Your systems stay as they are. The agents work on top of them.
We don't replace or rebuild your systems. We read from them read-only and add agents that understand their procedures and rules: they check, explain, analyse and build. We tested the method on the HR module of an ERP system, and it is the same for any system we can read.
HR and payroll
Leave and overtime limits, allowances, approval chains, and who holds each authority.
Finance and payments
Payments against contract terms, financial delegation tables, and segregation of duties between requester, approver and payer.
Procurement and contracts
Purchase method against value, split purchases, cumulative change orders, penalties and guarantees.
Projects and assets
Approved cost against the contract and its change orders, progress updated on time, and the asset register.
Warehouses and supply
Issue and transfer limits, receiving authorities, and stock counts against the records.
Citizen services and requests
Response deadlines, approval paths, reasons for refusal, and overdue requests.
What we need from you to start
Three inputs are enough to connect any system. We don't need copies of your data outside your network.
-
1Approved procedures
How the work is meant to run: steps, authorities, deadlines and limits. They go into the Reference Knowledge Base and are ratified by you.
-
2System design guides
Data tables, settings, workflows and integrations. They tell the agent where to find the evidence inside the system.
-
3A read-only account
On the test environment first, then production. The agent never writes to your systems or changes a record.
The result: agents that know your procedures, work on your system, and cite a text you ratified for every verdict.
One source of truth for laws, regulations and procedures
Governance and standards teams need one ratified copy that everyone refers to. In WakeelX that copy is the Reference Knowledge Base: every verdict an agent gives points to a text in it, with its version, the date it took effect and who ratified it. No conflicting copies, and no agent improvising.
Regulatory documents
Laws, implementing regulations, decisions, circulars and procedures, ranked by legal precedence.
Technical documents
System design guides, specifications, manufacturer manuals and approved security standards.
Contractual documents
Contracts, annexes and change orders, so delivery is measured against what the supplier actually committed to.
A ratified version never changes
Once ratified, a version is frozen. Any change needs a new version that goes through the same review.
In force on any date
It knows which version was in force on the day a case happened, and only ratified versions count.
Recorded human ratification
Reviewers verify, only the authorised owner ratifies, and who ratified and when is recorded. No agent can ratify.
Careful document intake
PDF in, Arabic OCR, and two independent readers that cross-check every number before review.
Saudi Arabia's new Government Tenders and Procurement Law takes effect in early 2027 and changes the limits for direct purchases, change orders and penalties. The knowledge base keeps both texts, so each case is measured against the text that was in force on its date.
An illustration of how one document keeps its versions: only the ratified version in force counts.
It checks what the system actually does, not what the documents say
The agent does not compare one document with another. It goes into the system itself, reads its settings, code and records read-only, measures what it finds against the ratified text in force, and gives a verdict with its evidence.
-
01
The ratified text
From the Reference Knowledge Base, in the version that was in force on the day of the case.
-
02
Read from the implementation
The system's settings, code and records, with read-only access and no changes.
-
03
Compare
What the system actually does, against what the ratified text says.
-
04
Verdict with evidence
Match, mismatch or insufficient evidence, with the source in the system and the article in the text.
Maximum emergency leave
- System setting: annual emergency leave limit7
- Ratified rule text, clause 4.25
- SourceERP system · read-only
Saudi governance frameworks that can be measured from the implementation
Scoped to your needsThe easiest controls to measure are the ones with a number, a limit or a deadline, because settings, code and records can prove them. As long as the standard is written down and the system is open for reading, the method is the same.
WakeelX is not affiliated with any government body. The names above refer only to their published frameworks.
Build what you need without waiting for a development project
Many needs are small but wait months in the development queue: a report, a tracking dashboard, a tool for one department. Describe what you want in Arabic or English, and the coding agent writes the code, runs it on your data in an isolated sandbox, and shows you the result to review before it is approved.
Data analysis
Ask about your data in plain language: trends, comparisons and outliers, with the code that produced the numbers.
Dashboards
Tracking dashboards for department indicators: overdue requests, spend against budget, compliance results.
Internal apps
Entry forms, calculators and small apps for one department, built on your system's data.
Automation and recurring reports
Converting files, merging spreadsheets, and weekly reports prepared automatically instead of by hand.
Ready-made agents, and agents you build yourself
Each agent works within the permissions of the person who asked, and everything it does is logged.
Compliance agent
Answers one question precisely: does our system apply our regulations as written? It reads from the implementation and gives a verdict with evidence, for any rule the system enforces.
Technical support agent
Connected to the organisation's system, it understands an error in context: why the transaction was refused, which rule blocked it, how to correct it, and who owns the fix.
Coding agent
Analyses data, builds dashboards and internal apps, and automates repetitive work, in an isolated sandbox. Details above
Build your own agent from the console
Lab-testedCreate a new agent from a template: set its name and job, the model it runs on, the tools it may use, and its permission and step limits. Then switch it on, run it and follow it on the task board. Governance and audit apply to it automatically from its first step.
- 1Pick a template
- 2Set tools and permissions
- 3Switch it on and run it
- 4Follow it on the task board
Lead agent and task orchestrator
One entry point: it answers itself, routes the request to the right agent, or splits it across several agents.
Governance built in from the start
AI is only useful when you can trust what it does. So no agent reaches a tool or a record except through the governance layer.
Data stays in your network
The models run on your servers in the Kingdom, with no calls to outside AI services.
Every tool call checked
Before it runs, against the user's permissions and your policies.
Tamper-evident audit trail
Every request and step is kept, and any change to the record is detected.
People decide
Agents suggest and explain. Sensitive actions wait for the person with authority.
Single sign-on and roles
Your organisation's own accounts, with role-based permissions.
Certificate-verified links
Mutual TLS between services, with your organisation's certificates.
Runs where your data lives
Your systems stay as they are: agents read from them read-only, and the models run locally.
On-premises, air-gapped
Lab-testedAn offline installer, on your organisation's own servers.
Private cloud
PlannedThe same installer on your private cloud in the Kingdom.
Connected today: ERP systems and PDF documents. Any other system is connected once we receive its procedures, design guides and a read-only account.
How we start
- 1
A WhatsApp conversation
The WakeelX assistant replies right away and books a call with our team.
- 2
Pick the system and the need
You share the approved procedures and system design guides, and we agree what to measure first.
- 3
A live demo
We show the agents on test data that resembles your work.
- 4
A trial in your environment
We install the platform inside your network and measure the results together.
Frequently asked questions
Do we need to replace our current systems?
No. The agents work on top of your systems as they are, and read from them read-only. We don't rebuild the system or change its data.
What do we need to give you?
The approved procedures, the system design guides, and a read-only account. We don't need copies of your data outside your network.
Which systems does it connect to?
We have tested it with the HR module of an ERP system and with PDF documents. Any other system, such as finance, procurement or projects, is connected the same way once we receive its procedures and design guides.
Does the agent replace internal audit or an accredited assessor?
No. The agent prepares evidence and finds gaps continuously. The final judgement and sign-off belong to internal audit and accredited assessors, and anything that needs interviews or paper records stays with them.
How can we trust the agent's verdicts?
Every verdict points to a ratified text in the Reference Knowledge Base and to evidence read from the system itself. When evidence is missing, the agent says "insufficient evidence" instead of guessing.
Does the coding agent need its own GPU server?
No. The sandbox that runs the code uses an ordinary CPU, and the model that writes the code is the same one the platform runs on. One model server with enough memory is enough.
Can we build our own agents?
Yes. From the console you create an agent from a template, set its tools and permissions, then run and follow it. Governance applies to it automatically.
Does WakeelX need an internet connection?
No. It installs and runs inside your network with no internet, and the models run on your servers. The interfaces and agents are Arabic-first, with full English.