Work
Work, shown as systems.
We publish what is real. Radd, our first AI Employee, works as a demonstration and is being deployed with our first establishments in Dakhla.
Systems, documented
Read our work like an engineer would.
Every case follows the same structure, from the problem to the stack. Where a number does not exist yet, we say so rather than estimate it.
Radd
Demo- Client
- Hotels, riads, guesthouses and surf camps in Dakhla. Being deployed with our first establishments.
- Problem
- Guest messages arrive at hours when reception is closed, in French, Darija, Arabic and English.
- System
- One Support AI on Radd Core, with a human branch to reception for anything outside the rules.
- AI
- Answers from the establishment's own knowledge, captures booking requests, switches language mid-conversation.
- Automation
- Message, retrieve, answer, capture, hand over with context.
- Integrations
- WhatsApp and a web widget.
- Outcome
- Measured after the first installations. We publish figures once they exist.
- Stack
- Radd Core engine · WhatsApp · web widget · four languages
Radd Core
Demo- Client
- Internal. Every JDM AI Employee runs on it.
- Problem
- A new role should not mean a new codebase.
- System
- Configuration-driven and channel-agnostic: identity, knowledge, memory, tools, channels, guardrails, logging, metering.
- AI
- Model calls wrapped in the role's identity, boundaries and confidence rules.
- Automation
- Every decision written to the trail.
- Integrations
- Channels plug in as adapters.
- Outcome
- A new AI Employee is a new configuration.
- Stack
- Zero runtime dependencies · 55 automated tests
Client portal · concept
What an owner sees each morning.
The roster sorted by what needs a person first, the day in four numbers, and every escalation with its conversation attached.
Dar Lagune · today
Needs a person first
- Conversations
- 18
- Handled by Support AI
- 14
- Handed to staff
- 4
- Median reply
- 9s
Airport transfer for 5 guests
Concept interface. Dar Lagune is a fictional establishment used for demonstration; every figure is illustrative.
Questions
What owners ask first.
What happens when it doesn't know the answer?
It does not guess. Below a confidence threshold, or on a trigger word you choose, it hands the conversation to a named person on your team with the full history attached.
Can I see what it said to my customers?
Yes. Every conversation and every action is in the trail, in plain language: what arrived, which source was used, which rule applied, what was sent and who approved it.
Which languages does it work in?
French, Darija, Arabic and English, including when a customer switches language in the middle of a conversation.
Do you leave after launch?
No. We operate what we build. Monitoring, fixes and knowledge updates are part of how every system is run.
How does a project run?
Discover how work flows today. Design what the system does and where a person stays in the loop. Build and test against real conversations. Integrate with your tools. Then optimise from the trail, month after month.
Is it running with clients today?
Not yet. Radd works as a demonstration and is being deployed with our first establishments in Dakhla. What we show you is the working system; it is not yet installed for you. You would be among the first five.