Customer-selected model providers
Operational Assist accepts customer-owned OpenRouter, DeepSeek, Google, Anthropic Claude, or OpenAI/GPT credentials, then verifies the connection and restricts work to selected models.
Assist & Automation
AI helps with the work. Your team keeps control.
Investigate, summarize and prepare actions using your chosen provider, with authority and execution kept explicit.
A change you can inspect
Separate the suggestion from the authority to act.
AI-assisted operations
Customer-selected models, bounded context, and human controlAI recommendation pathHuman reviewed
SignalSource attached
Scope checkPolicy applied
AI suggestionRecommendation
ReviewHuman decision
No model execution authorityRetry boundedRollback defined
Illustrative workflow · explore the steps · no action is executed
01
Choose provider and policy
Connect an approved customer-owned AI credential, select permitted models, and define automation level, budget, and data rules.
Confirm this step for your release, permissions and environment.02
Build bounded context
Retrieve only authorized tenant and customer records, remove secret values, and show the operator what informs the result.
Confirm this step for your release, permissions and environment.03
Review or execute
Validate the proposed output, require approval for risk, run through the existing module owner, and capture outcome evidence.
Confirm this step for your release, permissions and environment.Use model assistance without transferring authority to the model.
Use a customer-selected model to explain, draft, or diagnose only the permitted context.
Return any action to the owning module's permission, approval, signed execution, stop, rollback, and audit path.
Inside the workflow
The work this connects.
- Operational copilot
- Summarize alerts, devices, tickets, changes, backup state, vulnerabilities, and runbooks within the active scope.
- Plan and script drafting
- Generate reviewable remediation plans, scripts, ticket notes, customer updates, and documentation drafts.
- Software update assistance
- Compare installed and current versions, explain risk, and prepare policy-compliant update remediation.
- Controlled endpoint tasks
- Route supported Agent-S desktop actions through local policy, consent where required, sandboxing, recording, and audit.
- Customer provider choice
- Use permitted models through OpenAI, Anthropic, Google, DeepSeek, or OpenRouter with a customer-owned credential.
- Usage governance
- Set budgets, rate limits, model allowlists, automation levels, and billable outcome controls by tenant.
Try one useful scenario
Autonomy within a specific policy
Fictional scenario: an assistant proposes a read-only diagnostic under a preapproved policy. A later restart is outside that scope and waits for action-time approval; an expired approval blocks execution.
Illustrative evaluation scenario, not a recorded product result.What a useful result must show
Capture policy version, permitted target/action, provider/model, redacted inputs, token/spend limit, denial or abort, and a separate outcome check.
Get the evaluation worksheet →Check the support boundaries
- Advisory suggestions
- Preapproved actions inside a bounded policy
- Action-time approval for exceptional actions
- Provider/model support, Agent-S and spend limits
The human-review illustration is one scenario. Agent-S denotes the specialized execution agent described in the evaluation material; availability, installation prerequisites and tested release need confirmation. Review non-secret operational data sent to the chosen provider, provider retention/residency and key revocation.
Check public evidence and availability →Scope, availability and evaluation details
Model output is not operational authority and must not invent providers, routes, fields, or evidence.
High-risk actions remain blocked outside the configured autonomy, role, MFA, approval, and tool policy.
The proposed $99 monthly model includes the target product scope for evaluation, unlimited technicians, unlimited customers, and the first 50 managed devices. Final terms require confirmation. See transparent pricing →
Review full environment and integration scope
Actions return to the owning module
AI can recommend or draft across alerts, tickets, scripts, workflows, and documentation, but execution still passes that module's scope, autonomy level, approval, stop, rollback, and audit controls.
Customers bring their own provider connection and pay that provider directly; Dayrook does not add AI-credit pricing.
Questions, answered
Before you choose your next step.
Does Dayrook use one shared AI key?
No. Tenants connect an approved customer-owned provider credential and choose the models their policy permits.
Can Assist take action without a person?
Only within the tenant's selected automation level and risk policy. High-impact actions still require the same security and approval controls as manual work.
Are Digital Vault secrets sent to an AI provider?
Secret values are excluded from AI context. Context retrieval is tenant-bound, permission-checked, and recorded without logging sensitive content.
Which AI providers are supported?
Dayrook’s target scope includes approved models through OpenAI, Anthropic, Google, DeepSeek, and OpenRouter; availability follows the current provider matrix.