02 / Agentic AI company
A Saudi AI company I co-founded and lead technically: chat and voice agents grounded in each client's own knowledge, connected to the systems they already run.
Deliveries to Jeddah take 2 to 3 working days.
Searching knowledgeSource 01 / Shipping policy, p. 4
أبغى أحجز موعد معاينة يوم الثلاثاء الساعة ١١
Checking slotsTue 10, 11:00 booked
Agent handlingKhalid has it, with full context
02 / Overview
BotWisor helps businesses in the Kingdom put AI where the work happens: conversations with customers, the knowledge teams search for every day, inbound calls and the workflows behind them. I lead how we design and build it, from the agent core to the brand platform that sells it.
03 / Problem
The gap was never the model. It was everything around it.
04 / Define
Three rules every BotWisor program follows, whatever the sector.
Every interface, prompt and piece of copy exists in both languages, with full right-to-left layouts.
Agents retrieve from the client's own documents and data, and say which source an answer came from.
A delivery model from discovery to go-live in under 8 weeks, with integrations scoped up front.
05 / System
Website chat, WhatsApp and phone calls all reach the same agent core. It retrieves from the client's knowledge, acts through their tools, and knows when to hand over to a person.
06 / Decisions
The thinking
Adding Arabic at the end produces broken layouts and copy that reads translated.
The decision
All copy lives in one typed source for both languages. Layout, type and direction switch together, and the brand site was built that way first.
The result
Arabic pages that feel native, and no English string that can ship without its Arabic pair.
The thinking
An agent that never lets go frustrates the customers who need a person most.
The decision
Every agent has explicit handoff rules and passes the conversation, the customer's details and its own summary to the right team.
The result
People pick up mid-conversation without asking the customer to repeat anything.
The thinking
Building a separate bot per channel and per client multiplies work and bugs.
The decision
A shared agent core with per-client knowledge, tools and tone, exposed through chat, WhatsApp and voice.
The result
New programs start from a working system, which is what makes the 8 week go-live possible.
07 / Solution
Journey 01
What a single conversation does behind the scenes.
The visitor is answered instantly, in the language they wrote in.
Pricing, policies and product questions are answered from the client's own documents.
The agent captures what the sales team needs to know.
A complete lead lands in the CRM, with the conversation attached.
Sales picks it up with the full context, ready for a call.
Journey 02
The same core, shaped around how each industry works.
Inquiry handling, viewings and follow-ups.
Customer service with strict data residency.
Order questions, returns and recommendations.
Bookings and information in both languages.
08 / Outcome
Each new engagement can start from a working agent core, a bilingual brand platform and a delivery model clients can plan around.
Chat, WhatsApp and voice agents share retrieval, tools and handoff, so improvements reach every client.
Trust in AI comes from boring things done well: sources, handoffs and Arabic that reads like it was written in Arabic.
09 / Measures
The commitments every program is designed and delivered against.
Go-live
≤ 8 wks
From discovery to a live agent in production.
Grounding
Every answer
Traceable to a source in the client's own knowledge.
Handoff
Full context
No customer asked to repeat themselves after a transfer.
Next project
Every WhatsApp inquiry becomes a case. AI answers first, checks stock and credit in SAP, and the sales team closes the deal from one screen.