Government
Case work, policy analysis and correspondence — agents that read the entity's own regulations, circulars and case systems.
- Review this case, collect the related correspondence and policies, and prepare the decision package.
- Review this procurement package against the approved policy.
- Which circulars changed the requirements for this licence in the last two years?
راجع هذه القضية، واجمع المراسلات والسياسات ذات الصلة، وجهّز حزمة القرار.
The use case, on film
The agent pack
What it is made of
- Regulations and circulars
- Correspondence archive
- Case-management connector
- Approved government policy
- Human approval before a decision is issued
- Host workers (entity servers)
The examples below describe potential use cases. Execution depends on available skills and connectors, customer permissions and qualification in the customer environment.
What the entity is asking for
A case arrives with its attachments, two years of correspondence and three circulars that changed the rule since the last decision of its kind. An officer has to read all of it before the decision package can be written, every case is different, and the deadline is the same for all of them.
- Assemble a decision package from the case, its correspondence and the policy that applies today.
- Say which circular changed a requirement, when it changed, and what the requirement was before.
- Check a procurement package against the approved policy and list what is missing from it.
- Explain a service interruption on the entity’s own servers, with the evidence behind the explanation.
How the work is done
An agent is a model, a set of approved skills, tools and a policy, applied to a scope.
Agents reach your systems through the MCP servers you connect in the console. A token you enter there is sealed, never shown again, and acts only with the permissions your system gives it; a connector that runs on one of your hosts keeps its token in that host's own connector file.
Execution follows the configured host policy. A command can be approved, denied or referred for human review; individual approval depends on that policy.
The case system, the correspondence archive and the public services all run on servers the entity already operates. Those servers are where an agent starts, and the IT questions are the ones it answers first.
Read-only execution policy is the default on every enrolled host; the full policy is an explicit, per-host choice.
Host investigations show the commands that ran, their exit codes and output, per host.
A host joins the fleet with one install command and an enrollment token; revoking the host unpairs it.
Customer-hosted installations support an identity provider through OIDC with PKCE. Hosted workspace SSO must be confirmed for the agreed deployment.
Customer-hosted deployments use packaged deployment configuration. An offline deployment requires an accepted bundle, local model and disconnected installation tests.
Deployment guides for government entities
These guides cover how government entities deploy private AI agents on classified networks and under NCA requirements:
- Air-gapped AI agents — deploying on networks with no internet access.
- Private AI in Saudi Arabia — how KSA enterprises deploy private AI agents under NCA ECC and PDPL.
- AI agents and NCA ECC-2:2024 — mapping AcTree’s controls to NCA’s essential cybersecurity controls.
Guides for this team
Bring one question of your own.