Add AI to the software you already have
For teams that already use ChatGPT or other AI tools, and run on software of their own.
You don't need a new system to get real value from AI. I connect AI agents to the software and documents you already have: a WhatsApp agent that answers from live data and writes back to your records, an internal assistant your staff can ask in English or Arabic, and document reading that turns invoices and receipts into structured data. Everything runs on your systems, with your permissions, at a fixed price agreed after a free call.
What I build
A WhatsApp agent on your live data
Connected to your system through its API. It checks stock, order status and availability as it replies, creates bookings and orders, and hands over to your staff with the full context.
“Is order 4127 out for delivery?” answered from your system, in seconds.
An internal assistant for your team
Your staff ask questions in plain English or Arabic and get answers from your system and documents, limited to what each person is allowed to see.
“How many units are left in the Al Khor branch?” “Which invoices are overdue this week?”
Document reading
Invoices, receipts, delivery notes and quotations are read and turned into structured records in your system. A person checks before anything is saved.
A supplier's PDF invoice becomes a purchase record, ready to approve.
Answers from your own documents
Price lists, policies, manuals and contracts become something your team or customers can ask, with every answer pointing to the document it came from.
“What's our return policy for opened items?” with a link to the policy.
How it connects
How it connects
- Your system's API, or a careful read-only connection to its database when there is no API.
- Webhooks, so the agent reacts when something changes: a new order, a payment, a delivery.
- The WhatsApp Business Platform for customer chats on your own number.
- LLMs from the main providers, chosen per task, including the one your company already approves.
- Only the data a task needs is sent to the model, and nothing is stored outside your systems that doesn't have to be.
How I keep it accurate
- Before launch, it is tested against a set of real questions from your business, and the answers are checked with you.
- Answers come from your data and documents. When the agent can't find the answer, it says so and hands over.
- Permissions follow your system's roles, so nobody sees more through the assistant than they could before.
- Actions that change money, such as refunds or discounts, need a person's approval.
- Every conversation and every action is logged, so you can review what the agent did and why.
When you don't need this
- Your software already has a built-in AI feature that does the job well. Use it.
- Your business still runs on Excel and WhatsApp groups. Start with your own system, then add AI on top.
- You only need help writing emails and posts. ChatGPT on its own is enough for that.
How a project runs
A free call
We go through your systems, what they can connect to, and where AI would save the most time.
A working version in 7 days
Built on your real data, tested with your team and your own questions.
Test and tune
We check the answers together and adjust until your team trusts it.
Go live and review
It goes live with logs you can review, and I stay on for support and improvements.
Questions technical teams ask
Which AI models do you use?
The one that fits the task and your company's rules. I work with the main providers and can use the one you already approve.
Is our data used to train the models?
I use the providers' business API plans, where they state that your data isn't used for training by default, and I show you their terms before we start.
Do we need to change our software?
Usually not. If it has an API, I connect to it. If it doesn't, we look at the options together on the free call.
Who owns what you build?
You do. The agents, prompts, connections and code are yours, like everything else I build.
How much does it cost?
A fixed price after a free call, agreed before any work starts. Bigger additions later get their own fixed price first.
Keep reading
- Pretace, the AI teammate I built and runHow it answers customers and when it brings in the team.
- Why not just use ChatGPT?Where a general chatbot stops and an agent on your data starts.
- AI in plain wordsFor colleagues who are new to AI.
- The WhatsApp assistantWhat's included, how long it takes and how the price works.
Tell me what you run on. I'll tell you where AI fits.
Step one is a free call about your systems. If AI isn't worth it for you yet, I'll say so.