Synopsis: In this Techstrong AI Leadership video interview, Alice Steinglass, executive vice president and general manager of platform at Salesforce, explains how generative artificial intelligence (AI) will fundamentally change the way organizations think about building and deploying applications.

Mike Vizard: Hello and welcome to the latest edition of the Techstrong AI Leadership Series. I’m your host, Mike Vizard, and today we’re with Alice Steinglass, who’s Executive Vice President and General Manager for platforms at Salesforce. And we’re talking about how AI is going to transform the way we think about building apps and actually build them and deploy them. So it’s going to be like an end-to-end transformation or so we all hope. Alice, welcome to the show.

Alice Steinglass: Thank you. Thanks, Mike. Happy to be here.

Mike Vizard: Exactly how do you perceive the development experience kind of evolving? We’ve had low-code and no-code for a while, and a lot of people would look at a generative AI interface and it basically is a box that you just write into and maybe some code magically appears. What is the relationship between low-code going to be and generative AI and how will it all manifest?

Alice Steinglass: That’s a big question. I think there are multiple different ways when we think about, okay, so first of all, before we get into that, I want to take us outside of the box. There is a box, there is a box, yes, I can have a conversation with it and I can go back and forth and it’s amazing, but I think AI is going to be a lot more than that. It’s about integrating the AI throughout my applications, inside my workflows, inside my webpages, inside my button presses. The Copilot is great when I don’t know, for all the unexpected questions, for all the unexpected, allowing the user at runtime to be able to ask whatever they want to ask. But there’s a lot of cases where we know that they’re going to do it and they’re going to do it a lot and they’re going to do it quickly.
So a great example of that would be service replies or sales emails. We know our sales reps are going to be sending emails and they’re going to be sending a lot of the same types of emails again and again and again. So when we know that’s going to happen, that’s a great place to embed AI right inside of the flow of work. And so all of those places that’s inside flows, inside of pro-code, inside of low-code, we’re going to see AI just throughout all of our workflows, throughout all of our tasks.
And when we do that, I think there are two ways that I think about AI changing what it means to be a low-code app developer. The first is it makes it easier, right? If I’ve got an AI helper, we already showed last fall, we’ve got in pilot right now, being able to use natural language to create flows. That makes it easier. I can get started, I can get started with natural language. I can still edit that flow, and we still need those low-code capabilities to be able to create, edit, and test the right flows for my organization. But AI can help me out.
I think the second thing that gets interesting is actually creating those AI experiences and embedding them in the right places in my organization. There are opportunities both for low-code and pro-code there, but low-code developers really often understand what those business requirements are. They’re close to the business so they can say, “Hey, what is that sales email?” They can partner with the line of business owner to say, “How do we help you with AI in this part of your organization?” And then what we’ve done at Salesforce is built low-code tools that are going to let you create those experiences and embed them in the flow of work. And we can talk more about those if you want, Mike.

Mike Vizard: I mean, it sounds like to me a lot of the toil that I prefer not to do as a developer will get increasingly automated by generative AI, whether it’s a low level script or something of that nature. So as the developer experience evolves, what will I have more time to do do you think?

Alice Steinglass: Oh, I totally agree with you. I remember, so my first job in the nineties, I was making an active X control that could be embedded in a website and I was spending zero time on the actual goal. The actual control that we wanted to create, I would say that was like 2% of the time. I was spending 90% plus of my time looking up libraries, trying to figure out why it couldn’t connect, trying to debug some old code from the past. That’s not the fun part. That’s not why we get into this. We get into this because we’re developers, we’re builders, we want to create, and then we spend all of our time building out the boilerplate, the test cases, trying to connect. And I think AI is going to be an incredible assistant there.
One of our top requests from our developers has been test case generation. And so at TDX this year, we’re adding the ability, I can right click on an Apex class and be like, “Hey, can you generate a test for me?” And I get a test to get started, right? That’s important, but not the fun part. The fun part is building the technology. And so what I expect is that our developers are going to be spending a lot more time thinking about what are they trying to accomplish? What are those workflows where I want to use AI? What are those things that my team needs to do, and how do I enable them as part of their workflows?
So the other big shift that we’re going to see is a lot more focus on the data, thinking about data, understanding data and being able to leverage the right data for the scenario. If I generate, back to generating emails, we can do a lot more than emails, but it’s a good example. So I can write an email today to my aunt, to anybody, but if I don’t have the data there, it’s really just a party trick because even if I told an assistant, punt it to an AI assistant, “Hey, can you send an email to my customer?”
They’d be like, “Which customer? Who are they? How do you know them? What are you trying to sell them? What’s the background there that makes this email interesting?” It’s all of that data. It’s understanding my data. It’s having the right metadata on top of that data so I can understand what that data is and how to use it. That’s the part that is going to make the AI effective. And I think developers are going to be spending more time thinking about those scenarios. What is the data? What data do I need to bring in to make this effective? And how do I use it as part of my AI?

Mike Vizard: Will that increase the pace at which we’re able to build applications? And I’m asking the question because building software is not a factory. You have to have some level of inspiration and insight, and that doesn’t always strike every minute you want it to. So how fast will application development get or is it just more we’re just not going to burn out developers as much? But I’m not quite clear where that lightning in the bottle is going to strike more often.

Alice Steinglass: That’s a great question. I think, well, it’s very clear that when you add AI, you can go faster. You can be more productive, we’re already seeing that. So no question, but that we can do more and we can do it more quickly. Now your second question is, well, what do we do? Where do those ideas come from? How do we know what we can create and what we want to do? I actually think there’s a lot more out there than people are able to accomplish today.
When I talk to companies, and I think it’s not all in the developer’s minds, it’s throughout the company, right? When you talk to business owners, when you talk to a line of business in sales, in service, in marketing, but also in other areas like HR or across the team, what we see is that there are so many ideas. People have so many things that they’re like, “Hey,” or if you ask them, “What part of your day is boring? What is the thing that you do in a routine way that you wish you could get help with?” Everybody’s got one of those. Everybody has one of those in their job. And I think what we’re going to enable with this is the ability for developers and AI to be able to help out with a lot more tasks than they ever could before so that we can do that kind of work across the board.
I’ll give you an example. So Unilever, right? Unilever uses the Salesforce platform, not just for sales and service, but they have more than 50 apps that are built on the platform and it spans everything. It’s HR queries to scheduling volunteer sessions. And this is the kind of thing that wasn’t first when you’re saying, “Hey, what are the ideas?” Might not be the top of mind thing to say, “Hey, managing our employees volunteer activities,” but they built an employee volunteer management app on the Salesforce platform, and it worked. They were able to make that more efficient. They were able to give their employees more opportunities to volunteer, and overall, their deployment costs decreased by 40% by being able to leverage the platform for all of these different apps within Unilever.

Mike Vizard: Do you think that somehow or other we are restraining ourselves from having these great thoughts because we know there’s so much toil involved, so we’re kind of matching our ideas to the level of toil, and if we reduce the toil, then maybe we are freeing ourselves up to think about things we might not otherwise even bother thinking about?

Alice Steinglass: That’s a great point, Mike. Absolutely. So one of the pieces of AI that we’re rolling out at TDX is a tool called Prop Builder, and it’s a low-code playground where I, as an admin or developer, can just play around and see what the AI is capable of, but do it with the CRM in my organization. So I want to summarize a bunch of feedback survey data. What happens if I just take that data, I throw it into the LLM and see what that feedback survey looks like? I can try it out, I can test it, I can add other data. And what we’re seeing in the pilot, when we started working with our customers on this, is they went way beyond what we expected they would do with it. Because like you said, they’re unlocked and they can play with it and they can play with it in real time and see it.
So Andrew Russo from BACA Systems is a great example of this. He’s an MVP, and he went in and he generated, oh, actually, I’m not sure if he’s an MVP, but he is a core part of our team and he is a great user of Salesforce. So Andrew went in and he started out with the sales emails, and he started out by customizing those sales emails for his sales team. But originally he was just customizing the bodies of the emails. But what he found was that, as soon as he started to work with them, they were like, “Hey, no, can we get a custom subject line because I’m reusing my subject line for customer after customer, and I’m getting a low open rate on my emails?” They’re like, “Sure, let’s try AI for that.” So he generated custom subject lines and they almost doubled the open rate for the emails.
And then they started adding more content into the emails. And then he came back to us and he said, “I don’t just want to use this for emails. I want to use this in other parts of my system and other parts of my web pages.” And so what we’re shipping at TDX right now is something called a flex template. And the flex template is designed exactly for this. You can specify exactly what inputs you want inside of your prompts, and you can use it anywhere. You can use it in an invocable action on a web page, you can use it inside of pro-code, you can use it inside of low-code, inside of flows and what are people going to do with that?
I have a feeling we’re going to find out. I know a bunch that have requested it. I have know a bunch of ideas of people who have asked for it and reasons why they want the flex template, but there’s so much potential. And what I’ve seen is when you give people tools and you let them get their hands on it and you let them play with it, they’re going to surprise us and they’re going to do some amazing things. So I’m excited to give these tools to developers. And like you said, I think the potential is boundless here.

Mike Vizard: As we rediscover the joys of play, will application development become more democratized? Because I still feel like there’s this notion of business users and developers and the business users has an idea and the developers try to code it. I know we’re starting to see more business users involved with low-code and no-code tools, but still more exception than common. But are we going to get to the point now where everybody just kind of naturally is involved?

Alice Steinglass: That’s a great question. When I talk to our customers, they don’t want everyone involved. One of the challenges they have with rolling out AI is the first version of it that a lot of companies tried was everybody sort of directly creating their own prompts and you have thousands of prompts at your organization. Everybody’s doing different things, and you can’t monitor it or see it or even use the right data. I was talking to one company where they had this giant spreadsheet of prompts and everybody was supposed to, all their sales reps were supposed to copy and paste from that spreadsheet into the Copilot and then figure out what they were doing and then add their own customer data. That’s a mess. That’s a mess. Nobody wants that.
So when we’re talking to people, what they really want is they want the ability to aggregate, but more at a department level, to be able to talk to a department and say, “Well, what are the emails that you need to send for your department? What are the service replies that you want to do for this department? And how do we, in a central way, customize it so it really works for you?” And then be able to roll that out so it’s in the flow of work, so that your service reps aren’t copying and pasting from a spreadsheet that they’re just clicking the button, the service reply is already there, and then I can monitor it centrally. I can see, did it work? Did it not work? Do I need to edit it or adjust it? So what does that look like in terms of who needs to be involved? I definitely need someone from that department who understands those workflows, who understands that service or sales rep to be a core part of building what AI looks like for that organization.
And I think I want them to partner with somebody who’s understanding the system and the technology to see how we roll it out at the organization. But that is blurring, like you said, I think low-code and no-code and the abilities to do that on the Salesforce platform has already helped us get closer to that business user so that they can play a bigger role in how it works. And I think that AI is going to accelerate that trend. So what we’re going to see is that we open up these capabilities closer and closer to the business user so that they can roll out the right capabilities for their organization.

Mike Vizard: For as long as I can remember, business leaders have been saying IT is too slow, can’t keep up with the pace of the business, et cetera, et cetera. I’m sure you’ve heard it before. Are we on the cusp of flipping that and will IT be faster than the business can absorb it because we’re able to drive a rate of change using AI that the workflows and the businesses aren’t designed to absorb?

Alice Steinglass: I think, yes. I think we are seeing an acceleration. And I also think an interesting side effect of this is that it’s putting IT in a leadership position. When I talk to IT leaders, you can either be in a position where just the work is just piling in, it’s coming in from the business, list of never-ending list of requirements. Or you can be in this leadership position where you are taking the lead and saying, “Hey, how do we transform our business with AI?” And AI, it’s giving them a starting point to be that voice at the table who’s helping lead the possibilities and the opportunities and identify those opportunities for the business owners to help them grow their businesses in ways that maybe the business owner didn’t even realize they could do. Maybe they weren’t asking for it because they didn’t know that AI could power that decision or give that recommendation or give that insight, but now I think IT is going to be in a position where they can see it and help the business owners transform their businesses.

Mike Vizard: And how do we get to that? And I’m going to ask this question with all due respect to our friends in the business community, but people fall into ruts. You ask them why we do something a certain way, and they all look at you perplexed and go, “We’ve always done it that way and nobody remembers why.” Are we going to, as part of the spirit of play, will people be willing psychologically to go and tear up some of the processes? And as you and I both know, a fair number of those processes have more exceptions than rules these days, but no one seems to know that or they forgot about it, or it’s written down in a book that hasn’t been updated in 10 years. So can we get to this kind of sense of open experimentation?

Alice Steinglass: So what I’ve seen is when companies are nervous about it, when they’re not sure what AI is going to do for their company, when they’re just not ready yet, the place you start is you start with that routine, boring task that you have to do every single day and that’s where you begin to gain the confidence. And actually, even before you get to generative AI, I think predictive AI is another place that you embed to build that confidence. So Salesforce does a trillion predictive predictions on AI every single week. We’re embedded in things like, “Hey, what’s the next best action for this sales rep?” Or “What should I do with the service supply?” All of those pieces, revenue, intelligence, all of those pieces are already built in and used by companies every single day.
And to your point, maybe at first they didn’t ask for it, but they’re relying on it today. They’re using it to predict, to identify white space, to figure out where they should go, to figure out how to approach a customer. So when you start out with the real problem that the business owner has and you say, “Hey, here’s something that you’re doing that either you could do better if you had more insights or could be less boring or less manual, or you could be more productive.” And then we embed the AI into that system and they can see that proof point, they can see that it works. That helps bring people along.
Gucci is a good example of this. They have a bunch of new service reps and they new service reps need to be able to speak with the Gucci voice, right? They need to be able to provide that high level of quality service. So they used Einstein AI to help those service reps know how to answer questions, to suggest answers. And at first you’d be like, “Hey, is that going to work or is that going to be helpful?” It did. It worked. It was helpful. They saw better results. And because of that, now they’re looking at saying, “Okay, this works and we can do it even more.” So I think that’s where we start. I think we start by showing that it works, embedding it in the use cases.

Mike Vizard: All right, folks, you heard it here. All that stuff that you don’t like to do, it’s called work. You should automate that part. All the stuff you like to do, it’s called fun. You should do more of that, and that’s how you kind of work out this whole complicated AI equation and we’ll see where we all land at the end of the day. Hey, Alice, thanks for being on the show.

Alice Steinglass: Thank you, Mike. It was great to be here.

Mike Vizard: And thank you all for watching the latest episode of the TechStrong AI video series. Find this episode and others on our website. We invite you to check them all out. Until then, we’ll see you next time.