Apps Capture Value as Models Commoditize
Dave Rogenmoser, CEO and co-founder of Jasper, on the generative AI opportunity
Cheap, abundant models shift the battle for value creation up into workflow, data, and distribution. For Jasper, that means the durable product is not the raw model, it is the layer that knows a company’s brand voice, sits inside the tools employees already use, and learns from what users keep, edit, and publish. As model prices fall and quality converges, customers pay for better outputs inside real work, not for access to a specific foundation model.
-
Jasper already operated this way early on. It started on top of GPT-3, then added fine tuned models and feedback loops from user actions like save, favorite, and copy. That makes the app layer more defensible than the model layer, because the product gets better from proprietary usage data tied to a specific workflow.
-
The clearest comparable is Copy.ai. It also moved from being a simple GPT-3 wrapper to running dozens of fine tuned models plus monitoring and A/B testing, while aiming to automate go to market work across CRM and campaign tools. The pattern is that winning apps orchestrate systems and decisions, not just text generation.
-
The market already showed why this matters. Jasper and Copy.ai initially grew fast with template driven writing tools, but ChatGPT crushed their prosumer businesses by giving users a cheap general purpose interface. Their response was to go enterprise, where integration, compliance, brand control, and workflow depth matter more than model access alone.
The next step is more apps training small, narrow models on their own customer data while swapping among foundation providers underneath. That does not eliminate the big model companies. It turns them into infrastructure. The companies that capture the most value will be the ones that own the daily workflow, collect the best feedback data, and turn AI into measurable business outcomes inside the enterprise.