Company-wide AI workspace
One secure internal interface for every employee to chat with AI, summarize files, and get work done — in the web workspace or right inside Microsoft Teams, WhatsApp, and Telegram.
%20and%20must%20not%20be%20redrawn%20here.%20Light-surface%20treatment:%20ink%20%2314172E,%20brass%20%23C9973F.%20--%3e%3cg%20fill='none'%20stroke='%2314172E'%20stroke-width='9'%20stroke-linecap='round'%3e%3cpath%20d='M26%2022H78'/%3e%3cpath%20d='M42%2042H78'/%3e%3c/g%3e%3cpath%20d='M78%2062H46C31%2062%2022%2070%2024%2080'%20fill='none'%20stroke='%23C9973F'%20stroke-width='9'%20stroke-linecap='round'/%3e%3ccircle%20cx='48'%20cy='80'%20r='6'%20fill='%2314172E'/%3e%3c/svg%3e){kind=small}
ZIJEnterprise AI · Governed · Connected to your business systems
Intelligence thatunderstands, governs,and acts.
One platform where your workforce works with AI, your business systems answer in plain language, and every action taken on your data is authorised, attributed and auditable. ZIJ — formerly Employee Smart Assistance — brings a company-managed AI workspace and a governed execution layer together — so an answer can become an action without anyone losing control of either.
Arabic-first · Your data stays yours · Every action logged · Runs in your cloud or ours
What teams optimize with one platform
35-60%Lower AI spendvs individual subscriptions, in pilot deployments
70%Less manual back-office workin repetitive day-to-day tasks, in pilot deployments
<1sFallback switchingpremium to free model routing
100%Governed & auditabletracked per employee and action
The problem
AI is everywhere in your company — and nowhere working together. Meanwhile the systems that run the business hold the answers, but employees can't reach them.
- 01AI usage is fragmented across personal accounts and tools, making spend unpredictable and impossible to optimize.
- 02Sensitive product, customer, and supplier data gets pasted into unmanaged AI tools with no policy, no audit trail.
- 03Operational data stays trapped in the systems that hold it — employees wait days on IT or BI to answer questions they could ask directly.
- 04There is no consistent way to enforce role-based limits, token quotas, or model and data policies across departments.
The economics
Subscription sprawl + passive systems vs. one governed platform.
Today
- Each team buys separate AI plans; seats and tools are duplicated and underused.
- No company-wide quota strategy, no visibility into who uses what, or at what cost.
- Your systems only answer through reports — questions wait on IT, BI, or spreadsheets.
- Sensitive data leaves approved boundaries; no consistent policy or audit.
- Every new automation becomes a custom development ticket and a long backlog.
ZIJ
- One company AI strategy with centralized governance and predictable spend.
- Token limits by employee, role, or department — with automatic premium-to-fallback routing.
- Conversational access to your operational data: ask in plain language, get source-cited answers.
- A secure proxy and role-based controls keep keys, data, and actions inside your governance.
- New workflows ship in weeks on top of the systems you already run — every action logged and reversible.
What you get
One platform that understands, governs, and acts.
ZIJ replaces scattered AI tools and passive reporting with a single secure workspace. It understands your documents, your knowledge base and your live business data. It governs who may ask what, with which model, and at what cost. And it acts on your systems — creating and updating records — only after a human with the authority to approve has done so.
Conversational access to your data
Employees ask questions of the systems that run the business in natural language and get grounded, source-cited answers — no SQL, no waiting on reports or BI. Microsoft Dynamics 365 Business Central is supported today, with further connectors on the roadmap.
Governed action execution
The agent proposes and executes approved actions on your systems — like creating and posting sales, purchase, and transfer orders — under role-based controls, with full traceability and human-in-the-loop confirmation.
Token-based cost control
Set monthly or weekly token quotas per employee, department, or role to keep AI spend aligned with business priorities.
Automatic model fallback
Connect Claude, OpenAI, Gemini, DeepSeek, Kimi, and open/local models. When premium limits are reached, routing switches to a free fallback so work never stops.
RBAC & data-domain control
Define who can use which models, connectors, and data. Topic- and table-level data domains (finance, sales, costing…) are enforced even on generated SQL.
Knowledge base & documents
Your SOPs, contracts, and policies become first-class context via RAG. Upload PDF, Word, Excel, and images — with EN/AR OCR — for grounded answers.
Artifacts, voice & analytics
Generate downloadable Excel, PowerPoint, PDF, and images from chat, use voice input (EN/AR), schedule recurring reports, and watch usage, spend, and ROI on a live admin dashboard.
Runs inside your network
For data that cannot leave the building: a signed on-premises agent holds one outbound connection to the platform and runs a fixed list of named operations against your database. No inbound firewall rule. Your database password never leaves the machine. Read-only by design, re-validated locally, and audited on your own server.
Approvals that follow your org chart
Requests route to the requester's own management chain instead of broadcasting to everyone with an approver role. Segregation of duties means whoever asked cannot be the one who approves. Set monetary thresholds and compliance block-lists, and approve from Teams, WhatsApp or Telegram without opening the app.
Multi-step processes that wait for people
Record a request, pause for a human decision, then act on the answer — as a durable process that survives a restart mid-wait. Retries are bounded, side effects are never executed twice, and every step is journaled.
Dashboards & saved analytics
Versioned dashboards built by asking, refreshed in the background, and exported as branded PowerPoint or Excel. Save any result as a named dataset your team can reuse — and every viewer resolves it under their own permissions, so sharing never widens access.
Audit trail your SIEM can read
Every sign-in, permission change, query, tool call and executed action lands in one queryable trail, and streams to Microsoft Sentinel, Splunk or QRadar in a standard schema with signed delivery. Retention windows and legal hold are configurable.
Your brand, on everything it produces
One palette, logo and font applied across dashboards, generated documents, the workspace and every email — plus your own corporate PowerPoint, Word and Excel templates, so what the assistant produces looks like your company made it.
Built for Saudi data protection
Arabic and English natively, not by translation. Workspace-isolated storage, per-user data erasure with a compliance record, and deployment in your cloud, a private cloud, or with the data plane entirely inside your network.
Use cases
Built for every department — governed centrally.
Roll out AI safely across the company while keeping cost, access, and compliance measurable from day one. These are the patterns teams adopt first.
Resolve tickets faster, on policy.
Agents draft replies, summarize customer and order history from your systems, and suggest compliant answers — while managers control usage and quality.
Create campaigns and proposals at lower cost.
Teams generate content and pull live sales figures from your systems, while token limits and fallback routing prevent budget overrun.
Close faster, reconcile smarter.
Ask for pending approvals, invoice matches, and variances; let the agent draft commentary and propose actions under finance-domain controls.
Standardize internal workflows.
Draft SOPs, policies, and bilingual announcements, and query operational data — with clear access controls and full traceability.
Enable productivity with policy controls.
Developers get AI assistance for code and analysis while IT governs model access, connectors, data domains, and security boundaries.
Every prompt and action, accountable.
Model requests, token events, fallback switches, and executed actions are logged to a queryable audit trail — built for internal control and regulator readiness.
How it works
Launch one governed platform in five practical steps.
01
Connect AI providers & your systems
Link your AI APIs (Claude, OpenAI, Gemini, DeepSeek, Kimi, open/local) through a secure proxy, and connect your systems of record — read-only or governed read/write. Business Central is supported today.
02
Set roles, limits & data domains
Assign token quotas and RBAC by user, team, or role, and scope which data domains and models each group is allowed to use.
03
Ground it in your knowledge
Upload SOPs, contracts, and policies to the knowledge base so every answer reflects how your organization actually works.
04
Employees work in one interface
Teams chat, analyze documents, query your business data, and generate artifacts — with automatic fallback keeping sessions running past premium limits.
05
Admins monitor & govern
Leaders review usage, spend, and audit dashboards in real time, tune quotas, and enforce policy with complete visibility.
Example workflow
One email, end to end — drafted by AI, approved from WhatsApp.
Here is a real, governed workflow on ZIJ. An inbound supplier email triggers the assistant; it drafts a grounded reply from your knowledge base and your business data; and because it involves a price commitment, it pauses for one-tap human approval before anything is sent — with every step logged.
Scenario · Finance & Procurement
A supplier email lands in the shared procurement inbox. No one has touched it yet — and by the time a manager taps “Approve” on their phone, the reply is already on its way, with a full audit trail behind it.
A supplier email arrives in procurement@
The assistant watches the shared inbox. It detects the new email, classifies it, and starts a run automatically — no human action required yet.
It researches and drafts a reply
It reads the email, pulls the supplier's contract terms from the knowledge base, looks up the latest invoice and payment status in your system of record, and drafts a bilingual (Arabic + English) reply with a counter-proposal.
Policy decides this needs a human
The reply contains a price commitment, so the governance policy returns “require approval.” The run pauses and a checkpoint is created — nothing is sent without sign-off.
The manager gets a one-tap approval card
An approval card is delivered to the manager's chat app: the full email, the full draft, and Approve · Revise · Reject buttons. They never leave WhatsApp or open another system.
Tap Approve → the run resumes → the email goes out
On approval, the assistant sends the reply through the email connector. Tap “Revise” instead and it loops back to redraft. Either way, who approved what — and when — is written to the audit log.
The result
Hours of back-and-forth become one notification and one tap — with company data, policy, and a complete audit trail enforced at every step.
Inside the platform
Governed intelligence and live business data — in one employee workspace.
What you see below is the assistant employees use every day: an internal AI workspace that answers from your knowledge base and your live business data, with quota awareness, instant fallback, and an admin governance dashboard behind it.
ZIJ
Employee WorkspaceLive · Policy Enforced
Top 5 selling SKUs in the Eastern Region this month — and draft an Arabic + English summary.
Here are the top SKUs for the Eastern Region, resolved from your live data 40 seconds ago. Premium quota is at 94% — fallback will auto-activate at 100% if needed.
Top SKUs · Eastern RegionERP · Live
SourceLive · resolved 40s ago
Data domainssales, inventory
Premium tokens used47,100 / 50,000
Policy profileStandard Employee
One internal assistant — chat, documents, and live business data — with quota awareness and automatic fallback continuity.
erp.company.com / ai-ops
Modules › AI Ops › Control Center
AI & Action Governance Dashboard
Company-wide usage, spend, and executed actions
Active Employees
186
Premium Usage
82%
Fallback Sessions
1,248
Projected Savings
42%
Request Routing Flow · TodayWorkflow automation · live
Requests9842
Premium6310
Fallback3192
Actions Executed540
Recent governance eventsAudit trail
- 10:12Sales team reached premium cap; auto-switched to fallback modelWARN
- 10:04Purchase order SAR 84,500 — approved by Finance Manager, not the requesterAPPROVED
- 09:56Query blocked: user lacks the costing data domainBLOCKED
- 09:48Audit events streamed to Microsoft Sentinel — signed, 0 droppedINFO
- 09:41Daily spend report exported to finance workspaceINFO
Usage, spend, model routing, and executed actions in one view — designed for admins, finance, and audit.
Real-time
Usage, spend & model visibility
Grounded
Every figure traceable to its source
Enterprise
Governance and audit readiness
Productivity comparison · per department
Where the hours come back.
Conservative estimates at enterprise scale (1,000+ employees). Numbers are workflow time for one representative task — not aggregate FTE math.
| Department | Representative workflow · systems touched | Today | With Smart Assistance | Saved | Quality lift |
|---|---|---|---|---|---|
| Marketing | Campaign brief · analytics + Drive | 3–5 hrs | 11 min | ~95% | Consistent voice · AR + EN copy · last-year benchmarks attached |
| E-Commerce | Listing audit · marketplace + e-com SKU sync | ~30 min/SKU | 90 sec/SKU | ~95% | Catches taxonomy + image issues · auto-fixes copy · cites source |
| Sales | CRM follow-up draft · order context from BC + OTO | 12 min/lead | 2 min/lead | ~83% | Personalized in AR/EN · pulls last 3 orders · suggests next-best |
| Procurement | Supplier email triage + reply · BC contract terms | 15 min/email | 2 min (approve) | ~85% | Contract terms checked automatically · WhatsApp approval |
| Inventory Mgmt | Stockout decision · BC + OTO · PO via supplier | 2–6 hrs | 90 sec | ~99% | Voice-initiated approval · zero waiting on email |
| Warehouses | Pick-list rebalancing · OTO OMS · driver dispatch | 35 min/wave | 5 min/wave | ~85% | Optimal route + load · WhatsApp dispatch to drivers |
| Finance | Invoice reconciliation · Business Central exceptions | 25 min/batch | 3 min/batch | ~88% | Catches anomalies humans miss · cited audit trail |
| HR | Policy Q&A · onboarding doc · HR system lookup | 20 min/Q | 30 sec | ~97% | Always cites HR policy · AR + EN · 24/7 in-channel |
| Legal Affairs | Contract clause review · risk-flag · cite precedents | 45 min/contract | 8 min/contract | ~82% | Risk-tier classification · always cites the source clause |
| IT | Tier-1 ticket triage · diagnostics · BC + HR access | 22 min/ticket | 6 min/ticket | ~73% | Escalates with full context · pre-fetches diagnostics |
Methodology
Workflow times are based on internal interviews and comparable retail-IT case studies. The “With Smart Assistance” column assumes a post-pilot steady state — month 3+ after rollout — including the time the human still spends reviewing and approving. Pre-pilot months will see lower savings.
Quality dividend
Time saved is the obvious win. The bigger win is consistency: every employee gets the same quality of context, citations, and policy adherence — including the new joiner on day one.
Security & deployment
What leaders ask before rolling AI out company-wide.
Yes — same platform, same deployments, same roadmap. Employee Smart Assistance was renamed ZIJ (Arabic زيج, pronounced ZEEJ). A zīj was the astronomical table Arab astronomers computed from: a governed set of rows you could act on with confidence, which is exactly what this product makes of your business data. Existing customers keep their environment, their data and their agreements; nothing has to be migrated.
Control your AI. Unlock your ERP. Scale smart.
Deploy one governed AI platform for your entire company.
Stop paying for disconnected AI subscriptions and waiting on reports. Give every employee productive, secure AI access — grounded in your knowledge and your ERP — with quota controls, automatic fallback, and enterprise-grade governance from day one.
Bilingual enablement (EN/AR) · API & ERP onboarding support · No sales pressure