E078: The SaaS Shift: What Claude Design Signals for the Future of Software

(Transcribed by UniScribe (https://www.uniscribe.co). Upgrade to remove this message.)

Hello, welcome to the Big Strategy Podcast. I'm your host and fellow strategist, Jeff Ayatt. This is your go-to spot for real talk with the masterminds of business strategy. It's about sharing stories, insights, and those little nuggets of wisdom to help you unlock big growth.

Let's turn the key and spring into action. Welcome back. Last week, I started playing with Claude Design. Basically, it's Google Slides within the Claude environment that designs and allows for real-time editing of presentations.

It's pretty cool. Claude Design launched in April, and the day it dropped, Figma's stock dropped about 7%, and three days before that, Anthropics'chief product officer, who was on the Figma board, quietly left. The market called it the next chapter in the SaaSpocalypse. sitting right next door with image generation, background removal, and I thought, What's another example of this?

And I thought back to Garmin in 2008. Maybe it takes me back to like when my wife used to drive into San Francisco to visit and she, well at the time it was my girlfriend and she had her Garmin on the dash to get to my apartment. Back in the day you bought the device and then you paid for the maps. So if you wanted an updated map you had to buy the chip or the download whatever it was.

But then when the iPhone came out Apple. and then very shortly Google included navigation in their OS for free. Overnight, Garmin's whole value proposition became a checkbox. I remember when we sold my wife's car after we got married, and that Garmin was like, tucked in the door pocket gathering dust because we had switched using Apple Maps.

So I wrote it down and I thought about it. uh with in the current day with regard to canva and its design uh and ai tools and worrying that it might become a feature in somebody else's subscription Basically meaning: is chat GPT or Claude just going to make canvas core feet functionality a feature of its product then i did the thing that i tell every student and executive client to do and hopefully every one of you do the same i took my best idea and i went looking for the smartest person who disagrees with it And what I came back with was this idea that platform featurization always looks inevitable until the company you wrote off does something that you didn't predict.

The same week that Claude Design knocked Figma to a 52-week low, Canva had shipped AI 2.0, running on its own proprietary model, claiming roughly 7x faster and 30x cheaper than Frontier models. And this is not a company waiting to become a feature. That's a company building underneath the thing that was supposed to eat it.

And again, if you're Canva, you don't need a model advanced sciences. You need a model that can specifically create content and publish that content. And so they took that model and fine tuned it for their specific needs. And if you're an organization looking to cut your AI costs, I highly recommend connecting to talk about that.

But behind this, what was the framework? because it's still about the right framework. And I go back to one of the first lessons I learned in business. It was back when I was in college, back in the 90s, studying Jim Barksdale, who was the CEO of Netscape.

And he famously said at a Harvard Business School event, there are only two ways to make money in business, bundle and unbundle. And that's the engine of this entire AI era. A general that can do everything and then entrepreneurs unbundle it into separate workflows so you have like gamma right they were the first movers in generating slides or you had mid journey and generating videos or grammarly and checking your grammar etc then the model turns around and rebundle these workflows back into its existing product as features and Benedict Evans, the strategist, calls this the product problem.

An LLM that can do anything. So what's the product? In his mind, the product was never the AI itself. The product was what the AI sits inside of.

The raw chatbot makes you start from first principles. The product is everything that saves you from that. So hold on to that. That's the hinge for all three places.

where maybe my first... Look at. Canva proved to be a little thin. *music *Number one.

The first place that I looked and the first kind of not complete analysis in my initial thinking was that I treated Canva as one thing. And in reality, it's two. There's the design capability, generating images, writing headlines, knocking out backgrounds, etc. But then there's the design workflow.

And this is what its users use, where it has a template library stuffed with proven templates. You can upload your own brand. assets, et cetera, kind of like organizational level fonts, logos, et cetera. So no matter who from your company uses it, they have a standard set of, um, objects to add and what it really came down to was the approval chain the collaboration the publishing pipelines that went straight to social so by looking at Canva is one company, I skipped the capability layer.

That's the feature now. ChatGPT, Claude give it away, and that puts Canva in a little bit of a conundrum. If it charges for those services, it loses by comparison. But if they follow suit like Claude and ChatGPT, and give that design functionality away, it commoditizes their own upsell.

And again, this is the trap that Garmin sat in as well. But when we're talking about the workflow layer layer, taking your image and publishing it to social media doesn't disappear when the capability, such as generating the image using your brand guidelines, gets commoditized. When I use Claude Design, it gives me like the raw HTML of my slides. That's cool.

And I feel like I own that design, right? I'm not just giving it to an LLM, getting a result, not liking it, and then having to do some kind of back and forth that never really meets my expectations. I can go in and change the cell padding parameters. I can change the text and move things around on the screen.

And look, it's only going to get better. We know this about AI. But what it doesn't do is it hadn't, you know, Claude Design has no idea who my brand is, who are my collaborators, who signs off on the marketing campaign, and where or how the images that I create in Claude Slides get published. That's not a small gap.

That is like a grand canyon between becoming a feature of a model, the image background removal, and actually executing done framework where you don't buy a hammer because the hammer is a commodity what you buy is the picture hung on the wall that's the value added skill and canva is using that same model so when you when you name the risk you have to name the layer where that risk exists we're getting featurized is one sentence but it's actually two different fates So let's ask ourselves the question, which layer of Canva is the feature and which layer is the fortress?

the AI system, is. the feature, but The workflows is the fortress. Number two. Do we defend or do we relocate?

And so another example is Midjourney. Their image generator got featurized into every AI subscription. So meaning. I can go on to...

Chachipiti, I can go into Gemini, and I can get a video made within my existing subscription. So Midjourney was no longer necessary, even though they were a first mover in that space. So they escaped. by actually making a complete pivot and just recently announced their launch of Mid Journey Medical, which is a full body 60 second ultrasound.

In this case, the scanner doesn't run on Mid Journey's image generation AI. The imaging is licensed from a third party. In this case, it's a company called Butterfly Network and their ultrasound on a chip. And Mid Journey paid like Finally, they're not really using their AI yet.

midjourney is not using their own ai yet so whereas canva went and built their own ai midjourney is not using their own ai so what it did was it didn't carry this core capability into the next layer it completely abandoned the commoditized capability the model and it bought a brand new capability so this breaks my garmin parallel a little bit which makes it actually even more of different escape plays and a strategist needs to know which one they're choosing. Garmin relocated.

It took its actual capability, navigation, and it moved it into markets like marine, aviation, fitness, Same skill generating these maps, but where there's higher stakes, where life may be at stake, or where it's highly regulated. Nobody flies a commercial jet on Apple Maps, and that's where Garmin earned their keep. On the flip side, MidJourney just walked away from image generation straight up entirely into a new capability where it's hardware. And again, the most important thing to make that hardware defensible was the FDA pathways that could take years.

They purchased that capability through a licensing agreement with Butterfly Network. So both Garmin and MidJourney escaped featurization. Some were harder to commoditize, that was Garmin. And the other says, my capability is gone.

I'm going to go find a new one fast. That was Midjourney. Those are not the same bet. And for Canva, that's the live question.

Do you defend your capability or do you relocate the company? And there are two ways out of featurization. You carry your edge somewhere it still counts or admit the edge is gone and move your feet. Pretending it's the same move, is how you pick the wrong move.

Part three, rent versus own. Canva is not a passenger waiting for LLMs like OpenAI, ChatGPT, Claude, etc. to choose its fate. So looking back at Canva and their AI 2.

0 is a proprietary model that sets the speed and cost claims aside for a second and look at the structure of the decision. Figma rents its AI from Anthropic, the exact labs that are now building competing tools. said went and built its own. I tell this to people all the time.

It's do you invest your money in building your brand on LinkedIn where you are renting a network or do you take that time and effort and create something that you own your mailing list it's not capability versus workflow it's rent versus own so if your moat depends on a model you license from the company trying to absorb you in this case being figma renting a from anthropic that is now trying to eat you You don't have a moat, you have a lease. Canva looked at that and decided to own the foundation instead of standing on someone else's.

And this is textbook Clayton Christensen just flipped around a little bit. Whereas usually the incumbent dies because it defends its profitable core business and ignores the cheap, ugly new entrant. Canva's move is kind of the antibody. It defends the feature and went down a layer and owned the thing the feature stands on before the platform commoditizes underneath it.

it's relocating the floor. If you remember last week, this was jensen's competitive advantage in defining the foundation that everything else plays on so no competitor can really unseat in because nvidia in his mind is controlling the playing field so stop asking who has the better feature and ask who owns the layer underneath it that's where the lease comes due So where do I land in all this? After our analysis, I still agree that Canva is Garmin. I still think the capability layer gets featurized.

And I still think a lot of design tools are quietly updating their GPS. But I found three frameworks to remember. Let's walk through them. First, capability versus workflow.

I commodized the whole company, looking at Enva as one, when only one layer was at risk, the model, not the workflows. Second, do we defend versus relocate? This is Garmin versus mid-journey. Same play, opposite reactions.

And then third is rent versus own. I had thought of Canva as kind of this passive passenger, and instead, it's down digging a foundation, building its own model to control what its platform and workflows are built on going forward. Both things can be true at once. Canva can be a feature at the capability level and the infrastructure level at the workflow and model layers.

The pie can explode, and eventually the winner can still get disrupted someday. means the clock is longer and the stakes are bigger than i wrote so here's your exam question and it's one every company in your future portfolio is going to face don't answer are you the product or the feature that's too easy and the answer could be both instead answer the harder version if you're becoming the feature at one layer do you know which layer is still yours and are you You are, are you digging down to create an LLM that is your own, that you can comfortably build your workflows on top of?

Or are you still updating your GPS, i.e. trying to chase features on top of a model that you lease from someone else? As I always say, hold the framework, question the conclusion.

That's the move. Next week, we'll explore how two other companies are responding to AI-led disruption in uniquely different industries. In the meantime, I'm Jeff Ayat. This is the Big Strategy Podcast.

Thanks for listening. (Transcribed by UniScribe (https://www.uniscribe.co).

Upgrade to remove this message.)