I guest-hosted Sharp China with Bill Bishop; the episode, which covered China and Taiwan in the light of last week’s meeting between Speaker Kevin McCarthy and President Tsai Ing-wen and China’s military exercises in response, as well as China’s latest AI regulations, will be published later today. You can add Sharp China to your podcast player using the links at the bottom of this email.
Scott Belsky is the Chief Strategy Office and Executive Vice President of Design & Emerging Products at Adobe; prior to that he was Adobe’s first Chief Product Officer. This is Belsky’s second stint at Adobe: he was the founder of Behance, which was acquired by Adobe in 2012, and managed mobile and services for Creative Cloud.
I had a chance to talk to Belsky about not only his career path, but also about a momentous few years for Adobe: the company launched Photoshop on the Web, acquired Figma, and just a few weeks ago at Adobe Summit announced Adobe Firefly, its own AI image generation product. We covered all of this in this interview, along with discussions about Adobe’s business model shift, working with instead of against the Internet, and what he would tell future designers entering a world of AI.
To listen to this interview as a podcast, click the link at the top of this email to add Stratechery to your podcast player.
On to the interview:
An Interview with Adobe Chief Strategy Officer Scott Belsky
This interview is lightly edited for clarity.
The Shift to Creative Cloud | Background and Behance | Adobe and the Web | Figma | Firefly and Generative AI | Content Authenticity | Personalized AI | AI Personality | AI Risk and Opportunity
The Shift to Creative Cloud
Scott Belsky, welcome to Stratechery.
SB: Thanks for having me. Hey, Ben.
I have to admit, I’m actually kind of surprised, I had it in my head that you’ve been on Stratechery before, but assuming the search on my site is working correctly, my actual only mention of you was when I cited an interview you did with Nilay Patel a couple years ago. However, we have talked on and off over the last few years, so I’m excited to continue the conversation publicly.
SB: Thanks. You’ve covered some of the stuff we’ve shared at MAX before, I think, so you’ve covered some of our workings.
Well, it’s funny, I haven’t written about Adobe a ton. I wrote about Adobe back in 2013, my first year, and I wrote about the shift to Creative Cloud and how that was a very natural and appropriate shift, which at the time — this happens with every shift — people were very upset about it because they used to buy a product once, and I’m sure you still get people complaining about that, and then they would have it forever, but that this shift to subscription pricing was a better match for software generally. It’s kind of funny today, now SaaS is just like, “Of course that’s the way things should be done,” but it’s weird to look back at those transitions when they were still in flux and people were arguing about them. It’s like, “Look, no software is not a one-time thing. You want to maintain properly, it’s providing ongoing value. It makes sense to pay for it on an ongoing basis.” You’re obviously at Adobe, so you are compelled to support the model, but I’m curious, you were at Adobe when they acquired you, I guess around that time, what’s your view of that shift in general? Is it completely non-controversial at this point or do a lot of people still complain, “Oh, Adobe just wants to get more money”? Is it actually better for the products?
SB: Well, Adobe acquired Behance in late 2012 and it was right around the time of the transition, which I think happened in 2011 officially, and then the question was, “How do we bring Creative Cloud to the cloud”? The subscription was called Creative Cloud — how do we make sure we’re delivering services, value to our customers and what’s our roadmap to get there? The new customers that joined the franchise today, I’ve certainly never heard a new customer complain about our business model. They’re getting fonts lit up across all their products. Any customer who’s cloud-native expects that their styles or their colors or their fonts in Photoshop show up in Illustrator and that they can share libraries with other colleagues they’re working with. These are just things that people expect as table stakes and were enabled by the services that we infuse throughout the products. Also, by the way, as a company, we’re incentivized to bring services to the fingertips of our customers when we’re in this subscription-based business model, as well. I think it was one of those really healthy things, not only for the business, but also just to help us build for the cloud era, it kind of had to happen.
I’m already going off-base of my script here, but Adobe and Microsoft share a lot of similarities in a few different respects. One of these is that, and I think this is one of the reasons why the shift didn’t feel great to a lot of folks, is because it was actually similar to the Microsoft shift, where people don’t appreciate that Microsoft actually pioneered the subscription model and people didn’t really fully realize it because they were doing it with desktop applications. You had the whole Enterprise Agreement concept, that was the Steve Ballmer special, his biggest imprint on the company, you were paying on a subscription basis for applications and it was only down the road that those applications went to the cloud. That’s basically, I think to your point, the same path that Adobe had to travel. It started out as “It’s still the same desktop applications”, and while people just felt better about paying subscriptions for a cloud application, in your case they asked “Why am I paying a subscription for a desktop application?”. But to your point, it did make it easier to add on these other pieces once you had the business model in place.
SB: I think that’s right. Listen, at the time, we certainly weren’t thinking about bringing Photoshop to the web in 2013, but what we were thinking about was, wow, we can now ship updates every week as opposed to orient the company around an upgrade cycle every eighteen months. Then there’s also the matter of the fact that it’s just much more affordable. We still have a $10/month entry point for Photoshop that really opens the top of the funnel for us, whereas before, as you know, the price tag was a lot higher and it was all upfront.
Unless it fell off a truck!
SB: Yeah. There’s pluses and minuses to a transition like that, but at the time, we certainly knew the cloud was coming, we knew these services were going to be paramount. That business model change allowed us to then start building for iPad and start building for web and all these libraries and font services and everything else that I think has, in aggregate, added a lot of value to customers.
Background and Behance
Well, I didn’t mean to dive right into Adobe right away, because I actually wanted to dive into you first. You mentioned Behance, when did you start that? Give me the full Scott Belsky history.
SB: (laughing) The short version.
What was your background? How did you get into the design space in general? You sold Behance to Adobe, left for a few years and came back, it’s actually a very interesting career path.
SB: My interest has always been in both business and design and that was the theme, even in college. I took courses in economics in the undergraduate business program and the Design and Environmental Analysis major at Cornell, so I’ve always been spanning these two worlds. I worked on Wall Street after college and somehow, again, managed to find myself in a role in the executive office at Goldman Sachs, of all places, doing all kinds of odd jobs around the firm and guess what? Using Adobe Illustrator in the process of doing so. So I’ve always just had that natural affinity.
Behance was started in 2006 and the idea was to help organize the creative world at work. Everyone had these portfolios in all these places, but all the great creative campaigns and all of the awards went to agencies and creatives never got credit. Instead credit went to the people that they did work for, or sometimes it was the people that did work for the people that work for the brand. The idea was to bring attribution and some organization to the creative world.
Behance was bootstrapped for five years, venture-backed for two years, Union Square Ventures led our A round, along with a bunch of angels like Jeff Bezos and a bunch of other folks who were in that A round, but we didn’t stay as a private company for long. Within a year or so, we went through the acquisition process. Why? Because Adobe had just transitioned from software to service, I think the company recognized they needed to know who their customers were and what tools they were using and what they were creating and build a relationship.
So I came in through the acquisition in December 2012 and then quickly took on mobile and also services for Creative Cloud, so expanded the mandate in my three years there. But then I realized, “It’s been 10 years that I’ve been here” and made an interesting career mistake, which is when everyone tells you to do something, you start to think you should do it.
You became an investor, what everyone wants to do, right?
SB: Yeah. My friends were saying, “Listen, you’re a product guy and those make the best investors. You were a seed investor in a bunch of great companies, you should go and become a full-time VC.”
And you were at Benchmark.
SB: I was, yeah.
That was the hottest VC job, particularly at that point in time.
SB: I had a great opportunity to join an amazing team of partners and jumped in. I will tell you, Ben, probably three weeks in, I knew this was not my jam, but it took six months or so to find the right graceful way to move on to a venture partner role, which allowed me to open up my aperture again, knowing that I wanted to be a builder.
What’s different about that? Because I bet there are a lot of readers of Stratechery that, because obviously I’ve thought about that, as you can imagine, I’ve gotten pitched on the concept, so I’m curious about your perspective, why was it not the right fit for you?
SB: I think that venture capital is ultimately a transactional discipline. Even at a great firm with smart partners that you’re working with, you are chasing deals, you need to make sure that you close whatever has heat, and at the end of the day, you’re not building. You’re not building a team, you’re not building products, you’re not thinking about design problems. In fact, if you spend too much time with a team thinking through a product or design problem, you’re not doing your job, which is to go out and find more deals. As an angel investor, I’m brought in for product and design type stuff, but it’s just a whole different ballgame.
In that case they kind of want you because you’re bringing this unique perspective and you’re working for yourself because it’s your money, whereas at a VC firm, people forget they’re working for LPs, that’s who their bosses actually are, and the LPs don’t give a flying F about product or anything, they care about their return.
SB: Not at all! If you have a deal or two you want to do a year and you have a super high bar, you’re spending tons of time hunting those, and as soon as you realize, “This is not for me,” you could argue every second after that is wasted, which was just not in my own DNA, in terms of how I liked to use my energy and how I liked to build relationships. It just wasn’t my thing. I found myself listening to Johnny Cash music suddenly and I felt like I had hung up my spurs way too early in my life, and then I was like, “Wait a second-”
I like the spurs concept. I’ve always had the view of the climbing the mountain analogy and at some point, it makes sense to step off and build a nice chalet, but if you build that chalet at too low an altitude you start looking up thinking higher would have been better.
But why go back to Adobe then? It’s not like you’re saying, “Oh, I’m going to go leave for a startup,” which would be what you would expect, you actually went back to the big company behemoth. What was the thinking there?
SB: Well, I think a couple things. One is that I had realized that I was more of a mission-driven entrepreneur than a serial entrepreneur — there’s really two distinct types, I feel like. Some people I know, they just start company after company and it’s almost like this masochistic desire to go back to the early stages and almost kill yourself, whereas I was just always motivated by this idea of empowering creative people, making people more creatively confident, unlocking new tools for new capabilities, I just love that stuff. I have a super geeky level of love for that.
So that was one thing, and then it was Shantanu [Narayen]. I had a relationship with Shantanu and he was in town and we had dinner together and I figured I’m not working for him, so I might as well tell him exactly what I think should change at Adobe. I felt like design, engineering, and product needed to be in the same organization. They had never had a Chief Product Officer, I felt like that was missing. For the strategy, I felt like we needed to start getting these products to the cloud and to mobile, brimming with all these ideas and he called my bluff on it.
He’s like “Ok, how about you do it? You think we need a new Chief Product Officer, show me how it’s done”.
SB: Initially, I actually said, “I’m happy to be an advisor, but I don’t think I’m ready to jump in and come back,” that just wasn’t at the top of my mind at the time. When I jumped in to this advisor role for six more months with him, meeting with a number of the leaders and sitting in on product strategy sessions, I kind of got anxious, I was like, “I want to do this. Someone needs to fix this, change this, do this.” I was just like, “Okay, I want to,” and that’s when he brought me back. That was a little over five years ago.
I think one of the things that strikes me about your story is if you hadn’t left — you call it a mistake — but you would probably always feel like, “Oh, I’m just stuck at the big company.” There’s a bit where now I can imagine it’s easier to put up with all the big company BS because it’s like, “No, I chose to be here. I’ve been on the other side, the grass isn’t necessarily greener.” I think the payoff for being at a big company, that’s hard to keep in mind, is moving a cruise ship a foot makes a massive difference, even if in the moment, you don’t feel the massive difference, and knowing that actually was valuable to you and the way to have the broadest impact on creators, it’s hard to think of anywhere else you would be. So I’m not sure it was a mistake, I think maybe it actually sets you up well for this job.
SB: That’s fair. Listen, I remain an LP in Benchmark and I have those relationships still. I learned a lot working with the team and also now, being an executive in a big company, but having also been an investor and getting that sense also as an entrepreneur and going through an acquisition of my own and now doing acquisitions, I recently took on strategy and M&A for the company, so I am also now on the other side. It’s great to have that vantage point because you have empathy, that’s like the secret to any product leader, is if you’re creating a product for yourself or you have great empathy with those who are suffering the problem, you make a better product. I’m sure it’s the same for any job function.
Adobe and the Web
That makes a lot of sense. Well, to go back to that interview I referenced above, it happened after Adobe MAX 2021, when I think the big announcement was that Photoshop was coming to the web. I want to go back to that point, because I think the web bit here is very interesting, where in general — and I think this happens a lot when you look at a new platform — like Photoshop on the PC makes a lot of sense, it was built on the PC and it’s been thirty years or however long it’s been a developing on there, and you have the highest performance. If you’re coming from that mindset, you’re like, “Well, everything about the web is worse. It just doesn’t work as well, X, Y, Z.” It’s really important to understand what’s the key thing about this new platform that’s unique to that platform and only possible there?
In that case, from my perspective, it’s collaboration. Collaboration is native to the web, that is what the web does better than anything. You could do collaboration in apps, but then you have syncing and you have all these issues that are challenging, and it’s not native in the way that performance is not native on the web, but collaboration is. That was my big takeaway in why that was meaningful to go there. You started out just with commenting and feedback or stuff on those lines, which made sense.
I’m curious though how that fit in in that framework with then looking at the Figma acquisition? That’s what makes Figma unique, everyone talks about how they built this rendering engine and all sorts of stuff, which is important, but that’s table stakes so that they can leverage, from my perspective, their true advantage, which is native collaboration. To start off, is that the right framework? Am I thinking about the differences there in the right way?
SB: Yeah. A few thoughts here. First of all, clearly, the web is conducive for collaboration, not necessarily multiplayer collaboration, which requires another level of sophistication, and obviously, Figma is amazing, we can talk about that in a minute, but the ability for me to either pass the baton or for us to have some small number of people working together, the web is conducive for that. Another thing to keep in mind, by the way for the business, is that the web is an incredibly effective conversion funnel. If you go to get Photoshop, you’ve got to find Photoshop, you have to click download, you have to wait a long time in some cases for it to download, and then you have to remember to open it. Then when you open it, you have to go through this onboarding that cannot possibly be personalized for you because it’s a desktop binary that you just downloaded, as opposed to a truly web-enabled experience where we can have progressive disclosure and invite you into a product that is, by the way, forty years of features and possibilities and preferences and all this stuff.
One of the daily dramas I find managing products of this nature is you have established customers that are like, “Just don’t change anything, just stop. No, stop, stop don’t touch,” and then you have new customers that are like, “What the hell is this? How do I even orient myself and why can’t you make this simpler? It’s like a cockpit.” There have been a lot of scenarios — we just recently changed the onboarding to Premiere Pro, for example, which is what Hollywood uses, a lot of the Oscars were made in Premiere Pro — and we made a change that affected how you source the footage that you bring into a project when you’re making a film or a movie or whatever. It’s fascinating that when we shipped the change, the existing customers, a lot of them really hated it, they were like, “Why’d you change my flow?”. But the new customers were converting at a materially higher rate, and so you have this juxtaposition of you’ve welcomed new customers, but the NPS went down for a period of time. That’s an example of what we face. With Photoshop and Illustrator on the web, it was an opportunity to actually use these products for new customers that would convert in an easier fashion, and we also didn’t have to bring forty years of everything to the browser, because you know what? If you really want that, download Photoshop, but if you’re new to Photoshop, you actually don’t know what you’re missing, so we can bring you in with a much simpler and progressively disclosed experience. So there’s a key part of the strategy that is related to getting new people successful more quickly.
Was that the most important part of the strategy?
SB: I would say it’s equal. It’s collaboration and our mission for Creative Cloud, which is creativity for all, so it’s making sure that Photoshop is more available to more people.
The reason why it matters is because that priority still holds today. You still want a better top-of-the-funnel experience, you want to be able to differentiate the product that you deliver, and it means that looking to acquire Figma was not an admission that the strategy didn’t work, because no, they’re actually accomplishing two different things.
SB: Oh yeah, completely different things. Now you think about the fact that we’re in a world where Photoshop new users will increasingly be cloud-first and maybe web-only users of Photoshop, and same with Illustrator and maybe future products down the line, so realize all these products then are creating cloud documents as opposed to local files. That also helps you understand some of the strategy possibilities with a product like Figma, where Figma is in the product design and development space, and we can talk about why that is inherently multiplayer and involves developers and why it’s really different, obviously, than a Photoshop or Illustrator product. But if you take a cloud document of an image that you edited in Photoshop, you used to have to save that as a lossy local file, and send it to your friend in Figma who’s building a product experience. Now you can just actually embed the cloud document and when it gets updated, it updates in real time.
I have another way I’ve thought about Figma and Photoshop and I want you to tell me if I got it right or wrong, and this goes back to my analogy of Adobe and Microsoft. For Microsoft, Windows was the linchpin of everything they did and that was predicated on API control, but one of the issues with SaaS was API control didn’t really matter anymore, the new hardest job was coordination. The way that I’ve thought about Teams and why I think Microsoft’s executed so well with that is they they have set up Teams as basically the operating system in the cloud. They’ve leveraged up into owning the linchpin once again, but the nature of what makes it the linchpin is very, very different than when it used to be Windows. Fast forward to Adobe, the applications stay local for a very long time for very good reasons, particularly performance, there’s lots of legacy, all those sorts of things, and then Figma comes along and it’s not a Photoshop competitor, I think that’s super clear, I’m sure you would want regulators to agree with me, but it’s not a Photoshop competitor at all.
SB: You can’t edit an image, right.
Yeah. But what it feels like, it’s like the operating system for design, where it’s the central place where everything plugs into. The risk I saw for Adobe with Figma is not that Figma replaces Photoshop, but Photoshop is just a commodity, it’s just like a tool that plugs into Figma. I have a couple questions about this. Number one, do I have the wrong framework, but number two, if I do have the right framework, what was the Windows of design before Figma? What was the operating system, the core thing that tied everything together, or did that never exist and that’s why we’re emailing files with version numbers on the end of them?
SB: Listen, design is like the word “work”, it encompasses a lot of different stuff. Design includes illustration, you could argue it includes motion graphics, it includes certainly posters and splash pages and ads, banners, and everything else. The way I see Figma is that they really, first of all, there was a new segment emerging years ago called Product Design, and I think that people realized —
Right. Sketch was the first one there.
SB: Exactly, it had to be interactive. Sketch was really the first mover here and said, “Wow, people are tripping over themselves trying to make these graphics in a product like Photoshop and then printing them out and drawing red lines so people know which way to go. I have an idea, let’s just allow people to do it all in one space and let’s make it interactive and allow you to turn it into a prototype,” and so you had Sketch and InVision working together and that was the early days of product experience. Then what we realized is, well, Photoshop is a single document and it’s typically one person editing an image, but with product experience design, you have potentially thousands of different states of the users’ experience, and so you need to have tens, if not hundreds, of designers engaging with all those different states.
And you need engineers to implement it.
SB: Right. Actually, a lot of the early solve for that was version control, so Sketch pioneered some things in version control. And then Adobe was like, “Well, that’s interesting. We should probably have a desktop product also that does some of this stuff,” and so we built a product called XD, which was unsuccessful. Why? Because the whole product design and development space — it turns out there’s some underlying pillars that became absolute essential. One of them was the multiplayer capability for people to do all this together, and that was something that only the web enabled. Another one was the fact that wow, surprise, surprise, developers are a part of this ecosystem, and the developer handoff key piece, the fact that I think over 40% of Figma users are developers. Credit to [Figma CEO] Dylan [Field] and team, they pioneered this new approach to product design and development that was multiplayer, web-based, developer-centric.
I think that’s the key transition. Sketch created this concept of designing a mobile application and they really blew up with the iPhone. We need multiple states, we need to be able to flow through instead of a single document approach, so they got that part right. But this goes back to the web being important, by definition, that’s going to include a lot more people than just creating a single graphic and Figma put it on the web.
SB: It’s developers, it’s copywriters, it’s product managers, it’s legal folks, it’s everyone involved with an interactive product experience, and so it’s an entirely different genre. As far as I’m concerned, it’s like Illustrator versus Lightroom versus Premiere Pro, these are all design tools in essence.
For the record, I completely agree with you. The question is if the regulators agree with you, but I’m not going to push you on it because I agree. But sorry, continue.
SB: I think that we’re aligned on that. Then as to where this goes, we see the workflows that these product designers and developers go through and ultimately, these are digital experiences that include assets, like video, that need to be edited in Premiere Pro or products like it and images that you might want to edit in Photoshop and whatever, and so obviously we talked about enabling all these cloud-enabled products to work together would improve workflows and hopefully make people more efficient, so that’s the grand strategy.
One question in general is I think a perspective why Adobe would pay basically a strategic price for Figma is not that it’s a Photoshop competitor, but I sort of mentioned it before, it makes whatever tool you use into a commodity because it’s a centralized piece and it’s best for Adobe if their products are preferred or the best possible alternative. Where are you thinking big picture, and I’m going to come back to this later in the context of Firefly and AI, of how Adobe fits? Is the ideal experience everyone just lives in Adobe all the time, or you could make the Microsoft analogy here, like actually, it turns out that Teams is being the operating system of the cloud, but also in the cloud, you can actually favor your own services that much more, and people are pining for the days of Windows, when there was an open API and you could build whatever you wanted to. Now, Teams gives you a little box and you can slot into it or not. Where’s your view of that going forward?
SB: I have always taken the approach of, it sounds trite, but let’s do what’s best for customers and it will ultimately serve our business. You look at Frame.io, a company we bought for $1.3 billion that does video collaboration, a lot of their customers were using Premiere Pro and a lot of them were using DaVinci and Final Cut. That’s great, we continue to support those third-party tools and they’re integrated into Frame.io’s workflow and that’s great for Creative Cloud and for Adobe, because if you were using Final Cut or another product, you weren’t an Adobe customer and now you are because you’re using our collaboration service.
I’ve tried to have our teams have that approach to building our products now, which is we want all of these third-party products, and by the way, we know that there are going to be many, many of them, there’s going to be all these very niche specialized animation tools and all the generative AI stuff emerging, which we’ll talk about, if we don’t have an approach where we can accommodate that. But to your point, it’s a lot of collaboration and making these things work together and that’s to our benefit.
What’s the bigger mind shift internally for Adobe, is it shifting to a web collaboration being at the core or is it to a, “If we do that, that means we need to be open and more interoperable” and whatever that might be?
SB: My view is that there are going to be more tools, not less, in the future, and that our market is expanding. We’re not just focused on core creative pros that went through design school and hold their skills of using Photoshop as stripes. It’s now a far broader audience of folks who want to be a part of this, non-pros and pros alike, which means that there need to be collaborative capabilities for all of these products, and so the view has been that if we can ensure that Adobe’s providing indispensable value to all of these folks, they don’t need to be on our first-party tool, they can be using third-party tools, as well.
We’ve not just said that, we’ve delivered on that now. Frame.io has continued to ship releases for these third parties. We’ve also done things like the Content Authenticity Initiative, which we made open source and invited all competitors, anyone who wants to use it to use it. We have 900 partners now adding content credentials to assets using some of the technology we created. That is our ethos and I think that’s a smart way of going through this. Also, by the way, creatives are first adopters, so they love using new tools that are made by startups. Let’s just try to make sure that we can include them in ecosystem and foster collaboration in terms of how people use these tools together.
Firefly and Generative AI
Talking about generative AI, I think it’s actually kind of interesting in this individual versus collaboration framework. At least for now, it’s sort of an individual AI experience. You can have a group chat with AI, but the AI’s not really cognizant of that. I guess that will change, I’m sure everything’s going to change over time, but is there an aspect where, you just announced Adobe Firefly and I want to get to the other Adobe Summit announcements, as well, but does this feel like it’s back in Adobe’s sweet spot, where it’s like, “We want to give you a super high performance tool and you and your application can work together to create something unique”?
SB: The view on generative AI for us is really two things. One is we do believe that people should become more creatively confident in general and that serves our Adobe Express agenda and some of the things we’re doing for digital marketers, we can talk about that side of the equation and that’s a market expansion opportunity. In some ways, a prompt is the new template.
I wrote about this last year, I think, in the context of some of this stuff, where there is a split between creativity and execution that no one really perceived existed before, but actually, there are two distinct things. You can have creativity in an image space without being able to draw, that’s something that is certainly hugely expansionary.
SB: The way I’ve been saying it is there are creatives and there are creative directors, and everyone’s becoming a creative director, not a creative. You’re a creative director by giving a series of prompts to make an animated short, you’re not necessarily picking up your pen. It is the ingenuity or lack thereof that determines whether it’s a good output. I think that Firefly is only as good as your prompts and how creative your prompts are.
Well, give me the Firefly pitch, because I want to spend a little bit of time on it. What’s the elevator pitch for Firefly?
SB: Firefly is our family of generative AI models. We certainly have the classic text-to-image but we also have text styles, so you can actually, with any prompt, affect the style of fonts, which is something new that we brewed in the lab. And then we’ve also done a lot of work now around vector creation, text-to-vector creation, and also taking a sketch and then creating multiple vector options that you can then take and then have infinite points on Illustrator to further edit this vector. We’re taking a few different approaches to this than I think what you see elsewhere.
Number one is we’re very focused on how the model is trained. We want these to be clean models. When I go and sit down with the CIOs of some of these big marketing conglomerates and they say they don’t want to use generative AI because they don’t want to attest to their customers that they use something that they weren’t allowed to train off of, and so we want to have commercially viable models, that’s one important thing for us. Number two is we want to make sure that we bring these into the tools that people are already using. For the vector thing, for example, it’s made to work with Illustrator. Also, what we’re doing in terms of bringing this technology into Photoshop, we’re allowing people to further edit the prompts and have the outputs be nondestructive so people can actually further edit with it, as opposed to just be stuck with whatever the prompt gives you.
Firefly, I think, is just a web tool now, but you’ll be able to use it in Photoshop on the desktop application?
SB: Yep, exactly. What we decided to do is launch the Firefly suite as a beta. This is, again, something new we’re trying to do, we’re trying to get our teams to push things into private betas and then public betas, whereas I think years ago, it was like nothing would —
The hangover from 18-month releases is hard to escape.
SB: Yeah, it is, actually. It’s really interesting. People still think everything has to be perfect before you ship because it used to be printed on CDs, so it’s a different world. Then we are testing it and getting feedback, we have hundreds of thousands of people that are playing with this now, and then we’re bringing those capabilities natively into the tools, but we’ll also, of course, keep the Firefly suite of capabilities for people who just want to use it from there.
Obviously, a huge focus, and you mentioned it right here, is about this authenticity bit, and you’ve talked about that. Well, there’s two angles to that. We’ll get to the authenticity program, but as far as the training goes, you are pushing that it’s only on public domain images and Adobe Stock photo library. Now, I’m not sure if I can extract a truly honest answer from you because you are an executive at Adobe and so you have a motivation here. I will say, looking at it from the outside, number one, it’s not really clear to me why using photos is not fair use, it seems like a pretty textbook example of transformative work. There’s a bit where Adobe is leaning into the use of these photos for training as not being ok, and it kind of feels like a weird form of regulatory capture in a way, where it’s in Adobe’s interest to have actually a very narrow view of copyright and that this is not advanced by fair use, because “Hey, we have a huge stock photo library and we have the capability of attesting that this is the case”. I want to give you a good question that you can talk about here, but is this just your view? My guess would be that “Look, that’s what the big companies want, so we’re going to give it to them”, is that the way to put it?
SB: It was interesting, I was having a debate earlier today, actually, with a friend around — does this go the route of the Uber’s and Airbnb’s of the world, where they, on a regional level, were pushing the laws and then they just had to fight region-by-region and they were able to just get the galvanization of their users to allow them to become, in most areas, legal and accepted? Or is this more the Napster and LimeWire world, where there was an era where it was just a free-for-all and then suddenly, Congress and I guess on a federal level, there were just determinations about what the outcome of this would be, and then it actually all went away pretty quickly and then you had iTunes and Spotify and whatever else.
Here’s the thing, Ben, I understand today, it seems like it could go either way, but if you just play this out six-plus months from now, where you can actually make unauthorized sequels to movies and you can, say, you can watch Succession and be like, “You know what? I’m going to make my own script out of prompts with ChatGPT-5,” or whatever we have by then.
I can’t do that, I’m too busy making Balenciaga commercials.
SB: (laughing) Exactly. Then I’m going to feed it into a text-to-video model and I’m basically going to make an unauthorized episode using the exact actors, their likeness, their voices, and everything. I seriously believe this will all be achievable very soon and it’s just going to be a copyright dumpster fire. What is going to happen? There’s no way they’re going to be able to commercially monetize that content, there’s just no way, you’re using Spider-Man, you’re using The Matrix, you’re using whatever. If you believe that that’s an inevitability, which I do, then I think that there will have to be clean models that you can rely on commercially. Sure, you’ll make fun memes and unauthorized things that will creep around the internet, but from a commercial perspective, I don’t think it’s just a big company party line, I actually believe this is going to be a requirement for this technology to work commercially.
I could give the snarky response that it also is definitely in Adobe’s favor, but this kind of goes back to the tying in with other tools. You mentioned there’s a whole workflow component to this and you released a new product — I don’t want to mess it up because Adobe’s names are so great — Productivity Analytics. Honestly, I love the fact that Adobe’s names are so descriptive. Back when you shifted to the icons with just the letters, I actually thought that fit Adobe, from my perspective, that was years ago. But people are going to want to use Midjourney, they’re going to want to use DALL-E, is Adobe going to incorporate, are they going to have hooks so people can use those super easily or is that going to be, “Well, if you want to generate an image somewhere else and bring it in, but if you want the full integration, you’re going to have to use Firefly”?
SB: It’s a good question. Some of this stuff is obviously still under development, so we’re having these conversations, as well. My view is that people have always made mood boards in their creative process, they have all kinds of paths to inspiration and figuring out what they want to do, and in that stuff, everyone uses Google Images and other things. You’re allowed to use anything in a mood board. You’re not commercially selling it, you’re just leveraging it for inspiration, and so people should have access to third-party models with our tools. I hope we can allow people to use them for various purposes. I do think, again, that when it comes to commercially viable banners that you’re selling to a client, though, I also think the provenance part’s going to matter, too. The reason we’re adding content credentials to every single asset that’s made with Firefly is because we want people to know what model made it. We think that OpenAI and others should do the same so that any asset you see, you can see which model made it and then you actually know what that means.
Content authenticity, is that going to come to Photoshop, too, where if you edit a photo in Photoshop, it’s going to be there?
SB: It will.
Is that going to be an opt-out or it that just going to be a part of everything?
SB: We’re going to be bringing it into Adobe Express by default. In Photoshop we’ll have content credentials as an option. I don’t see us making it like locked down, opt-in-only type of model.
That’s pushing the old heads stuff too far.
SB: But we want people to do it. What it also does, by the way, is it shows how a piece of media was edited. That actually was the origin of content authenticity four or so years ago, when we got in a room being like, “What do we do about fake media and deep fakes?” Governments were asking us, other people were asking us this question. We realized we’ll never be able to have an algorithm that forever detects whether there was edits, it’s like a cat-and-mouse game, you make an algorithm and then someone figures out how to get around it. Instead, we wanted to build a world in which people could add content credentials as they wanted to and it’s almost like the equivalent of a verified badge on an asset as opposed to a Twitter account, which, again, is a bad analogy these days. For an asset, if you can tell if an asset, if it has that badge, it means that you can see the provenance of the asset and how it was edited, maybe you will have more trust than an asset that doesn’t have that badge. In the world we’re going into, where I can make a whole Ben Thompson interview using your likeness just with a script and it will be undetectable, you’re going to need that.
I agree with that view. I think the approach to this needs to be affirmative that this is true and the entire approach to misinformation, disinformation, is not that you’re going to have a top-down censorship and cut stuff off. There’s not only just fundamental philosophical questions, like who decides what’s true? There’s just a reality of the Internet perspective, which is, I made this analogy ages ago, I went back to, what’d I call it? Zero Trust Information, actually, an article I’m very proud of, where you can’t have a castle-and-moat sort of approach. What is zero trust networking? It’s the idea that you’re going to affirmatively say who you are, in every transaction, you’re going to assume everything is dishonest or not reliable, there’s just no default trust given and you have to affirm that. The reason why networking is moving in that direction is because you had this explosion of mobile devices, and now you have watches and whatever and all this sort of stuff trying to connect to a corporate network. It doesn’t scale, it doesn’t work on the Internet.
Actually, this goes back to, and I’m doing my little rant in the middle of an interview here, I think this goes back to the music thing. The key thing that I see from music is not just the lawsuits and government action, it’s that what is Spotify? Spotify is an embracing of the Internet — why would you not use Spotify? Instead of selling a disk, you’re selling convenience, where for $10, you get access to everything and do you really want to jump through hoops for $10? Your cost, it’s not just the cost of you might have the RIAA file a lawsuit against you, it’s that your time is, it’s $10. For $10, you could just get it all, why even bother? The music industry — dragged kicking and screaming by Daniel Ek — actually embraced the Internet. All solutions, in my mind, have to be enabled by the Internet, whereas in this case, what can you do? You can always connect to another server and that server could be Adobe’s content authenticity that says, “Yes, this is what it is.”
SB: To the analogy of music, the thing I should say about generative AI is I actually think there is a way to have content contributors get paid, much like Spotify figured out from music. One of the things we’re exploring right now is allowing artists to license their style. We talked about prompts and you have to be creative to make a great prompt. Well, imagine at the moment of prompt, having a style library that you can browse and leverage and add to your prompt, and then you pay that creative some small amount or they get some cut of the overall subscription or something to compensate them for the use of their likeness or their style or Mickey Mouse if you’re Disney and you’re licensing people to use it. I do think that there is a method, a Spotify-like approach, that, to your point, leverages the greatness of the Internet, makes it simpler, easier, and legal to use copyrighted content or characters’ likeness in interesting ways.
I think you said the keyword, which is attribution. The Internet makes it actually possible to track attribution. One of the things I’d love to do with Passport, my thing, is imagine there’s people who are actually really good at curating content. I always talk about my assistant Daman, he’s the podcast king, he listens at 3.5X speed, he plows through so many podcasts in a week. I would love to get the Daman podcast feed, where he actually chose all the best podcasts in the entire universe of podcasts for me to listen to and I just listened to his list. It would be great if he could get a little bit of cut and all the podcast people that he recommended got a little bit of a cut, and that’s only possible on the Internet. People always look at the Internet because you’ve brought offline business models online. All they see are the detriments and how it hurts their old business model and it’s important to look at what’s now possible that wasn’t possible before. I think your bit about the artist style, it’s flipping it on its head. This isn’t necessarily a bad thing, it can actually be a phenomenal thing in expanding the market for creators.
SB: I think that’s where we’ll ultimately go. Listen, I don’t like focusing on regulation as a way of dealing with things. I personally like allowing technology to lead us to a better way and finding a way to empower the folks that are involved with it, as opposed to have the grownup come in and be like, “This is how it’s going to work.” Look at Behance, it’s 40 million profiles now. If you look at Midjourney, a lot of the prompts that people use are artists’ names and even sometimes the word Behance, just because they want to make sure that it’s referencing that profile, so I therefore assume that they scraped Behance. It’s pretty wild that they do that as a third-party. Of course, we would never do that, even though we own Behance because we’re stewards, but we could very easily allow people to just make their style available, create a monetization model for it, and then surface this in the interface in unique ways and hopefully take the Spotify approach, where it’s actually net good for everyone involved.
One thing that’s interesting is you used the word attribution and I latched onto that, and the reality is is we know that attribution works very well because attribution is thought about in ad tech. When there is money attached to it, it turns out that the tech industry has been very effective at getting phenomenally good at attribution. We know it works, it’s just like “Can we attach the business model to it”? I am interested, one thing that Adobe’s talked about is you have Creative Cloud, which everyone knows about, but then you also have — what’s the marketing cloud’s name?
SB: Digital Experience.
Digital Experience. The Experience Cloud, I think you’ve started using that term, if I recall correctly.
One thing I am very interested for generative AI is we have A/B testing, it’s something that is an established bit. It seems obvious to me a super clear application of generative AI is not just A/B testing, but basically A/B generation and on-demand in figuring out the best image. Is that something that Adobe’s focused on?
SB: Yes. One of the things that we talk about a lot internally is how, in the future of personalization, right now, it’s still segment-by-segment. Even the best marketers in the world effectively segment all their customers and then they make various versions for those segments, as opposed to why shouldn’t I be able to make literally a custom asset for every single person in my database based on what I know about them, their profile of their gender and where they’re based?
The limitation was being able to generate the art. Now, that limitation is gone.
SB: Yeah, exactly, and that limitation is completely gone. That is a huge focus for us and part of it is building out this CDP, the Customer Data Profile service, and making sure that this collection of technologies that we’ve bought and built over the years on the Experience Cloud side of the business, that that’s all stitched together. We are talking a lot now about this vision of a content supply chain and we use that term to talk about the origin of the creativity, like the work that’s done on the Creative Cloud side and all of those assets and campaigns, all that stuff that’s set and put into the hands of marketers, and then the abilities for the marketers to say, “I’m going to deliver this campaign, but I’m going to add personalization with these parameters,” and then have generative AI do all the rest with those assets. That’s gone from a pipe dream a year ago to actually very achievable and should change everything. Can you imagine how that would impact conversion, just based on a very small amount of data and customer profiles?
You mentioned it seemed like a pipe dream a year ago. How much of the last year has been a surprise to you?
SB: I think that the pace and availability of these models, it has surprised me, even on a cost basis. When we were talking about some of this stuff a year ago, there was this assumption that it would just be massively and potentially prohibitively expensive.
That AI would always be centralized because only a big company could afford it, but now, you can run Stable Diffusion on your phone.
SB: It’s amazing. I think we’re actually still in the fire hose mode right now, where it seems like every week, there’s some new thing that a researcher internally realizes that we can do to either reduce cost or improve output. It’s just a super exciting time.
Should Adobe actually lean back into the desktop? Because one of the biggest things about AI and the biggest shock to me is exactly your point, where if you can run these models locally, just fundamentally, it’s transformative as far as the possibilities here. There’s obviously, if you run it locally, there’s no centralized control, but it’s also the case that inference costs effectively go to zero. What’s weird about AI today is we’ve been living in a world where you assume zero marginal costs for everything. Now, of course it costs money to do a Google search, but you operate your business as if it’s free, every marginal search is zero. With AI, a decentralized service, at least for now, you do have to think about marginal costs, but if it’s local, you never do. Guess what? What is the one local application that people still put on their PCs? They still install Photoshop.
SB: There are a number of these models that we are planning on doing locally and are executing locally in terms of actions and then there’s some that, for various reasons, we believe are best in the cloud. Also, of course, we’re trying to learn how our customers are interacting with these tools and then using that data to improve the models themselves, and so in some cases, it’s still better to have them in the cloud. But putting some things available locally in the products where we can makes sense, for sure.
SB: Syncing my AI with an app as opposed to using an app was kind of what I was trying to say. What I meant by that was we have all this preferences, data, we have all these things that are aggregated about us that conceivably, whether it’s this WHOOP band I’m wearing or my health or anything else that is tracking or recording us or whatever else, I have always had to use apps, that’s just the pattern that we have when we are doing anything or completing tasks or whatever. I am sure the days are numbered until we actually just sync a repository of our data with an application and suddenly, we’re not even really using it in the traditional sense. We essentially become the app.
It’s the plug-in idea, right. You just add another plug-in.
SB: It’s essentially like plugging in yourself. I guess it is a weird metaphor, but that’s probably true. Because I’m thinking about it as a product designer, I’m like, “Wow, so there’s no onboarding anymore.” If you literally just plug Ben in and then you turn on and it’s a health app, it’s showing you you and it’s showing you everything about you, and you can just ask questions and it will respond to you. But it very much changes a lot of the playbook for bringing new customers in, onboarding, all the various little tricks that product leaders use, it’s just fascinating to me how drastically that’s also changing.
How important is it for you then, to go back to the creative process and workflows to have people putting their assets into the Adobe Cloud so that you can deliver on that experience? I guess part two of that is when it comes to Firefly and these questions, is this going to be you’re going to be delivering that on a per-enterprise basis, as well, so they can incorporate their assets without knowing it goes back to Adobe? I guess this is a big picture privacy question, but it’s not just privacy, it’s also corporate data and things on those lines.
SB: Part of our strategy is we want creatives to be able to build their own models. The idea that your own style, all the work you’ve done, we have customers in Lightroom who apply the same presets to every single asset that they make for any customer that they work with.
That’s their style, that’s kind of where it comes from.
SB: It’s their style and why shouldn’t we automatically just do that for them? By the way, whenever they delete images or keep other images, we can also learn their preferences and we can start to automatically do a lot of that stuff for them. Any great photographer will tell you their dirty little secret is they just take a crapload of photos and they spend a lot of time sorting through them using their unique style and preferences, and that’s something that can be learned.
That’s a good example of how digital photo photography ended up being fundamentally different than photography previously. You’re still taking pictures, but it used to be you had to spend so long to get the picture right at the beginning and it took photographers a while to adjust their mindset to no, actually, just take as many as possible, and the workflow all shifted to post-snap as opposed to pre-snap.
SB: By the way, that’s a good example of a team that’s very sensitive to the cost of cloud storage and cloud operations. Some of these models on the AI side are being done locally, to the point you made earlier about just the efficiencies of that. Enterprise of Firefly, though, just to end the thought. Since we’ve launched Firefly, we have a lot of corporate customers that have reached out and they’re like, “We want this.” One of the reasons is that a lot of major brands need to think and act in real-time when it comes to marketing these days. If you’re a social media marketer, you have to respond in 30 seconds to something that happened.
The Oreo moment, right?
SB: Exactly, the Oreo moment, and so that is with the Super Bowl and the lights went out — it’s like how do you empower a team to do that? You need to give them tools like Firefly and an Adobe Express of course, the things that we’re building, enterprise has realized they need. I do think they’ll have their own private models, to answer your question, that leverage their own IP.
The Next Generation
This initial version of generative AI, and this happens with the new version of any sort of technology, is it makes the incumbents better. It’s super obvious in this case, Word is a word processor, what if it actually wrote the words for you? Photoshop is an image manipulator, what if it actually generated the image for you? It’s very straightforward. But in the long run, there are completely new things that are possible that were not made possible for, and it takes a while to figure out what those might be. How long do you think will be in this first era, that I think is obviously to Adobe’s benefit, before the second era comes along, which if it tracks to other shifts of this magnitude, Adobe will have a harder time responding to?
SB: I think that if we were only focused on bringing these generative AI capabilities into our tools, then we would get stuck, because you can only do so much of retrofitting and putting things into things. That’s why we’re trying to do both the AI-first creative approach, which is built out of Firefly as it is in the open today, as well as bring these capabilities into our tools, we have to play both. I agree with you, there will be some use cases where someone makes videos solely out of prompts and never even bothers to learn Premiere Pro. I want to make sure that we’re still giving them the opportunity to do that within the Adobe environment and also, if they want to take it to the next level and make a cut that they can’t really do with words, I think they’re going to want to bring it into Premiere Pro, but a lot of people won’t and that’s fine.
It will be interesting to see. I do have one final question on the personal level. My daughter’s an artist, she likes graphic design, in particular. What advice should I give to her as she’s thinking about her career going forward in this new world?
SB: The advice I always give to emerging designers and creatives is first of all to play. The way that these artists become working at the edge of their discipline is they just play with all the new stuff and they just follow the folks that they admire, because creativity is the world’s greatest recycling program. In some ways, generative AI is the computer doing what humans do today, which is you take all the things that inspire you, you put them in a mood board, you mush it up in your brain, and then you make something that’s inspired by what you saw. Every input yields outputs, so your daughter needs as many interesting inputs as possible, because that’s going to differentiate her output. She needs to play with a lot of these tools, to start to be creative in how she wants to use them. Listen, do I think that she should learn Adobe’s products? Of course. The reason I would say, though, is creative control. Every creative ultimately wants control. Control is not just that art director yelling at you and you want some autonomy from them, it also means that I want these pixels to be exactly as I see them in my brain and I want to have explicit, precise control. I just feel like every great creative ultimately lands in that same place, and I’m just not sure if prompts alone will get you there.
Well, Scott Belsky, it was great to talk to you. Now, there is officially an interview that will show up when I search for your name in Stratechery, and it certainly could not be a more interesting time to talk about this stuff.
SB: It is. It’s a lot of fun. I’m sure a year from now, it will all be different and that’s what makes our jobs interesting.
This Daily Update Interview is also available as a podcast. To receive it in your podcast player, visit Stratechery.
The Daily Update is intended for a single recipient, but occasional forwarding is totally fine! If you would like to order multiple subscriptions for your team with a group discount (minimum 5), please contact me directly.
Thanks for being a supporter, and have a great day!