boost.ai Launches Presis: A self-hosted AI model designed for regulated industries, balancing speed and accuracy
Classification fine-tuning models will now provide support for proxy routing, range determination, and security protection decisions on the boost.ai platform, maintaining stable performance during peak demand periods.
Sannes, Norway, November 10, 2026 / PRNewswire / -- Today, a global provider boost.ai focused on regulated industries and utilizing AI dialogue technology, announced the launch of Presis. This is a finely tuned self-hosted language model designed to handle classification decisions throughout every AI agent dialogue. These decisions include determining which part of the response should be provided by the AI agent, whether a question falls within the scope of handling, whether an answer is suitable for presentation to end-users, and how to conduct post-dialogue reviews.
According to internal tests, Presis is at least 2.5 times faster than the model it replaces. This difference is most evident during peak demand periods, as third-party computing capacity is most likely to slow down at such times; this point is particularly crucial in the deployment of voice AI, as routing and security checks must be completed before the response begins.
boost.ai CEO Jerry Haywood stated: "For regulatory authorities, every automated conversation is of great importance – speed and reliability are the core to building trust with customers. Presis reflects the experience we have accumulated over a decade of deep involvement in these industries, and it also gives us another way to earn the trust that customers place in us every day."
boost.ai indicates that over the past decade, the company has been deploying conversational AI within banks, insurance companies, and other highly regulated institutions. In these environments, incorrect or delayed responses can lead not only to inconvenience for users but may also trigger compliance issues. Based on this experience, the company has decided to provide customers with a model specifically designed for internal decision-making in each conversation, rather than relying on third-party models that are aimed at general scenarios and attempt to meet the needs of everyone.
boost.ai Chief Technology Officer Rasmus Hauch stated: "A model that is specifically tuned for classification tasks can achieve what general-purpose models cannot, including maintaining speed under peak loads. Moreover, since we control every aspect of its operational stack, it can continue to improve steadily. With the help of Presis, we are now able to provide faster and more consistent responses at critical points, while also integrating the reliability that customers expect from boost.ai more deeply into the platform."
Since Presis runs on an end-to-end controllable infrastructure provided by boost.ai, routing, scope, and security protection decisions are no longer exposed to potential downtime, timeouts, and capacity fluctuations that could affect third-party providers. Classification costs have also decreased: according to internal tests, the average token cost has been reduced by 33%; in addition, across the five classification functions it supports, the accuracy of Presis is higher than that of the replaced models, with improvements ranging from 0.5 percentage points to 7.9 percentage points depending on the function.
Presis will be gradually deployed to the boost.ai platform over the next few weeks. The company also stated that they are evaluating the use of similar methods for generative functions in the future.
About boost.ai
boost.ai indicates that the company is a provider of customer experience solutions for regulated industries, driven by AI. Its platform focuses on security, speed, and scalability, and achieves rapid deployment, high resolution in problem-solving, as well as comprehensive mixed control by seamlessly integrating traditional NLU with large language models. boost.ai states that the company has completed over 650 successful deployments, has 600 online AI agents, and has processed more than 200 million automated conversations, serving brands including Nordea, Credit Union, of, Colorado, Sage, DNB, Trading, 212, and others. boost.ai also mentions that the company was rated as a leader in Gartner's Gartner 2025 "Magic Quadrant for Conversational AI Platforms."
