Decisions and tasks — by QST
Qarar+ runs the full decision cycle: request, recommendation, approval, decision, and the tasks it creates — with AI that runs inside your own servers and draws on your institution’s own memory.
The problem
Most entities decide well and remember badly. The reasoning behind a decision lives in one person’s head, approvals circulate on paper, and the work a decision creates is chased by phone. Here is what changes.
The same subject is debated again because no one can find what was decided before, or why.
Ask in Arabic and get a written answer citing the entity’s own decisions, each one openable.
An analyst starts from a blank page, hunts for precedents and drafts the recommendation by hand.
Qarar+ suggests the classification and department, warns of duplicates, surfaces precedents and drafts a recommendation with numbered sources.
Files move between offices with no record of who held them, for how long, or why.
The authority matrix picks the route by classification and value; every step is timestamped, and delays escalate automatically.
Clauses are typed into a spreadsheet, if at all, and ownership is never settled.
Each clause becomes a task with an owner, due date and priority — proposed from the entity’s own past work and confirmed by a person.
Managers learn a task is late only after it is late, and closure rests on someone’s word.
Tasks close on delivered evidence, and Qarar+ flags the ones likely to finish late, with the reasons.
Staff report progress by email, then someone is expected to retype it into the system.
Local AI reads the reply, extracts the progress, applies the clear ones and holds the sensitive ones for a person.
The journey
Every request travels the same six stages. Each has an owner, a status and a next step — and every action is recorded with who did it and when.
The subject and its justification are registered; Qarar+ suggests the classification and department and warns of duplicates.
Precedents are surfaced and a cited draft is prepared, then edited by the analyst and submitted.
The route comes from the authority matrix. Each reviewer sees the file, the precedents and the chain; delay escalates.
The decision is issued with its clauses, numbered and closed to further editing.
Clauses become tasks with owners, due dates and priorities; progress is carried by evidence.
Work closes against delivered evidence, and the decision joins the institutional memory.
In-house AI
The AI in Qarar+ runs entirely on the entity’s own servers, on the central processor, with no external service — so no document, request or message ever leaves. Each engine shows its confidence and its sources, and its live figures are read straight from the records.
The engines
Neural models read documents scanned in Arabic and English, with a second pass over Arabic to correct digits and words, and extract the decision’s number, date and subject for an employee to confirm.
Appears in: Archive import, and adding a document to knowledgeAn Arabic search engine that understands word forms — the definite article, prefixes and suffixes, and broken plurals — ranks results by relevance and answers in writing, citing a source for each statement, within the asker’s own permissions.
Appears in: Institutional knowledge, and suggested questions on the home pageMatches every new request to the decisions and archived documents most like it, warns of open requests on the same subject, and measures how many of its precedents analysts judge relevant.
Appears in: Register a request › review precedentsA nearest-neighbour model proposes the classification and department for a request, and the owner, due date and priority for each decision clause, drawn from the nearest past decisions and tasks, with a confidence score and a reason.
Appears in: Register a request, and create tasks from clausesA first draft is prepared in seconds from the request’s own data and its precedents, with every paragraph referred back to its source, for the analyst to review and edit before use.
Appears in: Request page › draft with AIReads staff replies to task emails and extracts progress, completion, delay and new dates with a confidence score. Clear updates apply automatically and can be undone; sensitive ones wait for a person to confirm.
Appears in: Tasks › email updatesEstimates the chance each open task finishes late from progress against time elapsed, the owner’s record of meeting deadlines, and changes to the due date — and shows the reasons for every task it flags.
Appears in: Tasks › likely to be lateThe guarantees
The engines run on the entity’s servers, with no graphics processor and no cloud model.
Every result is drawn from the entity’s own records, with its source and its confidence.
Accuracy is measured against the entity’s latest real decisions, and every AI action is written to the audit trail.
No sensitive change is applied without a person confirming it.
Next phase — Qarar AI. A large Arabic language model, hosted locally on the entity’s own servers, for richer drafting and summarising — again without data leaving those servers.
The product
The working interface, shown with demonstration data — in Arabic, with Hijri and Gregorian dates throughout.
Each role opens on its own home: what waits on them, what is late, where cases are stuck — and what the AI has prepared since yesterday.
A written answer about past decisions, every sentence carrying the decision it came from, and only from records the asker may open.
As the request is described, Qarar+ surfaces the entity’s own past decisions on the same matter and warns of duplicates.
Every decision numbered, dated in Hijri and Gregorian, searchable by subject, classification or number.
The AI prepares a first draft with numbered sources; the analyst opens any source, edits freely, and decides what is submitted.
The platform
Four areas follow the decision, from the first request to the day its last task closes.
Who it is for
Qarar+ suits any body whose decisions must be justified, traceable and followed through — and whose data must stay inside the Kingdom.
In about thirty minutes we walk your team through one request becoming a decision, and that decision becoming finished work — on your own records.
