Answers from your own documents, with the source attached.
A knowledge assistant that answers your teams in Arabic and English from the documents you choose, cites the passage it used, and works under access rules your IT team sets. Every agent comes with a security review and an evidence pack.
With a source
Document answer
- What is the returns window?
- Fourteen days from delivery, as published.
- Policy v3 · p. 4
The problem
The answer exists. Nobody can find it.
Policies, procedures and project documents are spread across drives and wikis in two languages. Staff ask colleagues instead of searching, and IT is wary of any AI tool that might reach documents it shouldn't.
Use case 01
Answer with a source
A RAG assistant grounded in the documents you choose. Every answer shows the passage it came from, and chat logs and a cost dashboard let you see what was asked, what was answered, and what it cost.
Connects to your document store
A person approves which documents the assistant may read
Use case 02
Connected through MCP
Agents reach internal systems through MCP servers we build and test, with least-privilege access and a runbook. Reading is allowed where you allow it, and anything that writes needs approval.
Connects to your internal tools, through MCP
A person approves every write action
Least privilege
MCP trace
- search_docs
- get_policy
- create_ticket: blocked, needs approval
Use case 03
A security review before go-live
Every agent gets a supply-chain review, scoped permissions, logging and an evidence package your IT and security teams can sign off.
Connects to your IT review process
A person approves go-live
Before go-live
Security review
- Permissions scoped
- Logs on
- Evidence pack
More use cases
Scheduled briefings
Briefings assembled by agents from the sources you choose and delivered to the inbox on a schedule.
Cost you can see
A cost dashboard for each assistant shows what every request costs. If simple questions are costing too much, model routing can send them to a cheaper model.
Checked before go-live, scored after
Answers are tested against agreed questions before launch, and real conversations are scored afterwards.
Under your controls, on your accounts.
01
Discover
We agree with IT which documents and systems are in scope, who may ask what, and what the assistant must never touch. You get a written architecture and one success measure.
02
Build
Two-week increments, each ending in a working demo on your documents, with access read-only by default.
03
Launch and evaluate
The assistant goes live behind the agreed permissions, with logs on. Answers are scored, and gaps in the documents go into a backlog.
04
Operate or hand over
We run it on a monthly retainer, or hand it to your IT team with the runbook. The code sits in a repository you own, on your own cloud and API accounts.
From production
Source attached
every answer shows where it came from.
Read-only
by default, with writes behind approval.
EN · AR
questions and documents in both languages.
2 weeks
per delivery increment, each ending in a working demo.
Your repo
the code, on your own cloud and API accounts.
Why us
Answers with sources
Every answer cites the document it came from.
Permissions respected
The assistant sees only what the asking user is allowed to see.
Security built in
Least-privilege access, scoped MCP connections, data-loss-prevention masking, prompt-injection testing and an evidence pack for IT.
In your environment
It runs on your accounts, and the code lives in your repo.
Best for
- Organisations with large internal document sets in Arabic and English
- IT teams that need evidence before approving an internal AI tool
- Teams that want visibility into what every AI request costs
Related
Questions
- Can the assistant see documents a user isn't allowed to see?
- No. It sees only what the asking user is allowed to see, and we test that before go-live.
- Where does our data go?
- The assistant runs on your accounts and in your environment. Sensitive data is masked before it reaches a model, and every request is logged where your team can read it.
- What does IT get before go-live?
- An evidence pack covering scope, permissions, connections, test results and logs, for IT to sign off.
- Do you certify compliance?
- No. We provide the evidence and controls, and your IT and compliance teams make the call.
Start with the questions your team asks every week.
Book a free consult. Bring a handful of the documents your team searches most and the questions they ask, and we will show you how an assistant would answer, what source it would cite, and what IT would control.