David Shim and Rachana Rele: Read AI CEO and VP of Product Design for AI-native products at Adobe on amplifying creative work — not replacing it

0:01David Shim:We believe AI will be a creative partner. So removing the hidden work that most creatives have to go through, whether it's production of work at the far end or even during the front end of the process, being an ideation partner, if you will, but never doing things autonomously.

0:22Rachana Rele:Today we have two guests from two different companies who have one shared conviction. AI serves us best when it amplifies people, not replaces them. Today we're joined by Roshana Rayleigh, VP of Product Design for AI Native Products at Adobe, and David Shem. He's the co-founder and CEO of Read AI. Together, they're building very different products, but they share a vision of AI that removes the drudgery from creative work, makes room for the thinking that actually matters. In our conversation, we'll dig into some ideas that could genuinely change how you think about your work. David talks about a concept of storage of intelligence, the idea that your knowledge, meeting history, even your working style could be captured and made available as a kind of digital twin that keeps working even when you're not in the room.

1:10Rachana Rele:And Roshanna shares how Adobe's thinking about AI not as a one-shot creative output machine, but as a collaborative partner that helps teams break out of their own blind spots. We also push them on the harder questions, this job anxiety that's real right now in tech, and the surveillance concerns that come with recording your work life, and where they each personally draw the line. This is Design Better, where we explore creativity at the intersection of design and technology. I'm Aaron Walter. And I'm Eli Woolery. At Design Better, our primary mission is to produce work that helps people like you refine your craft, improve your collaboration skills, and get inspired by the creative process of others.

1:52Rachana Rele:If you enjoy what we do here, the best way to support us is to become a premium subscriber at designbetterpodcast.com slash subscribe.

2:08Rachana Rele:Roshna Rehle, David Shim, welcome to Design Better.

2:11David Shim:Thank you for having us.

2:13Rachana Rele:Excited to be here. You know, you're an interesting pairing. We've paired you together like a fine wine and a fine meal. Because both of you are working on a vision for what AI and the workplace looks like in different ways. Roshna, you're at Adobe, coming off of Adobe Max and probably trying to shake off some of the fatigue of the fun conference. David, you're at Reed, where you're co-founder and the CEO. Let's start with you, David. What's your vision of how AI is going to change the way that we work? I think where we're moving towards in the next year or two is this concept of storage of intelligence.

2:53Rachana Rele:And this is the idea that all the work that you do can be stored. And that storage of intelligence becomes incredibly valuable, not just for yourself, but from a multiplayer aspect. So when you think about the future of work, the ability to actually share your knowledge, to share your team's knowledge, and not do it proactively where you're kind of going and saying, hey, I'm going to forward you this report. I'm going to give you this insight. but rather it's there that you could start to, just like you do with Chachi, but you start to ask questions to go in and say, hey, what were the last 100 customer interviews?

3:24Rachana Rele:Give me an aggregation of what their feedback was. Now, can you bump that up against people who canceled the service? So now I know people who canceled the service and the reason why, and I can start to dive into understanding what's actually happening. And that's really rare to do today because it's hard to get all that information. If you've got teams in Tokyo, if you've got teams in London, Seattle, New York, you don't know who to reach out to. That data isn't readily available. And it's going to be more push than it is going to be pull. Like it is going to get pushed to you just like a TikTok story.

3:53Rachana Rele:Somehow gets pushed to you that you like cryptos on Sundays. And on Fridays, you like to talk about going to dinner. And it's just going to learn that and send you that information.

4:01David Shim:What I'm most excited about is AI playing a role of a mediator and a creative partner, especially in a team setting. Today's AI technologies are meant for a singular person, individual productivity. I would love a future where we go towards more around AI that makes us better teammates through being a creative partner with us, through mediating better practices inside a team collaborative environment. so that we don't sort of hit this local maxima, if you will, of our own capabilities, but rather tap into capabilities of people we work with.

4:42Rachana Rele:So, Rachana, at Adobe Max, some of the things that were featured, I think there's worry amongst the creative folks in our audience and all around that AI might replace their jobs or some portion of their jobs. But it looks like the focus for Adobe is largely how do you extend people's capabilities? How do you act like a partner? So maybe you could talk a little bit more about that.

5:01David Shim:Yeah, so if you see many of the Max keynotes and the demos that we did on stage and also on the side stages, it was really about using AI as an augmentative technology rather than a replacement technology. An example of that is using AI to auto-name your layers in Photoshop. That is completely, you know, augmentation as opposed to automation. So if you look at the theme of Max and theme of what we've released, it's really around removing the drudgery that people are most excited about. It's not about giving you this one shot creative output. That's not what our audience wants. And that's not what we are building.

5:39David Shim:We are building towards a future where we believe AI will be a creative partner. So removing the hidden work that most creatives have to go through, whether it's production of work at the far end or even during the front end of the process, being an ideation partner, if you will, but never doing things autonomously.

5:59Rachana Rele:David, there's a strong overlap in that idea of reducing the drudgery and the mundane in the way that Reid thinks about the workplace, too. You're gathering a lot of data. And I think you even have this idea of a digital twin in the workspace. How are you solving for that drudgery of work? I think the most basic thing is if you've ever been on a call or a meeting in person, the person who is assigned to take the notes is probably the most junior person in the room. And you're not going to contribute. You're not going to take any part in the meeting outside of just taking the notes and then distributing them.

6:32Rachana Rele:And what typically happens with that is like that person doesn't want to do the role. that person doesn't do a great job. Sometimes they miss certain facts, et cetera. That job is now eliminated when it comes to meeting notes with AI. And so with read, we join a call. We look at things like, are people nodding their head because that's really interesting? Or are they shaking their head violently because they disagree? So that kind of approach of measuring all those signals enables for better meeting notes. And no one's actually complained. No one has ever come back to us and said, you cost me my job.

7:01Rachana Rele:I can't do this anymore. It's really more like, oh, this is great. I don't have to worry about this specific task. And so that's been a great value return. Then it starts to go into things like, hey, following up. Normally, you would have someone follow up, a project manager or product manager, follow up with someone and say, hey, would you be able to give me an update on this project? What's the status on this? Now you can actually get those automatic updates where it's working in the background. It knows how the project has progressed. A great example that we've got is for salespeople, the biggest problem is getting them to update Salesforce or HubSpot on a regular basis.

7:33Rachana Rele:And so you don't understand what deals are coming in, what's going to close. And it's gotten so bad that, you know, chief revenue officers will go in and say, hey, you're not going to get paid your commission unless you update your Salesforce opportunity, your HubSpot opportunity on a regular cadence. Well, we've now gone in and said, hey, connect your emails, connect your message to us, connect your Slack, connect your teams. And with that, you're able to actually go in and see all that information and move things forward very quickly to go in and say, hey, AI is telling me that this opportunity should go from 25 % to 50 % because the person asked for a quote.

8:06Rachana Rele:It should go to 50 % to 75 % because someone actually signed it and said, let's move forward. And now it's moving to 100%. And so that's all automated now where we launched that about three months ago, and we've already had$100 million of deals go through that AI where AI initiated the chance for the change. And what we've learned is on average, it's about three to four days faster than if a human did it. But again, it's limiting the job that nobody wanted. Nobody wants to clean up their CRM system. No one wants to update their CRM system. I'm curious about this idea of a digital twin. And when you collect enough information about your work and the things that you're doing and how people could interact with that, even when you're absent, I see that as being like a huge unlock in some respects, especially if you're in an executive role.

8:55Rachana Rele:So much of what executive work is, is going from team to team to team to say, here's a vision, here's what we're doing. Are there ways where AI could help with that cross-pollination between teams and the queries that are kind of common types of questions people might have? Yeah, on our side, what we are looking at with the digital tool in it is, we've got over 20 integrations that it brings in all that information to understand who they are, how they respond to things, what they're interested in, what they're not interested in, as well as bringing in the information about the team. But it's also bringing in things like basic things like calendar.

9:30Rachana Rele:So if I email somebody and say like, hey, can we schedule this meeting? Internally, we have this digital twin called Ada. And with Ada, we can go in and say, hey, let's schedule a meeting across these three people. And it will just do that automatically. I use that for external meetings as well. And what Ada does is it actually goes in and schedules those meetings seamlessly because and has access to my calendar, understands what blocks of time I'm most available, that I'm most engaged. I will schedule that meeting. And what happens a lot of times is the end user will just reply back and say like, hey, this is great.

9:58Rachana Rele:Really appreciate it. Totally surprised you're replying back five minutes after I sent the email at 11.35 PM. And then I have to jump in and say like, hey, appreciate it. And just as a heads up, this is Ada. It's my digital twin. It's doing my scheduling. We've actually opened it up now a little bit more to when I send my investor letters to the board, they'll get the letter that talks about her performance, but I'll also go in and include ADA and they could actually ask questions and ADA will actually pull in the answers and I've got it set to auto respond. So we're transparent with our investors.

10:28Rachana Rele:So we're like, go ahead. Any question that you want to ask me, you can tap my knowledge page and it will immediately answer back. And so that's actually gotten to the point where it gives more trust to the investors to go and say, David's not hiding anything. The AI is actually pulling that information in. It's giving my answer immediately. And a lot of times it's giving a better answer than I can because I don't have context on what certain teams are working on, but they've given me access to their information. So that ability to fill in and be a more knowledgeable, not necessarily smarter, but more knowledgeable person can go in and say, I have access to this information.

10:58Rachana Rele:Let me answer this question. And where we see the immediate impact is going to be on things like you're out of office. If you're out of office for a week and there's a fire drill, normally you would get a page and you'd have to pick up the phone and then you get out of the mindset of being on vacation and the other person on the other side feels bad. Imagine if your digital twin can fill in for you. Hey, why did you decide to do this? When did you check this in? And I can say, oh, it was because of this bug, this happened. And now you've got full context that you can go solve that problem. Imagine you're on maternity or paternity leave where you don't want to be out of commission where all of a sudden when you come back, people have forgotten who you are and they've gone and they rely on other people.

11:34Rachana Rele:Now imagine if your digital twin steps in and says, hey, I would recommend we do this. This is my design preference. This is what I think the customer would want. And it continues that conversation for you, and it fills in while you're out. I think that's where that extension comes in, where it's about augmenting who you are, not necessarily replacing you. This concept is interesting, and it actually came up. We're doing some AI and design thinking workshops, and we're trying to develop one for managers. And with one of our clients, it came up that they wanted to experiment with something like this digital twin.

12:03Rachana Rele:And there is a tool out there called Delphi, which can go and funnel in all the content that you have online, essentially all your writing, all your talks, and then create a digital version of yourself. A friend of ours, James Buckhouse, has one of these. And it's pretty impressive what it can do. You can go and essentially call it up and ask it questions. I guess one open question I have about it is, how are people going to react to that if I'm an IC and I want to talk to my manager, but they're off on vacation? Am I going to be comfortable talking to this digital twin, or are there guardrails around it?

12:34Rachana Rele:just like with ChatGPT, just like with anything that does automatic scheduling, with meeting note takers, there's going to be a little bit of hesitation. But as they start to see, oh, I can ask this question, and my manager is not going to get annoyed because I'm asking a question about process that I've asked three other times before that I just forgot what to do. Manager loves that because, okay, I don't have to get annoyed that I'm getting asked this question because I want to help them, but they're still asking me the same question. The AI can fill in. The AI can actually move things forward.

13:01Rachana Rele:And it also takes things off the plate of the manager where they don't have to do that anymore. They can go in and say, hey, if it's a basic question about process, go in and ping my digital twin and potentially CC me. Or the digital twin can go in and say like, hey, this is a question that came up from Eli. He wanted a little bit more context on this. This is what I said. If you want to follow up, feel free to do that. So it becomes an extension of you. And I think people will get more comfortable in doing that. Just like with search. A lot of people will first go to their internal wiki, their internal Atlassian, Jira, Confluence, whatever that they have to search for that information rather than bother somebody.

13:34Rachana Rele:Because people are generally thoughtful and they will go in and say, let me try to find it on my own. But that's going to take a long time. It might take you an hour or two versus like, I can go in and bother my boss. I can get that answer very quickly, but I feel bad about that. What's the intermediary? It's the AI. It's the digital twin that can go in and step in and give me that answer. Roshna, how do you think about this stuff and agents? Is that something that is on Adobe's radar?

13:55David Shim:One place that we see for digital twins, not at a individual level, but again, at a team level to be useful, is when you are actually reviewing designs. from a standpoint of, you know, send this to a client. I constantly work with this client. This client has certain patterns. So before I send off my new designs to the same client, let me just go and see the patterns at this client level on the comments that they've given, the common themes that they touch upon usually so that I don't make the same mistakes before sending these designs off or the deck off to the client. So I think at that level, at a team level, it can be tremendously helpful as the force check, not as end all be all, but helps you get 50, 60 percent there with respect to sort of making sure the baseline feedback is in corporate and the client doesn't have to repeat themselves.

14:52David Shim:And then the second place we are seeing is also in cases of collaboration, for example. What if I have a specific style as an individual working for a company and a colleague of mine wants to call upon that style or learn from my style or learn from the codified workflows that I might have had generated through AI? So we see uses and use cases for that sort of work where, Aaron, as you, as my colleague, can say, hey, slash Rachna's agent on graphic design, particularly. So I do believe that there is goodness in improving productivity through this sort of general concept of twinning as opposed to specifically about digital twins.

15:39Rachana Rele:You know, there's one thing that's particularly interesting with what just was released at Max, and that's in Firefly that you can feed like a design system or design styles, like brand guidelines and so forth. And one of the big challenges in larger companies is you've got so many different people who are making so many different things for different platforms, different channels. And inevitably, people, they go astray from brand guidelines. And I've seen this firsthand. It's a mess. And then you have to go have uncomfortable conversations with people and say like, hey, you can't do that. Pull that right now and let's go fix that because you're representing our company the wrong way.

16:17Rachana Rele:Could you just talk to us a little bit about what Firefly is doing in that respect?

16:21David Shim:Yeah. So we released something called the custom models, which has been in the market for, I would say, over a year and a half. What it is, is you can train a model using Firefly on your specific designs, your specific style, your specific composition, and use that, release that within your enterprise for other people to sort of build upon and generate net new designs based on that specific model. So going forward, you can imagine a world where I actually invoke a model while I'm creating new designs from a colleague from, let's say, marketing studio team or on a specific campaign for that matter, that you could have multiple custom built models from that standpoint.

17:05David Shim:So that's how we are doing it for custom models on Firefly. And then the other thing that we released is very bespoke kind of services we are calling Foundry, Firefly Foundry. And what it is, is for large companies who want to have bespoke custom models built on their own IP. And so that's specifically for large brands who want models trained on their own IP.

17:29Rachana Rele:David, I've been using Read a lot for obvious things like note-taking. The other thing that I found that's really helpful with is, you know, Aaron and I have a lot of conversations, not just podcast conversations like this, but with sponsors. And over time, even if you have pretty good notes, you might lose track of where a conversation took place even. And so that's one of the interesting things is it can kind of help you kind of like organize your brain a bit and help you get context on things where you're like, I remember somebody said this thing about this thing, but I don't remember when or where that conversation took place.

17:58Rachana Rele:So I think that's kind of an interesting effect of having a lot of these conversations on read. 100%. And that full memory is the thing where it's like, you've got something that triggered something in your brain. And I've done this a couple of times where it's like, what's the thing that we talked about a couple of weeks ago that was about optimizing something and something. And then it would go in and say, hey, what about these four things? And it's like, oh, this is great. I know exactly what we're talking about. Where we're focusing in on is around proactively pushing some of those things out.

18:24Rachana Rele:So now if you do have a meeting and these are the five key questions that come up, we enable you to search for those answers, but we want to proactively push those to people. That's what we've really seen in this age of AI is search is kind of where we're at today, where you think about chat GPT, where you ask it to build something, where you ask it to find something, you narrow it down to go in and say like, hey, I like that result, but can you adjust it this way? I think what you're going to see is a very quick evolution into more of a push system where it's more TikTok meets Tinder, where it's going to go in and say, hey, I see what you're doing.

18:54Rachana Rele:You don't have to prompt anything. I'm going to deliver content to you that we think is relevant, either updates on different projects or actions that we think you should take. And you're going to go swipe right or swipe left to actually trigger that action. And then you're going to go in and maybe do a couple of edits, but then you're going to activate against it. And I think that type of motion is what we will see very shortly. And it's kind of the motion that consumers are already doing today. Like if you look at consumption of content, it is through things like TikTok and Instagram reels now.

19:23Rachana Rele:It is no longer through doing search, finding something, and then reading an article about it. It's more like, hey, I want the content served to me. I want to figure out what's going on. And it is going to be interesting on how that switches to work, but I think that's where we're going to go. What's the net of all of this? So there's tons of things that we can do to improve productivity. In fact, it's something we've been doing since the beginning of the industrial era. We want to try to be more productive, improve our capabilities. But when you think about what's possible in this near-term future that you're both describing, and then we push that out to maybe like five to 10 years, a farther term, which is a long time in an AI age.