Value Shifts From Models to Workflows

This line matters because it means the model itself stops being the scarce asset, and shifts value to the company that best turns cheap intelligence into a sticky workflow. Jasper already treated the model as interchangeable infrastructure, routing different actions to different models, layering prompt design and fine tuning on top, and focusing on marketer specific outputs rather than owning the base model. As model access gets cheaper and more open, that product layer matters more than raw model ownership.

• Jasper was built as a layer on top of foundation models, not as a lab training a frontier model from scratch. By late 2022, it was already using a mix of vanilla GPT-3, fine tuned variants, and some open source or in house models for specific actions, which is exactly the behavior expected in a commoditizing base layer.
  1 sacra 2 sacra
• The economic logic is simple. If many apps can rent similar models, and open models can be downloaded, tuned, and run locally, then model vendors face price pressure while app companies can swap suppliers. That is why Jasper compared model access to cloud infrastructure and why open model hubs became important distribution for alternatives to closed APIs.
  1 sacra 3 sacra 4 sacra
• This also explains why the first wave of AI writing apps was fragile. Jasper and Copy.ai reached roughly $72M and $10M ARR in 2022 by wrapping GPT-3 in templates and workflow UX, but ChatGPT later flattened the prosumer market by giving users a cheaper general purpose interface. When the base intelligence gets cheaper, shallow wrappers get squeezed first.
  2 sacra 3 sacra

Going forward, the winners in generative AI writing will look less like model vendors and more like workflow software. Jasper’s path is to own brand voice, approvals, integrations, and embedded usage inside the tools where work already happens. In a world of abundant low cost models, the durable advantage comes from being the system a team actually works inside every day.