Member-ops AI inside an existing health fund stack
CBHS · Embedded AI · AU health fund
Member services were still split across claims, policy, and inbox tools. Straightforward questions moved between teams, and AI remained in presentations rather than in the work.
We placed an agent in the systems they already used: retrieval over approved policy, suggested replies for first-line staff, and a human approval step before any response was sent.
Frontline staff keep the same consoles. AI prepares drafts. People decide what is sent. Repeat contacts fall.
31% faster first-response on routine member queries.
Context
Mid-size health funds do not have a budget to replace core claims, policy, and CRM systems, and they cannot wait for a separate AI platform. Member questions still arrive in the same inboxes, and staff still work in the same consoles.
The gap was not the absence of a chatbot on the marketing site. First-line teams were answering the same policy and claims questions by searching documents and previous tickets, while leadership had seen AI demonstrations that never reached those tools.
What we focused on first
We mapped the member journeys that already existed: eligibility, extras, waiting periods, and “where is my claim.” For each, we listed the source of truth the fund already trusted — PDFs, knowledge articles, and structured policy fields — and the people who must still approve the response.
The insertion was deliberately narrow: draft a first response, cite the source, and leave sending with a person. There was no new member portal and no replacement of the claims engine.
Delivery approach
We connected retrieval to the approved knowledge the fund already maintained, then presented suggested replies inside the existing ticketing view. Controls included allow-listed sources, a visible citation, and a stop on clinical or financial advice the model is not permitted to give.
Staff were trained on when to accept, edit, or ignore a draft. Exceptions — anything retrieval could not ground — stayed on the manual path. That kept risk inside the operating model they already ran.
Outcomes and how we measured them
Time to first response on the targeted routine categories fell by about 31% after the insertion stabilized, measured on the same queues and staffing levels. Repeat contacts for “I already asked this” fell as answers became more consistent with the written policy.
The operating change mattered as well. Member operations stopped treating AI as a side project. It became part of the work they already did.
What we would emphasise for similar programmes
Place AI in the queue you already have. Do not add a parallel chatbot that staff will not use. Ground every draft in sources the business already owns, keep sending with a person, and measure first response and repeat contact.
“We did not rip out our member stack. AI just started sitting in the queue with us — and the easy tickets stopped eating the day.”
Liked this insight?
See how these concepts apply to your business. Book a short fit call and we'll map a practical next step.