From GPT-3 Wrappers to Workflows
Cohere
OpenAI turning the base model into the product compressed the room for thin AI wrappers almost overnight. Jasper and Copy.ai had grown fast by taking GPT-3 output, packaging it into marketer friendly writing flows, and selling subscriptions on roughly 60% gross margins. Once ChatGPT let users paste in drafts and iteratively rewrite them in a simple chat box, the core job of those apps became much easier to get directly from the model maker, which pushed surviving startups toward deeper workflow products and left enterprise first labs like Cohere with a clearer lane.
-
Before ChatGPT, Jasper and Copy.ai were essentially productizing GPT-3 for copywriting, reaching about $72M ARR and about $10M to $11.6M ARR in 2022. Their wedge was not proprietary model tech, it was packaging, templates, and workflow around someone else's model.
-
After ChatGPT launched in November 2022, customers could swap a $50 per month writing tool for free ChatGPT or the $20 premium tier. The hit was strongest in prosumer and SMB writing use cases, because the buyer mostly wanted text generation and editing, not a deeply embedded business system.
-
The response was to move up the stack. Copy.ai repositioned from single asset generation into repeatable GTM workflows that research accounts, draft multi step outreach, and write back into Salesforce or HubSpot. That is much harder for a general chat app to replace, because the value sits in orchestration and system integration, not just text output.
This pattern keeps pushing model companies toward owning high frequency interfaces and pushing application startups toward workflow depth, compliance, and distribution inside existing systems. That is why Cohere's enterprise focus matters. As horizontal chat absorbs generic consumer writing tasks, the durable winners are either the labs with direct distribution or the software companies that turn models into repeatable work inside the enterprise stack.