Deployment
Your environment. Your controls. The same Percidian.
A Percidian-managed environment, or your own cloud boundary. The same understanding, memory, permissions, integrations, and delivery either way — what changes is who holds the controls.
Percidian-managed
- Understanding
- Memory
- Permissions
- Integrations
- Delivery
Isolated environment · operated by Percidian
Customer-controlled cloud
- Understanding
- Memory
- Permissions
- Integrations
- Delivery
Your account, network, keys, and models
Choose the operating boundary
Who holds each control in each boundary.
Data, encryption keys, network, model endpoints, and operations — who holds each in a Percidian-managed environment and in your own cloud.
| Isolated environment, Percidian-operated | Your cloud account and storage | |
| Key model selected per deployment | Customer key control | |
| Percidian’s isolated boundary | Your network boundary; controlled outbound | |
| Customer-approved endpoints only | Customer-approved endpoints only | |
| Percidian operates and scales | Your infrastructure; Percidian manages the platform |
Managed is the fastest path; customer-controlled is for infrastructure, data-boundary, or model-routing requirements. Both run the full product.
Select a control to compare it across both boundaries, or a boundary to read it top to bottom.
Governance in either boundary
The controls that hold either way.
The controls a security review asks for, and how each holds in either boundary.
- Source-aware permissionsSource access decides every request. Connecting a system grants no one new access. How the decision is madePer request
- Enterprise isolationCustomer information and institutional memory stay inside the customer boundary.Per customer
- EncryptionIn transit and at rest; customer key control where the selected deployment requires it.TLS 1.2+ · AES-256
- Evidence and reviewRequests, access decisions, and delivered context stay traceable to source.Every request
- Retention and deletionDeployment-specific policy; source deletion and retention reach derived context.Follows the source
- Approved destinationsOnly endpoints you approve receive context. Customer data never trains third-party models. What an unapproved destination receivesNo third-party training
Your keys. Your infrastructure. Your model endpoints.
In the customer-controlled boundary, Percidian runs inside your environment; data, keys, and model endpoints remain under your control.
Bring your AI strategy
Bring your models. Keep your memory.
Percidian is the understanding layer beneath the AI systems your organization approves. Use customer-selected model endpoints and change the model serving a workflow without rebuilding the business context, permissions, or source connections behind it.
- Model independence
- Persistent institutional knowledge
- Destination policy
- Workload-specific model choice
- One governed business understanding
Deployment journey
From first outcome to enterprise coverage.
The foundation is built once and reused at every step after it.
Define
First outcome, users, sources, AI surfaces, security requirements.
Connect
Enterprise identity and approved source access.
Understand
Business foundation, domain modules, your organization’s model.
Validate
Representative users, questions, evidence, permission changes, destinations.
Activate
Context into approved assistants, agents, applications, workflows.
Expand
More teams and use cases on the same foundation.
Your systems of record stay authoritative. Start with one use case; expand on the same foundation.
Percidian does not replace existing systems: it connects their information, relationships, permissions, and history into business understanding for AI. Begin with a defined domain, team, and outcome, then reuse the same foundation across more sources, workflows, and AI systems.
Deployment request
Design the right Percidian environment.
Bring your business into every AI decision
Give every approved AI system the context to do its best work.
Connect the knowledge your organization already has. Preserve what it learns. Apply the permissions it already trusts. Make that understanding available wherever people and agents work.