AI consultant · for small business
You don't need an AI strategy.
You need one thing that works.
Most small businesses are sold a roadmap and a slide deck. What actually moves the number is one automation that removes a job somebody does by hand every single day — built properly, shipped in weeks, and handed over so you own it. I'm the engineer who writes it, not the consultant who bills for the meeting about it.
The work
Four things small enough to finish.
Each one is scoped to ship on its own. Start with whichever removes the most hours from your week — the rest can wait until the first one has paid for itself.
01
The assistant that answers for you
A bot on WhatsApp, your site, or both — trained on your catalogue, prices, and policies, not on the open internet. It answers the same forty questions your team retypes every day, hands over to a human the moment it should, and logs every conversation you'd otherwise never see.
Best if you lose evenings to the same customer questions.
02
The paperwork that fills itself
Quotes, invoices, contracts, reports, product descriptions — generated from your own templates and your own data, with a human approving before anything goes out. The model drafts; you stay the one who signs.
Best if someone bills hours for copy-paste.
03
Search that understands the question
Retrieval over your documents, catalogue, or archive so staff and customers find the right answer by meaning instead of exact keywords — with citations back to the source document, so nobody has to trust it blindly.
Best if the answer exists but nobody can find it.
04
The store or platform underneath it
Sometimes the AI is not the problem — the business needs the product first. Storefront, payments, cash on delivery, admin, deployment. I build the whole thing, then add the AI where it actually earns its place.
Best if you're starting from nothing.
The receipts
Small businesses, live systems, real numbers.
818 / 629
Registered clients and orders in the first two months
Belmo — a K-beauty store built from an empty repo: storefront, cash on delivery, 24–48h fulfilment across Morocco, self-hosted to keep the running cost low.
Read the Belmo case study →~500
Students enrolled on a live e-learning platform
Magic Hands LMS — courses, video lessons, quizzes, instructor dashboard, and Stripe/PayPal enrolment for a training academy that was running on spreadsheets.
Read the Magic Hands LMS case study →~500
Suppliers onboarded to a B2B marketplace
Maroc Fournisseurs — sub-second search across the catalogue and real-time negotiation chat between buyers and providers.
Read the Maroc Fournisseurs case study →Darija
A WhatsApp agent that talks like your customers do
RESO Khdma — an AI agent matching workers to jobs entirely inside WhatsApp, in the dialect people actually type, because that's where the users already were.
Read the RESO Khdma case study →
How it runs
No retainer, no discovery phase, no surprise invoice.
Step 1
A call, free
Thirty minutes. You describe where the hours go; I tell you which part is worth automating and — often — which part isn't. If nothing here fits, I say so and you've lost half an hour.
Step 2
A fixed scope
One page: what gets built, what it will and won't do, what I need from you, when it lands, and what it costs. One number, agreed before any code is written. No hourly meter running in the background.
Step 3
The build
Weeks, not quarters. You see it working while it's being built, not at the end — so if it's going the wrong way, we find out in week one instead of month three.
Step 4
Handover that holds
It ships to your hosting or mine, with the documentation to run it and the option to keep me on call. You own the code either way. Nothing is locked to a platform you have to keep renting from me.
Written up properly
Three builds, from empty repo to paying customers.
The questions
Asked before every first call.
Still deciding whether to build anything at all? How to build an MVP that ships walks through scoping a first version honestly.
My business is small. Is AI even worth it for me?
Often it isn't, and I'll tell you that on the call. AI is worth it when the same task is repeated many times a week and the cost of getting it slightly wrong is low — customer questions, drafting, sorting, searching. It's a bad idea when the task is rare, or when a wrong answer is expensive and nobody will check the output. Small doesn't disqualify you; the shape of the work does.
What does it cost?
It depends entirely on scope, so any number I put on a web page would be a lie. What I can commit to is the shape: fixed price agreed in writing before the build starts, quoted after a free call, not an open-ended hourly arrangement. If the honest answer is that your idea costs more than it will return, that's part of what the call is for.
Do I need my data organised first?
No — that's usually part of the job. Most small businesses arrive with the knowledge spread across a WhatsApp history, a few spreadsheets, a website, and one person's head. Getting that into a shape a model can use is normal work, and I've done it for a legal corpus, a supplier catalogue, and a job board.
Will it invent things and embarrass me?
That's the real risk, and it's an engineering problem, not a prompt problem. I ground answers in your documents and show the source, keep a human approving anything that goes out under your name, and design the system to say 'I don't know, here's a human' rather than guess. On FASL — legal drafting, where a wrong answer is expensive — that structure is the entire product.
Who actually does the work?
I do. There's no agency, no account manager, no junior the work is quietly passed to. You talk to the person writing the code, which is why the scope is honest — I'm the one who has to build what I promised.
Where are you, and does that matter?
Agadir, Morocco, working remotely with clients worldwide. Practically: I overlap the full European working day, most of the American morning, and I work in English, French, Arabic, and Darija. Every project on this site was delivered remotely.
What if I need a website or app before any AI?
Then that's what we build. Belmo, Magic Hands LMS, and Maroc Fournisseurs were product builds first — the interesting technology only matters if there's a working business under it. I'd rather build you the thing that makes money than the thing that sounds modern.
What happens after it's delivered?
You own the code and it runs on infrastructure you control. I can stay on for maintenance and iteration if you want it, but it's an option you choose monthly, not a dependency I build in on purpose.
Free call · no deck, no pitch
Tell me where the
hours go.
Thirty minutes, no charge, and an honest answer at the end — including “don't build this.” I read and answer every email myself, within 24 hours.
Email [email protected] →