ELARD · Meeting summary
Our meeting at ELARD: summary, IT infrastructure and next steps
Wednesday 18 November 2026, Zalka · Prepared for Ricardo Khoury and the ELARD team · From Rawad and our team
Our aim
It takes the repetitive work off your engineers. It is built to assist them, not to replace them.
It drafts and searches. Your engineers check, decide and sign.
What we discussed
Starting a new report from similar past studies; you mentioned four types of studies.
Preparing proposals and answering tenders.
Reviewing mail, and how incoming questions are handled.
Mapping the sensitive receptors around a project.
Gathering information about a site or a facility: what it is, its data, its location.
How your engineers can verify what the AI brings back.
Keeping your data inside ELARD, and your concern about public AI tools.
Built to assist your engineers and analysts, not to replace them.
What we commit to
1
Nothing of yours is sent to an outside AI unless you approve it in writing, job by job.
2
Drafts stay marked as unverified until an engineer confirms each source. Before go-live we measure, on your own studies, how often answers are right.
3
The system has no means to send mail, sign or issue a price. We show you this in the acceptance test.
4
The AI takes the repetitive part of the work. Staffing decisions remain yours.
5
You can export all of it at any time.
Offline and online work
Everything with ELARD material stays inside, on your own machine.
Online is optional: only research of a kind you approved in writing goes out.
A checkpoint you control holds back your documents and client names.
Without the online part, everything still works. No checkpoint can stop every copy and paste, so a short ELARD rule covers the rest.
Why your own system
01
It stays on your own machines, at your own sites, under your own keys. No cloud contract can say that.
02
In 2026, online AI services have had outages, policy changes and retirements.
03
by people who come to your office.
What to expect. A system on your own machine is narrower than the largest public assistants. Its strength is that it works on ELARD's own studies. The pilot shows you how well, before you decide.
The first job, and how we prove it
1
One type of study you choose, before any machine is bought.
2
A handful of your past studies already published, on our own test machine: nothing confidential leaves ELARD.
3
How often the answers were right, how long a first draft takes, how well it reads Arabic and scanned reports.
4
Marked unverified until an engineer confirms each source. Maps drawn by your mapping software; your specialist checks.
5
The result sizes the machine and how many people it serves at once.
How the system sits inside ELARD
Each person has their own login.
Only the people and devices you authorise reach the system, over an encrypted private link.
Encrypted backups in three places; two named people hold the key, and we hold no copy.
Option 1. The people drawn are an illustration only; the real numbers come from your answers.
The IT infrastructure ELARD will need
At your main office, specified after the pilot.
A small room or cabinet with cooling, and two named key holders.
A backup power unit at each site; a clean shutdown and restart if all power fails.
And a second line if possible: your other offices reach the system through it.
Only the people and devices you authorise. Tested with you before go-live.
Kept in three places of your choice.
Trained by us; every user also gets a short training.
On a rhythm agreed in writing; every action is recorded.
Why we cannot name the machine yet. One machine, a larger professional machine, or more than one: the pilot measures the real work first, and the cost comes with our options.
Who can see what
Every request goes through the front door, which checks who you are and your role.
The AI reads only what that person may read.
Our team has no access unless ELARD opens it, and ELARD can close it at any time.
The roles are an example; ELARD's management sets the real ones before go-live, and we test them with you.
Two options
You can start with the first and add the second machine later: nothing is thrown away.
Option 2
On a normal day both machines serve your teams and share the work; machine 1 holds the main copy of your data and machine 2 a live copy.
Side by side
| Option 1: One machine | Option 2: Two machines | |
|---|---|---|
| On a normal day | One machine serves everyone | Both machines serve your teams and share the work |
| If a machine stops | Work continues without the AI; the machine is repaired or replaced, then restored | The other machine carries everyone; if it was machine 1, your IT contact or deputy moves the main copy in one step |
| Work that can be lost | Since the last daily backup | Very little: the second machine holds a live copy |
| Fire or theft at one site | Restored from the daily copy kept outside the building | The other site keeps running |
| What ELARD prepares | One site | Two sites, and a third office for the small third machine |
| Costs, time frames, recovery times | With our options after the walkthrough and the pilot | Measured on a real test before we commit to them |
Keeping ELARD running
Teams keep working as today; restored from backup, or the other machine carries everyone.
A daily copy and the weekly unplugged copy are kept outside the building.
Several weeks of copies are kept; a copy is checked before it is restored.
A backup unit carries short cuts; if all power fails, a clean shutdown and restart.
The system keeps working where it sits; a second line reduces the risk.
Authorised people reach it from wherever they work, or the second site serves everyone.
Only authorised people and devices, over an encrypted private link; every use is recorded.
It can mislead a draft but cannot act: the system cannot send, sign or price.
Applied as soon as needed; your IT contact opens our access; every action is recorded.
Nothing in the system depends on one. It runs on your own machine.
A named deputy, trained in the same way, and the written guide.
You own the machine, your data and the keys, and can export everything.
Before go-live we restore a full backup in front of you and show you how long it takes; then we test it regularly.
Next steps
1
Signed before the walkthrough.
2
Of your departments, with the people you name.
3
On one type of study, before any machine is bought.
4
Time frames, costs, support terms and contract basics.
5
What we build, and how.
Your contact, and the colleagues who met us
People in each department and office
Where your archive is kept, and who looks after it
Who runs IT, a deputy, and your software
What must stay strictly private
Who approves the budget
Client contract and tender rules on data and AI
For option 2: which site, and in which country
A handful of published studies for the pilot
The languages of your archive, and how much is scanned
Your mail system
The two people who would hold the backup key
ELARD · Next step
Rawad leads our team and is your contact.
Before the walkthrough, we sign a confidentiality agreement with you.
If we misunderstood anything, please tell us and we will correct it.
Turn your phone sideways for the full view