Platform version 14 moves boost.ai's core classification features onto Presis, its own native language model, replacing the mix of GPT-4o and GPT-4.1 models those features used before. Agent orchestration, few shot predictions, out of scope detection, the hallucination guardrail and AI review switch to Presis automatically on update, and customers can no longer choose a different model for them. Boost.ai says the move lowers token cost and latency by taking a third party provider out of the path.
Our readBanks and insurers running boost.ai assistants now have a vendor built model, not an OpenAI model, making routing and guardrail decisions, and bring your own model is no longer available for those features. Model inventories and third party risk files need updating before the version 14 upgrade.