A new artificial intelligence startup is addressing the frustrating experience patients can have when calling a health plan to understand their benefits.
With Trampoline AI, health plans, third-party administrators and supplemental benefit companies can upskill their human call center agents with AI. Trampoline leverages large language models to help call center agents address the swath of members that don't get their questions answered on their first call to their plan.
The company has been incubated at Redesign Health since November 2023, and it made its public debut with its first two customers HealthCheck360 and Alliance Health on Thursday.
Though healthcare call centers increasingly try to upskill their agents with tools to help customers, agents have lacked easy access to member benefit information, real-time quality assurance checks and summaries of member sentiment across the call center.
Similarly to AI copilots that surface insights for clinicians at the point of care, Trampoline AI pulls relevant member benefit information and prompts questions for call center agents to ask members during their interactions.
Trampoline’s products will also provide quality assurance checks of every call and offer coaching feedback to agents in real time. The generative AI writes notes from the interaction and documents them in the call center system.
Call center systems are often complex and hard for agents to navigate. Despite some advancements in call center technology, co-founder and CEO Mike Bourke explained that calls to health plans still take too long to identify the member and their plan, let alone answering their question.
“[Agents] put people on hold. They put them on mute. They spend a ton of time looking for answers and doing keyword searches and things across lots of different systems," Bourke explained. "And with these conversational AI tools and RAG, Retrieval Augmented Generation, techniques, we can now go to the right source, find these answers, and bring them back to the agents.”
Trampoline AI allows call center managers to analyze trends in the center's calls. Managers could understand the average caller sentiment, if the caller has called before, if the call was an audit or if the caller asked to speak to a supervisor. This level of analysis was not previously able to be done before LLMs, Bourke said.
“These conversational AI tools allow [Trampoline] to be in the conversation and to know what's happening in every step of the of the way so to do things like listening in and making sure that the agent is following the right process, and they know the right steps to take to stay in compliance with the process that answers that question,” he said.
Trampoline’s first two customers are HealthCheck360 and Alliance Health.
HealthCheck360 offers program management and incentive benefit design to employer-sponsored health plans. Trampoline is specifically working with the company’s new health advocacy group, which the company is trying to scale quickly.
“By helping them with the process, helping them with information, helping them write up their notes for them, we're really helping to shorten that training time and make their agents and this team really effective in providing that next level of service for their members,” Bourke explained.
With Alliance Health, Trampoline is focused on bolstering compliance and quality assurance reviews ahead of potential government audits of their Medicaid benefits.
“For them, it's following the right steps in compliance … so that when they get audited on any of these government plans, that they know ahead of any audit exactly where things are going to fall, and so they're getting ahead of any potential issues that could happen,” Bourke said.
In the rest of 2025, Trampoline AI hopes to contract with more health plans and supplemental benefits companies. It eventually hopes to be able to divert easier low calls like getting a new copy of an insurance card to a fully automated agent. That way, human agents can handle more complex issues.
“They can focus on that human connection, and have the AI do the mundane work of looking at the answers, writing their notes and all the other stuff, so humans can do what humans do really well,” Bourke said.