Shadow AI risk
Client matter numbers and IP in ChatGPT logs create confidentiality incidents waiting to happen.
On-premises · Australia · GST inclusive installs
AI for firms that take confidentiality seriously
SovereignDesk installs a complete private AI platform in your office — chat, session-based document intelligence, engineering CAD, and draft-only admin agents — so your team works faster without sending client data to US cloud providers.
Legal & accounting
Engineering consultancies
Defence supply chain
5–25 users
Why firms switch from public AI
100%
Inference on your hardware
6–10 wks
Typical install timeline
0
Client files sent to OpenAI
Full
Audit trail & SSO
Your staff are already using AI. Just not safely.
Every week, employees paste contracts, designs, and client emails into public tools. SovereignDesk gives them something better — the same speed, on hardware you control.
SovereignDesk answer
All inference runs on your LAN. Optional air-gap. No training on your documents.
One platform
Replace fragmented chat, doc search, ad-hoc CAD experiments, and risky email drafts with one approved system.
Everything your team needs. Nothing leaving your network.
Three GPU tiers work together — fast chat, deep analysis, and multimodal generation — routed automatically to the right model.
Smart model routing
Quick questions hit a fast LLM. Contracts, tenders, and engineering tasks route to a deep-reasoning model — automatically or by user choice.
Session workspaces & RAG
Each matter or project gets its own folder. Upload PDFs and specs; ask questions grounded in your files — not the public internet.
Engineering CAD suite
Describe a part in plain English → Fusion 360 script → STEP export. Engineer sign-off built into the workflow.
Image & document generation
Local image generation, large-document parsing, and report drafting from your session corpus.
Draft-only agents
Email, calendar, and requirements agents draft work for human approval — nothing sends without a click.
Enterprise security
Microsoft Entra ID SSO, audit logs, VLAN segmentation, and on-premises storage on Synology NAS.
Choose the system that fits your firm
Installed price includes hardware, software, on-site deployment, and training. Annual support keeps models and security current.
Essential
Private chat & documents for small teams getting off public AI.
$52,800 inc GST installed
+ $9,600/yr support
- Fast LLM for everyday chat & drafts
- Session folders with document RAG
- Large document parse (to 200 pp/file)
- Basic fast/deep routing
- 1 day on-site training
- Deep-think GPU server
- CAD / Fusion suite
- Email & calendar agents
Professional
Full private AI for privacy-sensitive firms with up to 10 daily users.
$98,500 inc GST installed
+ $18,000/yr support
- Fast + deep-think LLM servers
- Full automatic model routing
- Session RAG + image generation
- Large docs (to 500 pp/file)
- Fusion 360 text-to-CAD pipeline
- Email, calendar & requirements agents
- Microsoft Entra ID SSO
- 2 days on-site training
Enterprise
Redundant GPU, 25 users, and dedicated CAD workstations for larger practices.
$168,000 inc GST installed
+ $36,000/yr support
- Dual fast + dual deep GPU (redundant)
- Org-wide document search
- Unlimited batch document processing
- 2× dedicated CAD workstations
- Priority deep-model queue
- Redundant PSU & gateway
- Enterprise SLA (2 hr response)
- 3 days on-site training
| Capability | Essential | Professional | Enterprise |
|---|---|---|---|
| Concurrent users | 5 | 10 | 25 |
| Deep-reasoning model | — | ✓ | ✓ ×2 |
| Session RAG folders | ✓ | ✓ | ✓ + org-wide |
| Fusion CAD suite | — | ✓ | ✓ + CAD WS |
| Admin agents (draft-only) | — | ✓ | ✓ |
| Image generation | — | ✓ | ✓ |
| SSO (Entra ID) | Optional | ✓ | ✓ |
| GPU redundancy | — | — | ✓ |
Start with the ebook. Train the team for $9.99.
Free executive governance guide → low-cost video fundamentals → specialist tracks when you're ready.
Applied AI for Business & Corporations
Governance, technology, and practical pathways — free Amazon ebook for directors and GMs. Preview the web edition on site.
Download free ebook →Applied AI Fundamentals
44 video lessons that turn the ebook into team literacy — ML, LLMs, agents, certificate on pass.
Get course — $9.99 →From signed order to trained team
We handle hardware, rack install, model deployment, and handover — typically 6–10 weeks for Professional tier.
1
Discovery
Site survey, network design, SSO plan, and rack location.
2
Staging
GPU servers configured in our lab while hardware is in transit.
3
Install
On-site rack, power, VPN, vLLM, RAG, portal, and agents.
4
Train & handover
On-site training, UAT sign-off, 30-day hypercare support.
Built on commercial hardware you own
No vendor lock-in to a US API. Standard servers from Supermicro, Dell, and Synology — fully auditable and air-gap capable.
[ Your team ] → VPN / Office LAN → SovereignDesk Portal (SSO)
│
┌────────────────────┼────────────────────┐
▼ ▼ ▼
Fast LLM Deep LLM Multimodal
RTX 4090 RTX 6000 Ada RTX 4090
Llama 8B Qwen 32B Flux · Docs · CAD
│ │ │
└────────────────────┴────────────────────┘
▼
Synology NAS (sessions & RAG)
Qdrant · PostgreSQL · Audit log
See it working with your documents — not a slide deck
Book a demo at our Perth lab or request an on-site walkthrough for metro Australian firms.
Book a demo Download package summary