The Pulse
Boost.ai introduces Presis model for agent routing and safeguards
Boost.ai announced Presis, a self-hosted model that classifies conversations and supports routing, scope and safeguard decisions. The company says internal tests found it faster and less costly than the models it replaces.
AI.info Team · Written with AI · fact-checked ·
A model for decisions around each conversation
Boost.ai has introduced Presis, a fine-tuned, self-hosted language model for classification decisions within and around AI-agent conversations. The model selects which part of an agent responds, checks whether questions fall within scope, determines whether answers are safe to show end users, and supports review of conversations after they end, the company said.
Boost.ai said Presis now powers routing, scope and safeguard decisions across its platform. The company described the model as a dedicated system for classification, rather than a general-purpose model built to handle a wider range of tasks.
Company reports faster results and lower costs
According to the company, internal tests found Presis at least 2.5 times faster than the models it replaces. Boost.ai said the speed difference is most noticeable at peak demand and in voice deployments, where routing and safeguard checks must finish before a response begins.
The company also reported an average 33 percent reduction in token costs in internal testing. It said Presis was more accurate than the replaced models on all five classification features it powers, by 0.5 to 7.9 percentage points depending on the feature.
Infrastructure and regulated customers
Boost.ai said the model runs on infrastructure it controls end to end, putting those decisions outside the reach of third-party providers' outages, timeouts and capacity swings. CTO Rasmus Hauch said the company can improve the model over time because it controls the stack on which it runs.
The company said its decade of work with banks, insurers and other highly regulated organizations informed its decision to build a dedicated model for these tasks. CEO Jerry Haywood said that for regulated organizations every automated conversation matters, and that speed and reliability are core to how they build trust with their customers.
Rollout planned for the coming weeks
Boost.ai said Presis is due to roll out across its platform in the coming weeks. The company is also evaluating a similar approach for generative features in the future.