Local only
Confidential values and work that must stay on hardware you control are processed locally.
Grant Analytics · Private AI
Useful AI without treating confidential work like ordinary cloud data.
A private AI setup for solo and small-firm lawyers, therapists, accountants, advisors, and other professionals who need practical AI with explicit information boundaries.
Local confidential layerDe-identification before cloud useHuman approval gatesDocumented controls
Private by architecture
Use the right processing path for the information in front of you.
The system applies functional handling rules before a model receives the task.
De-identify, then use cloud
Appropriate context is masked locally before an approved cloud model receives it. Restoration happens locally.
Public cloud use
Public or explicitly approved material can use a cloud model directly under the selected service terms.
The offer
A bounded implementation with training and handoff.
Private AI Setup · one time
From $2,000
Typical full setup range: $2,000 to $6,000. A limited number of founding-client projects may be scoped at $1,500 to $2,500 in exchange for structured feedback and an approved testimonial.
- Information-flow discovery
- Local and approved cloud paths
- De-identification workflow
- Evaluation and failure tests
- Training and operating guide
Managed support · monthly
From $300 per month
Typical support range: $300 to $800 per month, depending on tools, review frequency, updates, and service level.
- Updates and health checks
- Workflow tuning
- Evaluation review
- Support and documentation
What the work includes
The system, not a slide deck.
Discovery
Map information types, tools, tasks, users, and approval points.
Implementation
Configure the selected local and approved cloud components on hardware you control.
Evaluation
Test routing, masking, task success, and failure behavior with representative cases.
Training
Document the operating rules and teach the people who will use and review the system.
Fit call
Start with the information and workflow you need to protect.
A 15-minute call determines whether the architecture fits the practice, hardware, and operating needs.