Government and public sector
Citizen tax assistant
How can a tax authority answer citizens with AI that cites its own rulings?
A citizen tax assistant answers taxpayers' questions on the agency's own website, using only the agency's published rulings and citing the one it used. Each version is scored against the agency's test questions before release, and no citizen question leaves government infrastructure.
At a glance
| Industry | Government and public sector |
|---|---|
| Who it is for | Head of Citizen Services, tax administration leads, contact-centre managers |
| Data that stays in place | Citizen questions, answers, rulings and logs |
| Runs air-gapped | Yes |
| Sources | The agency's published rulings and guidance only |
| Infrastructure | HPE Private Cloud AI, or HPE GPU servers in an HPE AI Factory design |
The problem
Tax authorities field the same citizen questions by phone and email, and the answers are buried in rulings and guidance.
A public chatbot that invents an answer is worse than none, and citizen questions cannot be sent to a public AI service.
How it works
- The agency's published rulings and guidance are loaded into the Knowledge Engine.
- A citizen asks a question on the agency's website.
- The assistant answers only from those rulings and cites the ruling it used.
- Before any version is released, it is scored against the agency's own test questions.
- Every question and answer is logged for 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 assistant uses the models served on the system inside the government data centre.
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 runs on government-owned GPU servers, alongside other agents for the same agency.
More on both routes: Katonic AI on HPE.
Where it stands
Demonstrated in October 2026, using a fictional ministry. A recording is available on request.
Questions and answers
What happens when the rulings do not cover a question?
The assistant is set to answer only from the agency's published rulings. Where they do not cover a question, it says so instead of guessing.
How does the agency know the answers are right?
Each version is scored against the agency's own test questions before release, and every answer cites its source.
Do citizen questions go to a public AI service?
No. Questions and answers stay on government infrastructure.
Has this been delivered for a tax authority?
It was demonstrated in October 2026 with a fictional ministry. Ask us for the recording.