Financial services
Member fund assistant for pension trustees
How can a pension trustee answer member fund questions with AI without member data leaving?
A member fund assistant answers member and staff questions about funds and holdings in English and Traditional Chinese. Every figure is traced to the trustee's own fund fact sheets, and each answer is limited to the member who asked. The assistant runs on the trustee's own GPUs, so member data does not leave the trustee.
At a glance
| Industry | Financial services |
|---|---|
| Who it is for | CIO, Head of Member Services, Chief Compliance Officer |
| Data that stays in place | Member records, holdings, fund fact sheets, questions and answers |
| Runs air-gapped | Yes |
| Languages in delivery | English and Traditional Chinese |
| Infrastructure | HPE Private Cloud AI, or HPE GPU servers in an HPE AI Factory design |
The problem
Pension trustees answer member questions on funds and holdings by hand, from years of fund fact sheets in two languages.
Public chat tools are ruled out. Member data cannot leave the trustee, every figure must trace to a source, and each answer must be limited to the member who asked.
How it works
- The trustee's fund fact sheets and holdings data are loaded into the Katonic Knowledge Engine.
- A member or a staff member asks a question through the channel the trustee chooses.
- The agent retrieves only the sources that person is entitled to see.
- It answers in the language of the question and links each figure to the document it came from.
- Every question, source and answer is logged for compliance review.
On HPE infrastructure
On HPE Private Cloud AI
HPE Private Cloud AI is HPE's turnkey private AI system, co-engineered with NVIDIA. Katonic AI is a partner in the HPE Unleash AI program. On Private Cloud AI, Katonic connects to the models the system serves and adds the agent platform on top. The trustee keeps its documents and member data on the system, and the assistant uses the models served there.
On HPE GPU servers
Customers who build their own estate on HPE GPU servers, as part of an HPE AI Factory design, get compute, storage and networking from HPE. Katonic supplies the software layer above the hardware: open-weight models served on the customer's GPUs, the AI gateway, the agent platform and governance. The assistant, its knowledge store and the models all run on GPU servers inside the trustee's own data centre.
More on both routes: Katonic AI on HPE.
Where it stands
In delivery for a pension trustee in Asia-Pacific, selected after a scored proof of concept.
Questions and answers
Does the assistant give investment advice?
It is built to answer factual questions from the trustee's own fund documents and to cite them. What it may and may not answer is set by the trustee through guardrails.
Which languages does it support?
The current delivery answers in English and Traditional Chinese, from source documents in both languages.
Where do the models run?
On the trustee's own GPUs. No member question or member record is sent to a public AI service.
How is accuracy checked before release?
Each version of the assistant is scored against a set of test questions with known answers before it is released to members.