The platform

A governed execution layer for enterprise agents.

Zahen is a controlled execution layer for business agents — retrieval-grounded, role-aware, human-approved, and audited end to end. Agents plan and carry out real work inside your workflows, but every sensitive action stays under human control and every step is recorded.

Two ways to work

Ask a question, or hand over a task.

Zahen · Ask Read-only

What is our parental-leave entitlement for a 9-month tenure?

Employees with 6–12 months' service are entitled to 45 working days of parental leave at full pay 1, with an optional 15 days unpaid extension 2.

1
HR Leave Policy 2026
§4.2 · parental leave
0.94
2
Employee Handbook
§7 · extensions
0.89

Ask

Ask a question in plain language. Zahen retrieves the approved documents you're allowed to see, composes a grounded answer, and cites its sources. Asking never sends an email or changes a record.

Zahen · Agent task Multi-step
Plan generated
4 steps · grounded in policy
Sources retrieved
Permission-filtered
Draft prepared
Awaiting tool call
Human approval
High-risk step · held

Agent Tasks

Give the agent a goal that needs more than an answer. It plans, acts through approved tools, checkpoints its progress, and pauses for human approval on high-risk steps — then resumes on the decision.

How governance works

Six layers, one controlled path from question to action.

Approved knowledge

Documents and policies are ingested into an access-controlled index. Retrieval is pre-filtered by each user's permissions, so agents only see what a role is allowed to read.

Workflow orchestration

Stateful, multi-step runs that plan, execute, checkpoint, and resume. Agents complete real work, not one-shot answers.

Tool execution

A governed tool registry mediates every action through approved tools, with schema validation, timeouts, and rate limits. Agents never hold credentials — calls are brokered.

Human approval

Sensitive actions pause for a human to review, edit, approve, or escalate — and no one can approve their own request.

Audit trail

An append-only, exportable record of every request, retrieved source, tool call, approval, and outcome.

Observability

Live visibility into what agents are doing, where they're grounded, and where they're escalating — plus usage and cost by model and department.

What's inside

The surfaces your team works in.

Operators get the full console; everyday staff get a simpler employee mode. Both are bilingual English / Arabic and RTL-first.

Agent Workspace

Ask questions and read grounded answers with citations.

Agent Tasks

Stateful, resumable runs with a live execution timeline.

Approvals

A human-in-the-loop queue for high-risk proposed actions.

Audit Log

An append-only, exportable record of every privileged action.

Knowledge

An access-governed index of approved policy documents.

Tool Registry

Register, enable, disable, and gate the tools agents may use.

Integration

Agents work through the systems you already run.

Time-to-value depends on integration. Zahen connects to the enterprise systems where your knowledge and work already live, and acts through approved, governed tool adapters. Notifications and approvals reach people in Slack and Microsoft Teams.

Common connectors
Microsoft 365SharePointSlackMicrosoft TeamsCRMERPHRMSTicketing

Slack supports notifications, starting an assistant, and approving in-message. Microsoft Teams today posts notifications to a channel with a link back to Zahen. Connector availability and scope are confirmed during the readiness workshop.

Platform FAQ

How the platform fits.

Is Zahen a single platform or per-workflow tools?
One governed platform. Assistants (workflow packs) are configured agents on the same controls — approved knowledge, approvals, audit, and observability — so you expand department by department without re-buying governance.
Can agents take real actions, or only answer questions?
Both. “Ask” is read-only and cited. “Agent Tasks” plan and act through approved tools — with high-risk actions paused for human approval and every step logged.
How does Zahen keep answers grounded?
Retrieval is filtered by role and department before the search runs, and answers cite the approved sources they're drawn from. If the documents don't contain the answer, the assistant says so instead of guessing.
Who uses the operator console vs. employee mode?
Approvers, admins, security, and developers get the full operator console (workspace, approvals, audit, administration). Everyday staff get a simpler employee mode — pick an assistant, fill a short form, and track requests.

See where governed agentic AI fits in your business.

Book a readiness workshop with our team. We'll map your highest-value, lowest-risk first workflow — no obligation to proceed.