Name: MPW Summit 2023: View From Silicon Valley
Uploaded: 2023-10-11
Duration: 14 min 59 s
Description: Asha Sharma, Chief Operating Officer, Instacart Clara Shih, Chief Executive Officer, Salesforce AI Moderator: Ruth Umoh, FORTUNE

Asha Sharma, Chief Operating Officer, Instacart Clara Shih, Chief Executive Officer, Salesforce AI Moderator: Ruth Umoh, FORTUNE 

MPW Summit 2023: View From Silicon Valley

Conferences

Hello, everyone. Welcome back. We can also please settle down so we can get this show on the road. Thank you so much. Welcome back. All right, we want to talk now about the state of the tech industry which is still in the effects of several shifts all at once from the impact of A I to a slowdown in fundraising and the growth of tech hubs outside of Silicon Valley. How these shifts in tech affect the broader business world? And what should corporate leaders know about implementing A I at scale safely and responsibly? Joining me on the stage now are two tech industry leaders with a view from the inside. ASHA Sharma is the co OO of Instacart and Clara Shai is CEO of Salesforce A I. Please welcome them to the MP W stage. Welcome. Welcome ladies. We're so delighted to have you. Thank you so much for joining us at NPW. Thank you for having us. Quick reminder for the audience. I will be opening up the floor for a question. So do keep it in mind as we kick off this conversation. Uh ASHA I'd like to start with you, machine learning. Has long been at the center of Instacart business. But the advancements in generative A I over the years uh have obviously turbocharged and unlocked the potential for new and innovative use cases. So how are you leveraging genitive A I across your marketplace? From customers to retailers to your more than 6000 shoppers who pack and deliver orders across the US and Canada. Yes. So to your point, we call it the invisible ingredient in Instacart. So we have a catalog of 1.4 billion items. It updates 6000 times every single second. And we are constantly thinking about how we connect the millions of customers to the hundreds of thousands of shoppers and the tens of thousands of retailers with generative A I I think the big innovation is that it's natural language. So all of a sudden we can democratize how people express themselves. So for consumers, the number one thing we're thinking about is how do we help the meal plan? How do we help them find what's for dinner? If you think about it, a grocery store has 30,000 items. It's really hard to spear fish by shopping for an avocado or cardamom instead of actually shopping for the meal. And that's what generative A I is going to do. At the same time, we have 5500. And so we are thinking about how we create the perfect ad for the perfect moment, an explosion of creative, that generative A I can help us with. So that's how we're thinking about utilizing it. Well, that's certainly useful for this audience because women never know what they want to eat so we can use generative A I in that regard, Claire, I'd like to turn to you. You were tapped with sales forces, new head of A I in the spring of this year in the last five or six months or so. What core areas have you been focused on to accelerate A I solutions across the company? And certainly among your customers. Well, it's really amazing because A I is not new, it's everywhere around us. Every time we use Instacart or Netflix or Amazon, any of sales forces, you're using salesforce, your salespeople are using A I over a trillion A I inferences every week now across every salesforce cloud. So what's new in the last six months is to Asha's point expanding upon traditional classical A I and now making generative A I safe, secure and trusted for enterprise use cases and sales force really thinking about this in four different ways, right? The 1st, 1st and foremost is data, right? Because A I is only as good as your data, you need data to train the A I models, you need A I to fine tune models and then of course to ground and give context into the prompts so that the models don't hallucinate. So for us, it really starts with our data cloud and so many companies B to BB to CS MB, all the way to enterprise, have their customer data already in salesforce. And so that's really where we start, I think from there, it's phase two and what we've delivered out into the market. And we have amazing four leading customers like Instacart and Gucci and William Sonoma using our generative A I now is bringing these use cases into the flow of work for sales teams, for customer service teams for marketers to really address operational bottlenecks and issues and kind of undesirable tasks that have slowed teams down. For example, if you're a customer service rep, customer calls in expecting a quick answer, instead of having to look through pages and pages of knowledge articles to read the solution and then let the customer know, generative A I allows us to automate that right. And that's just one example, salesperson walking into a customer meeting, instant account summary of everything that's happening at that account, kind of it's completely up to date. So that's what we have out in the market. And then what we announced last month at Dream Force is creating a platform around this so that our customers, they want to be able to customize those prompts. You want to be able to create your own prompts and build your own applications with no code. And so allowing customers to do that, we announced it's called Einstein Co Pilot and co pilot studio. And last but not least, it's our research team continuing to innovate across all areas of the stack from vector databases to the tooling training, infrastructure, ethics, toxicity, filters citations and more. We'll touch on epic shortly. I want to turn to you. How are you using A I in fulfillment logistics and to solve for or avoid inventory challenges? Yeah. So in our business, we often have to know what's on the shelf more than our retailers know. And so like I mentioned, we have a huge catalog that catalog updates, you know, multiple times every single second. And so when a shopper, one of our 600,000 shoppers are going into the store, we update our catalog real time when they're taking items off the shelf that feeds all the way through to our advertisers. So they're not advertising on products that aren't even available to customers. We use this term, we use A I to match shoppers with the right customers. We use A I to match them to the right stores to actually navigate through the stores, you know, in a business like ours, every second matters, every basis point matters and every moment of quality matters. So we do this from, you know, start to finish and the the customer delivery. Yeah. And the impact on the consumer is what? Talk to me about that consumer experience. Yeah, I mean, so when I open the app, what I want to be able to promise our customers is that you have the best selection, you have the best quality and it's going to be delivered when you want it A I is the way that we do that at scale for 250 million orders every single year. Question from the audience. If you have a question, please raise your hand and Mike handler will come to you any questions at this point? OK. All right. Let's keep going. If that changes, I will circle back around both of you view A I as a lever for scale and productivity to that end. What is your approach to using it internally? To simplify day to day workflows and to expand its adoption in the company outside of your typical tech savvy functions, whether it's hr legal marketing, etcetera. Yeah. Well, we have our CMO or CH I run in the room so they'd be great at answering this. But I mean, we look at, we look at A I as some that can augment the workforce offload tasks, but I also think it's a massive accelerant. So yes, generative A I is automating a lot of the code that is written, but it's also the first knowledge base we've ever had. We hired more than half of our employees during COVID and we're all in different locations. And so for the first time, we can build a knowledge base, we call it Ava named after avocado where anybody can query Ava and actually learn about the business learn what's happened in the past and can actually grow together. So that's something that I'm really excited about. We also are using it with customer service. So it's helped us decrease escalation rates, improve contact rates. That is a really important part of our operating leverage that we get out of that. Our legal team, we have, you know, 1200 retailers that we partner with thousands of pages of agreements. And so we are able to actually look at all the bespoke nature of those agreements, understand them and then build off of those, all the generative A I. So we're trying to use it and really explore and power the workforce with it and same question for you. Yeah. Well, we're, we're of course in the business of building A I for our customers, but we refer to our own company as Customer zero and we're always, you know, eating our own dog food or drinking our own champagne. And so we, every product we've released in the market, we test on our own teams first. And so some examples, all of our developers are now using developer GPT Apex GP T to do automated code generation, test case and comment generation. Our sales people are using sales GP T our own customer support team using service GP T both as an assist for our customer support engineers as well as there's a lot of common questions that customers just want to go directly to our website and be able to self service and ask. And so that's another great use case for A I. Yeah, going back to Instacart. Um you believe that the future of grocery shopping in commerce really in general isn't online or in store? It is both, which is great news for me because I love perusing through Trader Joe's Isles. How are you using the online experience? So then augment or compliment in person shopping? Yeah, it's a great question. I can also talk about why generative A I is important there. I think for decades, we have, since the advent of e commerce, we have taught people to express themselves in keywords, which is a very bizarre and unnatural thing to do. Again, if you're thinking about the 15 ingredients that go into a meal is so odd that you can't just ask what is for dinner. And so I think that's a big opportunity when you think about going into the store, it is a very generic experience. You're walking through the aisles, you're going in a, you know, a non linear pattern to get through it and you're spending 45 minutes in the store. So we've invested in a company called Caper Cards and it is in our opinion, the best computer vision technology to help you when you are shopping. And so now you can connect your Instacart account. You can connect your loyalty for the first time. Retailers can know who you are. When you get into the store, advertisers can know who you are. When you get into the store, they don't have to just wait until checkout. We have a screen in front of you. You can check out as you go and we know exactly what's going into the cart and you can pay and we can develop preferences. We can suggest things based on your past purchases. So I think that those are the types of experiences that are better because you're online or better because you're in store and only manically unlocks your technology. Absolutely anything to add to you were not in your head. So you had something to, I mean, just to build on that. I think that the first big UX revolution was when we went from very basic assembly language to higher, higher order, more abstract code. Then the big revolution. I think a lot of reason why companies use platforms like sales force is going from code to clicks, right? Being able to democratize who can build a business out. Now, we're seeing yet another UX revolution come from A I which is to go from code and click to conversation. And you know, exactly as you're describing whether it's consumer who's doing it that way or all the employees and partners that you have in your organization. And I think too about, you know, we've got Lydian and my colleague who leads Slack in the original conversational interface. It completely changes how accessible intelligence and creativity can be to every worker. Yeah, I want to talk about some of the concerns around A I because it obviously offers up a mixed bag of risks and rewards. Uh It can certainly help to enrich and enhance the customer and employee experience, but it can also fabricate information or promote bias or discriminatory information. Uh which is why data privacy and security is a major concern. How do you leverage A I at scale while ensuring ethical and responsible use within your respective organizations and perhaps we can kick off with you this time? Sure. So I mean, right now, this is our 1997 moment for A I, right. We know that it's going to be big. We know that everything is going to change every job, every department. We know the customer experience is going to change for the good. But we also know that there's going to be new risks that will be introduced some of those risks we've identified, we're having really healthy discussion and dialogue about as we have the last couple of days, some of them, they're unknown, unknowns and we have to be very agile and vigilant at looking at those. And so, I mean, if I look at salesforce 25 years ago, the company convinced organizations to put their most treasured data, their customer data in the cloud. That was, that was a controversial concept at the time. And in order for the company to do that. We all had to invent new ways of securing and encrypting data and do it in a trusted way. Right. That's why trust is the company's number one value. 10 years ago, we had to do the same thing with predictive. There were new risks that came with predictive A I, including some of the same risks that we are talking about today, like bias and unintended training data sets. Now we have a new set of risks and it's on a completely new scale because of how accessible and how broad the applicability that the A I is. And so we have to come with a trust first mindset which includes both data security and data privacy, not just across companies, but even within an organization, different employees and departments have access to different data. And so two people asking the same prompt can and should get a different response. And so making sure that access control is a part of it. The other side of it is is on the ethics side, right? Making sure that these large language models, they are trained on the corpus of content. On the that unfortunately, there is bias, there is hate speech, there is a lot of not trustworthy sources out there. And so we have to compensate for that with really good toxicity filters, citations. Again, going back to the data is so important to ground all the prompts in trusted reliable, truthful data to ensure that the outputs are reliable and accurate. And then I want to with you, ASHA very quickly A I can deliver real value to companies that are early adopters. What advice do you have for leaders who have yet to fully embrace the technology in a meaningful way within their organization? The last 30 seconds. Yeah, I mean, a couple of things, one, I think it needs to be leader led. This is the largest technology transformation since the shift to mobile, potentially since 1997. And I think it needs to start from the top and I think every single person needs to be experimenting and exploring with A I in order to get the full power of it. Third, don't just deploy A I for A I's sake, deploy to solve real world problems. That is what we're in the business of doing as technologists, as business leaders, as operators, use it to solve the most pressing customer problems and we'll succeed together. Absolutely. Very well said. Thank you both so much.