The Great AI Unbundling

Was this newsletter forwarded to you? Sign up to get it in your inbox.

Before Excel came VisiCalc, a spreadsheet often known as the first killer app for computers. It transformed how professionals did their jobs and became so essential that people bought computers just to use it. In the 1980s, companies fought a full-on battle for spreadsheet dominance. Entrants like IBM’s Lotus-1-2-3 gained ground, and in 1985, when Microsoft released Excel, VisiCalc encountered fierce competition. With the release of Windows 3.0 in the 1990s, Excel became the dominant spreadsheet program. And we haven’t looked back since.

Excel is a good product because it’s easy for beginners to get started: All you have to do is start typing in a cell. But it became a great product because it’s also incredibly powerful: Expert users can do everything from complex financial modeling to data analysis and visualization to building video games.

Excel was originally designed for business users working in finance and accounting roles, but its versatility has made it a ubiquitous, general-purpose tool. It’s also become the source of startup ideas. Just find a customer who’s manually operating a process in Excel and build a SaaS app designed for them.

Stripe’s Patrick McKenzie summed up Excel’s generativity best when he wrote, “My favorite symptom of an unmet need for software is any Excel spreadsheet which is ever updated by one employee, sent to a second employee, updated, and then sent back. Every time that happens a SaaS angel gets its wings.”

The venture capitalist Tomasz Tunguz was the first to observe that over the last 15 years, Excel has been unbundled into a host of other apps, such as Asana, Looker, and QuickBooks. But this unbundling was only possible once Excel became ubiquitous enough for users to know they wanted a purpose-made alternative. In order for that to happen, Excel needed widespread adoption with power users who began using it for niche workflows that it wasn’t originally designed to support.

Once those workflows were created, the power users realized that parts of their workflow were inefficient, or that features were missing for their use cases. They felt a need for purpose-built tools—and that became the opportunity for B2B SaaS to develop into a $327 billion market.

Just like Excel spawned the B2B SaaS era, general-purpose chatbot applications like ChatGPT, Claude, and Gemini will spawn a new era of startups for a new generation of computer users. AI chatbots have the same blend of accessibility and power as Excel did. They’re also trending toward the same level of general-purpose ubiquity—in far less time. 

Those who use AI today are becoming familiar with the atomic units of AI-based software: prompting, context windows, few-shot learning, and multi-modality. They’ll use ChatGPT or Claude to create workflows that work for their niches. In the process, they’ll develop opinions about how those workflows could be better, easier, cheaper, faster, and more secure.

That creates the opportunity to unbundle those workflows into a totally separate app. As LLM adoption increases, so, too, does the opportunity for startups.

I’ve seen this first hand. About a month ago we released Spiral, an AI app that automates repetitive creative work like X posts, LinkedIn posts, headline creation, and product release notes. It already has more than 3,500 users and is growing every day.

In essence, Spiral is a prompt builder. It uses Claude on the backend, so, technically, anything you can do in Spiral, you can do in Claude. 

But Claude’s chat interface isn’t built for the tasks that Spiral does well. In Claude, it’s difficult to keep track of the complex prompts you need to input in order to get good results. It’s also hard to share prompts you’ve discovered with your team. 

Spiral is a better interface for a specific kind of workflow: creatives and business people looking to translate their content into different mediums. But it could only have become popular now because people have had a year or two of getting used to prompting ChatGPT and Claude for similar tasks. They’re ready for something purpose-built.

The term “AI wrapper” has been a pejorative term. The implication is that no one is going to use a purpose-built AI tool when they can use a chatbot like ChatGPT or Claude for the same result. 

But the opposite is true. ChatGPT and Claude increase demand for wrappers as people use them to find use cases that a general-purpose tool isn’t designed to accommodate.

In fact, rather than ChatGPT and Claude steamrolling startups, they are great places to find startup ideas.

If you want to build a startup in the AI era, just watch how you use ChatGPT or Claude. If you’re doing something repeatedly—and the outputs are impressive—there’s a chance it could be its own app.