Over the past few months, it seems like every company has launched an ✨AI-powered ✨ feature (or at least published a thirst trap GIF of a feature). It’s become a bit of a meme. Ever wondered why so many teams do this? Read on.
The AI roadmap hijacking (teams reallocated resources to focus on AI projects) deserves a post of its own, but I’ve been particularly interested in how all of these launches are, without fail, marketed as containing “AI powered” or “Built with GPT4”. On the surface it seems obvious – everyone is doing AI, we want to signal that we’re doing it too (maybe I’ll give some unsolicited advice on how companies and teams can experiment with AI more substantially in a future post). But on the other hand, it flies in the face of traditional product thinking that suggests users only care about outcomes technology can achieve for them, not about the stack a product is built on.
“I don’t care what alloy the fork is made out of. Tell me why it’s a better fork.”
Although most companies doing this are, in my experience, doing this mimetically – they see other companies doing it and understand that they should do it to – they are all actually engaging in a well-defined product marketing strategy called “Ingredient branding”.
Ingredient branding is what it sounds like – companies market a product by showing off an ingredient which their customers signals quality (along some dimension they care about – more on this later).
The textbook example of this strategy is Intel’s famous “Intel Inside” campaign. Intel made CPUs for computers. Intel didn’t actually need to convince PC buyers that their chips were good. The people that actually buy the chips are PC manufacturers (OEMs). But Intel’s Idea was that if consumers started to associate Intel chips with quality PCs, then they would kick off a virtuous cycle where:
- End users would buy PCs that proudly incorporated Intel chips
- OEMs would be incentivized to advertise their products as containing Intel chips (“Intel Inside”) because it was an efficient way to signal quality.
- More end users would become aware of Intel and the quality of their products
There are other examples of this strategy too. For example, Teflon-coated pans (spectacularly backfired when Teflon was discovered to be bad for us when heated) or Gore-Tex for waterproof clothing. Not that many in software, but we aren’t many years removed from an example: crypto / blockchain.
Just a few years ago we saw an explosion of tools advertised as “built on blockchain” or “the decentralized alternative to X”.
It turns out though, that most users didn’t care about this ingredient, because that phrase doesn’t communicate value propositions that resonate with most people. When I see “Built on blockchain” I think censorship-resistance, which most of the time I don’t care about (maybe I should).
In fact, you could argue that these ingredients batsignal attributes that users find actively distasteful. Many crypto applications were slower to use than traditional stacks, and often required clunky UX like downloading an external wallet to interface with the surface.
It would be like if snack food manufacturers decided to put Moringa powder in all of their products. Moringa is a gross-tasting (IMO) powder that has some purported (IMO promising) health benefits. Most eaters of CPG food don’t know what Moringa is, but pretty soon are going to learn that all of these “Morgina Inside” products taste bad. Even if there are some benefits to suffering through Morgina Mayhem flavor, the value proposition isn’t salient enough to consumers for the ingredient branding to work.
AI-powered is different, because there are several examples of massively popular consumer apps that have made the benefits of AI salient (even if their long-term value is unclear). Hundreds of millions of people have used ChatGPT and at least viewed if not created images with Midjourney. So AI-powered is shorthand for characteristics that users know they find interesting: raw natural language input and a remarkable ability to understand and respond to that input
Another difference between AI and Crypto is that the AI-powered ingredient marketing phenomenon has focused mostly on a single ingredient manufacturer: OpenAI. Every “GPT-powered” launch is actually just saying “OpenAI inside”.
While Intel kickstarted their campaign with marketing dollars, OpenAI did with an historically viral consumer product (ChatGPT). The OEMs in this case are the software providers integrating with OpenAI’s APIs. It’s worth dwelling on just how rare this is in software. Think about all the other services in the software stack that most consumers of software don’t care about:
- Infrastructure: “Built on AWS”, “Powered by Google Cloud”
- Database: “MongoDB Inside”
- Email marketing: “Your emails will be sent by Braze”, “Texts fulfilled by Twilio”
- Frameworks: “Brought to you by React Native”
If you’re evaluating a product you may come across the fact that an OEM uses these vendors. For example, if you’re the reads-the-fine-print type, you can find a list of data subprocessors in a company’s terms of service or privacy policy). But I’m willing to bet that in the vast majority of products they aren’t marketed as a critical ingredient that the buyer should consider when making a buying decision.
It’s not like these ingredients aren’t important or that these providers are operating in commoditized environments. Instead, it’s about consumer/buyer recognition. Going back to the snack food analogy, a lot of things we eat contain emulsifiers to improve texture. They have very different properties and trade-offs. But you probably don’t see “Mouthfeel powered by Soy Lecithin on packaging”. Because that means nothing to most people (in fact, many people are more aware of the negative health impacts of emulsifiers than the value they create).
I’m curious whether this surge of free marketing for OpenAI will result in their brand words becoming synonymous with the value of the category. OpenAI clearly doesn’t have a monopoly on the magic of LLMs or AI generally. But for so many people, ChatGPT is synonymous with AI. As a result, it’s possible that software companies will feel the need to brand their products not only as “AI-powered” but specifically as containing the variety AI that buyers are fiending for.
It’s interesting to consider where this ends. On the face of it, it’s a flywheel. Presumably the more end users that are exposed to “OpenAI inside” products, the more they’ll come to associate them with magic, leading more vendors to market their stuff as OpenAI-powered, requiring them to actually work with OpenAI. So what happens in the limit?
- OpenAI’s brand continues to strengthen, giving them a long-term competitive advantage over competitors. This could cause them to develop a long-term technological advantage too (e.g. from more human feedback on their models). Or it could just persist as a sales and marketing leg up.
- The advantage persists until OpenAI does something to mess it up (like the Teflon example – where the ingredient marketing began to backfire).
- The advantage gradually wanes over time as competitors’ models outpace OpenAI and “Intel Inside” becomes a marker of old/cheap instead of new/fresh. Kinda like how if you see a MacBook with “Intel Inside” today you’d think “Don’t the latest Macs use bleeding edge Apple Silicon?”
