Risks of Writer's multi-model approach

This risk is really about R&D leverage. Writer is trying to win both as an enterprise app company and as a model builder, which means every new vertical model adds another system to train, evaluate, update, and support while frontier model labs spread those costs across much larger revenue bases. That burden matters more because Writer was at an estimated $47M ARR in 2024, far smaller than adjacent enterprise AI rivals that can amortize model work across broader platforms or bigger deployment footprints.1 2 3

• Writer has expanded from brand safe writing tools into a full stack AI platform with specialized models like PalmyraMed for healthcare and Palmyra X5 for long context workflows. That broadens product surface area, but it also means more domain tuning, safety work, and maintenance for each model family.1
• A close comparison is Cohere, which also sells industry specific models, but does it from the model layer first. Cohere can reuse one base model stack across banking, healthcare, government, and private cloud deployments, and was at an estimated $150M ARR in October 2025 versus Writer at $47M ARR in 2024.1 2
• Glean shows the opposite design choice. Instead of owning a broad family of in house models, it uses multiple outside models inside a search and agent product, letting it focus engineering on connectors, retrieval, and workflow. That helped Glean reach an estimated $100M ARR in November 2024 while competing in the same enterprise AI budget pool.1 3

The likely direction is fewer companies training many separate enterprise models, and more companies either building one reusable core model or orchestrating third party models above the workflow layer. For Writer, the durable path is to make its proprietary models matter in the specific tasks where compliance, brand control, and speed inside enterprise workflows are valuable enough to justify the extra model complexity.1 2 3