to be as big as the internet. Yes, but almost everybody who could afford it got the internet relatively quickly and in a few years – electricity took two decades. And this couldn’t be more like that. I don’t think it’ll take decades. But it might, you know? This is really about us upskilling humans and their processes and how they approach work and what the calendar looks like. And it’s just going to take, I think, a little bit of time, for us to, ourselves, learn what the best practices are and then teach them to others.

Mike Vizard: Do you think we have to rethink the way we engage with machines? And I’m asking this question, because if I look at the history of IT for the last four decades, it was about creating all these abstractions, so we could talk to a machine and get it to do something for us. Seems like now the machine can actually talk to us in the language we understand, and maybe that will change that relationship.

May Habib: Yeah, I love that. I do love that. Yeah. Anything that speaks to us in human language, it’s going to automatically get human attributes slapped onto it. One of our own brand guidelines is we never say, “The AI.” You know, like, it’s kind of like a holistic thing or like a being – the AI. And we also don’t talk about AI as like, you know, a robot. The robot emoji is banned at Writer; you won’t see that in any of our marketing any of our communication. And both those things come back to my belief which is that AI is human, right? Like what is the foundation of large language model? It is all the shit we’ve ever written before. It’s all the things you’ve ever felt before and thought before. It is us just kind of regurgitated back to us faster, the smartest of us. And I do think that being able to engage with our collective consciousness, in a way, changes how we engage with AI and technology. I think it really humanizes it because it is human. It’s us. The logic is us.

Mike Vizard: Do you think people understand the pace at which this innovation may be occurring? Because we humans get smarter perhaps at a rather glacial pace and then machines order at a much faster pace and people are talking about multimodal models and all kinds of fun stuff coming down the pike. But do you think we have access to it?

May Habib: Absolutely we are. Things that people thought would happen in years are happening in months. We have some very exciting multimodal announcements to share very soon Mike. I hope we’re talking shortly. Especially in the next 18 months there is a commercialization overhang; like what is in the research right now that has not been productized is mind blowing. There simply aren’t enough humans that can, you know, can take the research and turn it into working code and working prototypes for us to go as fast as we want to go. Anyone listening who wants to join a generative AI startup – I am May at writer.com. We need you because there is so much to do, and to do responsibly and to do in a way that brings joy to people’s lives and workdays.

Mike Vizard: Alright. There are lots of executives out there that are looking at this with equal measures of probably fear and delight. And what’s your best advice to these folks and and the leaders of these businesses about what should they be focused on now to kind of be prepared to hook up to what may become a firehose?

May Habib: Really good question. This is going to be one of those transformations where the productivity gains are going to outweigh the short term security concerns. And I say this as a privacy for our security first company. And the fastest way for CTOs, CIOs c-suite to understand how this is going to change their businesses is to interact with the hands on keyboard people who have become the fast adopters within their companies. And then explore solutions that really result in business process change at the team level.  ChatGPT is a fantastic individual productivity tool. What Writer does is take very similar technology, adapted to your company’s needs, use cases, own data set in a secure way. And, all of the end users, empower all of them with AI augmentation. And so you know, we are really here at the center of a brand that executives can trust and end users can love. And this is the first generation of AI innovation that has resulted in so much like just tactical end user product. And that’s a really fun place to be as an executive because you get to make a productivity gain kind of decision that gets people very excited. And you know, we’re not always in positions to kind of make decisions like that. So we’ve got a lot of heroes within the writer, customer portfolio, people who have won themselves a lot of love by providing such productivity and augmentation to folks.

Mike Vizard: Alright folks, well, you heard it here. We have nothing to fear but fear itself. Thanks for being on the show.

May Habib: Thanks for having me.

Mike Vizard: All right. And thank you all for watching the latest edition of the Digital CxO Leadership Insights series. You can find this episode and others on the digitalcxo.com site. We invite you to check them all out and once again, thanks for watching.