In January 2026 the Privacy Commissioner for Personal Data opened compliance checks on 60 Hong Kong organisations. The results published in May found that 95% were using AI in day-to-day operations. Almost none of them could say precisely where the documents they fed into it ended up. That gap is why private AI knowledge bases moved from a niche request to a board agenda item this year, and why most enterprise buyers now start with a demo rather than a trial.
What is a private AI knowledge base?
A private AI knowledge base turns your existing documents into a question-and-answer system that runs entirely inside infrastructure you control. Unlike a public AI subscription, the documents, the index and the computation stay on your own servers or a dedicated private cloud, so nothing leaves your environment.
UD's product in this category is Private AI BizHub. The buying decision it answers is narrow and specific: can we let staff ask questions of ten years of contracts, SOPs and quotations without any of it reaching a public model?
How much does a BizHub demo cost?
The demo is free and booked through a form on the product page. There is no card, no subscription and no trial clock. UD's published limited-time offer adds an enterprise cloud security assessment valued at HK$10,000 for booking the session.
What is not published is a list price for the deployment itself. UD quotes BizHub per engagement because the cost depends on document volume, whether the workload runs on your servers or UD's private cloud, and how much GPU capacity the deployment needs.
If your procurement process requires a rate card before a first meeting, say so at booking. It is a fair question and it has a fair answer, but you will not find the number on the website.
What does the demo actually cover?
The session walks through the four capabilities the product page names, applied to your own use case rather than a generic script. Expect a working demonstration rather than a slide deck, and bring a real document set if you can.
--- Knowledge base construction. Word, PDF and Excel files are imported directly. The system indexes them and identifies usable content, turning SOPs, product specifications, meeting minutes and contracts into a queryable base without manual tagging.
--- Standardised management. Every employee sees the current, approved version according to their own AI permissions, which is the control that stops version drift across departments.
--- Asset activation. A decade of quotations, tender documents and transaction records becomes retrievable for analysis, so management can pull comparable past cases in seconds.
--- Private computation. The deployment sits on internal servers or a private cloud with dedicated GPU infrastructure. Data is processed in an isolated environment rather than sent to a public endpoint.
Which AI models can a private deployment use?
BizHub supports switching between leading commercial models including ChatGPT, Claude Opus, Gemini and DeepSeek, selected per workload. Data analysis, product research and internal enquiry each reward a different model, and being locked to one vendor is a cost you pay quietly every month.
Model choice also matters for continuity. Providers retire models on their own schedule, and an architecture that assumes one specific model is a dependency your board should know about. UD has covered that risk separately in what AI model deprecation means for enterprise continuity.
Private AI BizHub at a glance
Cost and access
--- Demo: free, booked via form on the product page.
--- Included with booking: enterprise cloud security assessment valued at HK$10,000, limited-time offer.
--- Deployment price: quoted per engagement, not published as a list price.
--- Card required: no.
Deployment and data
--- Location: your internal servers or a private cloud on dedicated GPU infrastructure.
--- Storage: 100% independent, data does not leave the environment.
--- Inputs: Word, PDF and Excel documents, indexed automatically.
--- Access control: per-employee AI permissions.
--- Models supported: ChatGPT, Claude Opus, Gemini, DeepSeek, switchable per use case.
Vendor facts
--- UDomain has operated in Hong Kong since 1998, 28 years.
--- 20,000+ business clients, government-recognised cloud provider.
--- 20+ years of network security and data centre operating experience.
Where Private AI BizHub is not the right choice
An honest answer serves you better than a flattering one, so here is where this product loses. Four situations where a different route is the better decision.
You want self-serve, per-seat SaaS. BizHub is a deployed system on dedicated infrastructure, not a sign-up-and-go subscription. If your requirement is twenty licences by Friday, a public enterprise plan will get you there faster.
Your procurement needs a published rate card. There is no list price on the site. Organisations that must compare three published prices before a meeting will find that process harder here.
Your documents are not in order. A knowledge base inherits the quality of what you import. If your SOPs are three versions out of date, the AI will confidently repeat the outdated version. Clean-up is work you own, not work the demo removes.
You need frontier model research or custom fine-tuning. BizHub orchestrates leading commercial models rather than training its own. If your requirement is a bespoke model trained on proprietary data, a research lab or specialist ML consultancy is the right conversation.
One more caveat worth stating plainly: the HK$10,000 security assessment is described as a limited-time offer with no published end date. Confirm it is still running when you book, and get the inclusion in writing.
Who should book the demo, and what should they bring?
The session is most useful for organisations with a real document problem and a real compliance constraint. If both are true, the demo answers in an hour what a written proposal takes three weeks to approximate.
--- Best fit: 50 to 500+ staff, regulated or contract-heavy sectors such as financial services, professional services, logistics and property management, with document archives that predate anyone currently in the role.
--- Bring: a representative document set, your data residency requirement in writing, and the three questions your team currently cannot answer without asking a specific person.
--- Ask: where the index physically sits, who can delete an entry, how permissions map to your existing directory, and what happens to the knowledge base if you end the engagement.
That last question is the one most buyers forget and the one that matters most eighteen months in.
The next step
A demo is a low-cost way to convert an abstract governance worry into a specific technical answer. You either see your own documents answering your own questions inside your own environment, or you learn quickly that the fit is wrong. Both outcomes are worth an hour.
The organisations that get this right are not the ones with the biggest budgets. They are the ones who asked where the data sits before they asked what the model can do. We understand AI. We understand you. With UD by your side, AI never feels cold.
Reviewed by the UD enterprise AI team. Product facts, offer and pricing position verified against the Private AI BizHub product page on 26 August 2026. Offers change, so confirm current terms at booking.
Book the session
Bring your documents and your compliance constraint, and we'll show you what a private knowledge base looks like on your own data. We'll walk you through every step, from the demo to deployment, access design and ongoing operation, backed by 28 years serving Hong Kong enterprises.
