Workflows, Not Models, Drive AI Value
Dave Rogenmoser, CEO and co-founder of Jasper, on the generative AI opportunity
The durable value in generative AI shifts upward from the model to the workflow. Jasper’s bet is that once good models are broadly available, the winner is the product that turns raw text generation into a repeatable business outcome, like helping a marketing team draft blog posts, ads, emails, and on brand copy inside the tools they already use, with better prompts, templates, feedback loops, and integrations than a generic model alone.
-
Jasper already treated the model layer as interchangeable. Different actions inside the product route to different models, sometimes multiple in sequence, and Jasper fine tunes on user ratings and template level feedback. That means the moat is not owning one big model, it is knowing which model to use, where, and how to shape output for a specific job.
-
This also explains why margin pressure cuts both ways. Jasper described itself as having normal SaaS margins, with model costs falling as the foundation layer commoditizes. But cheaper models also lower barriers for rivals, so the defense is shipping a better application, not relying on exclusive access to underlying AI.
-
The market validated both the opportunity and the risk. Jasper grew to about $75M ARR by the end of 2022, but later AI writing tools saw their SMB and prosumer base hit by ChatGPT. The surviving path was to go deeper into enterprise workflows, where brand controls, integrations, and team specific use cases matter more than raw text generation.
Going forward, AI apps that look like thin wrappers around a public model keep getting squeezed. The stronger businesses are likely to look more like workflow software with AI built into every step, embedded across apps, trained on company context, and judged on whether they save labor or drive revenue, not on which base model they call underneath.