ENTERPRISE AI, WITHIN COMPANY POLICY

Do not block AI.Put it insidecompany policy.

Q: What does AgentDeck Enterprise govern?

It manages members, AgentDeck seats, approved AI engines and cloud routes, regions, permission ceilings, devices, usage and security-policy events from one organization console.

Q: Can administrators read employee prompts and responses?

No. Administrators can see operational metadata such as the member, route, model family, time, usage, policy revision and enforcement result, but not prompts, responses, files or command content.

Q: Can employees keep existing Claude, ChatGPT and Gemini team subscriptions?

Yes. The company can manage seat assignment while each employee signs in directly with the provider on their own device. AgentDeck does not centralize provider passwords, cookies or tokens.

Let people keep the AI and subscriptions they already use. Before data leaves the device, protect personal and business context, then allow only approved accounts, models and regions.

Request a pilot See the real console

The screens below are captured from the actual AgentDeck Enterprise admin console with representative organization data.

AgentDeck Enterprise organization overview — ACTUAL PRODUCT SCREENOrganization overviewSeats, policy revision, retention and privacy principles at a glance.

9-language admin console Signed policy, fail-closed enforcement Metadata-only audit Provider credentials stay off the control plane

PEOPLE AND SUBSCRIPTIONS

A company seat is assigned centrally. Each person still signs in directly.

Manage members, roles and company-purchased Claude Team or OpenAI Business seats. The console stores assignment state—not passwords, cookies or provider tokens.

AgentDeck Enterprise members and invitations — 01Members and rolesInvite members, separate owner, admin, auditor and member access, and revoke cleanly.

AgentDeck Enterprise subscription seat assignments — 02Subscription seat poolAssign and reclaim company seats while sign-in remains on each member’s device.

Operational boundary: reclaiming an AgentDeck seat stops new organization sessions. Cancel or revoke the upstream provider seat in that provider’s own admin console.

SIGNED POLICY

The stricter rule always wins.

Set ceilings for permissions, engines, security checks, workspace roots, MCP destinations, remote shell, versions, model families, regions and usage. Personal settings can be stricter, never looser.

01 · PERSONAL DATA

Keep personal data from leaving as-is

Supported names, contact details and identifiers are protected on the device before sending, then shown normally again in the response.

02 · BUSINESS CONTEXT

Protect customer and project names by company rule

Your team chooses which customer, product and project terms to hide, so protection follows the work instead of relying on employees to remember every rule.

03 · EXTERNAL CONTENT

Do not let a document quietly take control

Web pages and attachments stay reference material. A hidden instruction cannot turn into file access or command execution without the allowed scope and approval.

04 · CHECK FAILURE

If protection cannot be checked, do not send

A required check never becomes “send anyway.” AgentDeck stops the request, explains why and leaves the decision to the right person.

Protect before sending Restrict where it can go Never grant execution authority to external content

AgentDeck Enterprise policy editor — 03Policy editorPermission ceilings, security mode, workspace boundaries and connectors in the actual console.

Signature verifiedDesktop enforcement accepts only signed, monotonic policy revisions.

Safe on failureA device that has never received policy starts from the safest defaults.

Local security gateRequired PII and context checks block the turn if they cannot complete.

Pseudonymization and masking reduce exposure before an approved provider receives the request. They do not automatically remove every privacy or cross-border transfer obligation.

VISIBILITY WITHOUT SURVEILLANCE

See operating risk and cost. Keep the work private.

Audit policy decisions and security events, then understand usage by member and model family. Conversation text and local work content are not collected.

AgentDeck Enterprise metadata-only audit log — 04Metadata-only auditTime, user, tool and policy decision—never the conversation or command content.

AgentDeck Enterprise usage by member and model family — 05Usage and limitsCompare members and model families with explicit estimated-cost labeling.

ADMIN CAN SEE

Seat, user ID, provider route, model family, time, usage, estimated cost, device posture, policy revision and enforcement event.

ADMIN CANNOT SEE

Prompts, responses, conversation history, files, command content, provider credentials, profile directories or local project paths.

AgentDeck Enterprise organization AI providers — 06Organization AI routesManage Bedrock, Vertex AI and Azure AI Foundry configuration and allowed model families.

CLOUD CHOICE AND DATA RESIDENCY

Bring the cloud relationship your organization already trusts.

Connect AWS Bedrock, Google Vertex AI, Azure AI Foundry and explicitly approved compatible gateways. Members use organization SSO or IAM on their device; long-lived keys are not distributed through policy.

Allow only registered routes and canonical model families
Pin approved regions and model versions
Show unknown model cost as unknown—not a false estimate

DEVICE AWARENESS

Detect what needs attention. Revoke what should no longer connect.

Track app version, platform, presence and policy state, then cut relay and push access for a lost or retired device. These are actionable signals—not a claim of hardware remote attestation.

AgentDeck Enterprise registered devices — 07Registered devicesVersion, platform, last contact, status and immediate access revocation.

BEFORE YOU ROLL OUT

The questions security, IT and employees ask first.

What does AgentDeck Enterprise govern?

It manages members, AgentDeck seats, approved AI and cloud routes, regions, permission ceilings, devices, usage and security-policy events from one console.

Can administrators read employee prompts and responses?

No. Administrators see operating metadata such as the member, route, model family, time, usage and policy result—not prompts, responses, files or command content.

Can employees keep existing Claude, ChatGPT and Gemini team subscriptions?

Yes. The company manages seat assignment while each employee signs in directly with the provider on their own device. Provider passwords, cookies and tokens are not centralized.

How is customer data protected before it reaches an external AI?

Supported personal data and company-defined customer, product and project terms are protected on the device before transfer. Required checks block the request if they cannot finish.

Does AgentDeck Enterprise guarantee data-residency compliance?

It can restrict approved providers, model families and regions and reduce exposure before transfer. Legal obligations and provider contracts still require review during rollout.

ADOPT AGENTS WITHOUT LOSING THE BOUNDARY

One place to govern access. No place to browse private work.

Bring a team, its existing provider seats and its cloud accounts. We will map the rollout boundary before the first organization policy is published.

Request a pilot Read the technical case study