AI your team uses on Monday morning.
Practical AI fluency for people who have a day job — marketers, sales, operations, finance. We have taught AI tools to thousands of people who had never used them, inside programmes that were measured on whether the work actually changed.
We train the people who use AI, not the people who build it. If you need engineers taken through model APIs and evals, we are the wrong call and will say so on the first one.
What the engagement gives you. The programme figures behind it — people trained, hours delivered, sessions written — sit with the client work below, where they evidence the capability instead of opening with our own output.
The training ends. The habits do not start.
A demo is not a skill
Somebody shows the tool doing something impressive on their own example. Nobody in the room has tried it on the thing they were doing that morning, so nothing transfers to Monday.
Generic prompts, generic output
Training built on a public prompt library produces work that reads like everyone else's. Your team needs the tool pointed at your catalogue, your tone and your customers, which means the session has to be built after we have seen them.
The wrong people in the room
A single all-hands session serves nobody: the finance team sits through content marketing, and the marketers sit through spreadsheet automation. Both leave having watched rather than done.
No one owns it afterwards
When the trainer leaves with the deck, the capability leaves too. The next cohort needs the same outside spend, and the second year costs what the first one did.
Built from a curriculum that already ran.
Everything below has been taught to a live cohort. None of it was drafted for this page. The AI content began as session seven of the Masar 3500 curriculum, where 4,100 Saudi home businesses were taught to write product content with AI tools — people who, three weeks earlier, had never sold anything online.
01Where AI actually helps you
A short, unglamorous audit of the work your team does every week, sorted into what a model does well, what it does badly, and what it must never touch. This is the session that saves the other twelve from being wasted.
02Writing that sounds like you
Product descriptions, listings, campaign copy and customer replies — drafted with a model, then edited against your own tone. Taught at scale on Masar 3500, where the output had to survive being read by a real buyer.
03Research and summarising
Getting a usable brief out of a long document, a competitor set or a spreadsheet, and knowing which parts to check before anyone acts on them.
04Images and product photography
What a model can and cannot fix in a photograph, and where the honest line sits for a product image a customer will buy from.
05Checking the output
How the failure modes look: fluent wrong answers, invented figures, a confident citation to nothing. Your team leaves able to spot all three.
06Data you must not paste
Customer records, contracts, pricing, anything under an NDA. One session, plainly delivered, and a one-page rule your team keeps.
Six sessions is the usual shape. The order changes once we have seen what your team does — the audit in session one is what decides it.
We hand over the guide, so the second cohort is free.
Most training leaves with the trainer. Ours is built to be run again by your own people, because that is how it had to work on a programme training four thousand families across a country we could not be in every week.
A trainer guide written for somebody who was not in the room when we built it — the script, the exercises, the questions people actually ask, and what to do when the answer is wrong.
On Masar 3500 the trainer guide reached version nineteen and the trainee guide version twelve. That is not a document that was written once and filed — it is one that was corrected after every cohort until it stopped producing questions.
Nothing. Running it again with your own team lead is the point. We would rather be called back to build the next thing than to repeat the last one.
Plotted against the alternatives.
Competitors are plotted at the bottom of their published range and we are plotted at the top of ours, which is the least flattering way to draw this table.
| Option | Year one, 40 people | Year two | Built around your work |
|---|---|---|---|
| A self-serve course library | From ~$15 per user per month | The same again | No — a catalogue everyone gets |
| A one-day external workshop | From ~$4,000 per day | The same again per cohort | Sometimes, at extra cost |
| Umbrella500, six sessions | Up to ~$9,000, fixed | $0 — you run it from the guide | Yes, after the session-one audit |
Sources: published per-seat pricing for the major course libraries, and published day rates for regional corporate training providers. We will put your real numbers into this table on the call, including the ones that do not favour us.
Six weeks, and the audit comes first.
The sequencing is deliberate: we look at the actual work before we write a single slide, because a session built on assumptions is the failure mode this whole page is about.
- Week 1
Watch the work
Two short sessions with the people who will be trained. We come out with the list of tasks worth automating and the ones that are not.
- Weeks 2–3
Build the material
Exercises written against your catalogue, your tone and your customers. You see the draft before anyone is booked into a room.
- Weeks 4–5
Deliver, in small groups
Delivered by Saleem Najjar and Noora Shanak, in Arabic or English. Groups stay small enough that everyone works on their own task rather than watching someone else's.
- Week 6
Hand over the guide
The trainer guide, the exercises and the data rule, so your own team lead can run it for the next intake without us.
Cohorts that had to produce something.
The last programme we taught on was published by the World Bank, with our co-founder as a named author. The teaching was the deliverable, and the report says what it produced.
Three times we are the wrong choice.
You need engineers trained
Model APIs, evaluation harnesses, fine-tuning. That is a different trade and we do not pretend to it.
You want a certificate scheme
If the outcome that matters is a credential rather than a change in the work, a course library will do it cheaper than we will.
You want it next week
We will not run a session before the audit. A generic day delivered fast is exactly the thing this page argues against, and we would be selling you the problem.
Our corporate track record is honest and short: this practice was built delivering national programmes for development clients, and we are taking it to companies now. Ask us for that detail on the call rather than reading it into a logo wall.
Counted from three named programmes — the World Bank in Jordan, Masar 3500 in Saudi Arabia and SPARK in Türkiye. These are development programmes. Our corporate track record is short, which is the honest shape of it — the skill being sold is the teaching.
The ones that decide it.
Which tools do you teach?
Whichever ones your team is allowed to use. We are not a reseller and have nothing to push — if your organisation has standardised on one assistant, the whole course is built on that one.
Arabic or English?
Both, and the material exists in both. The SPARK curriculum we wrote was bilingual across thirteen modules, so this is a thing we have already done rather than an offer.
How many people per session?
Small enough that everyone works on their own task. Beyond about twelve it becomes a lecture, and a lecture is what fails.
Who actually delivers it?
Saleem Najjar and Noora Shanak. The same two people you meet on the first call are in the room on the day.
What if our team is already using AI badly?
That is the most useful starting point there is. The session-one audit finds it, and the checking session is built around what it turns up.
Can we run it again ourselves?
Yes, and that is the design. You get the trainer guide and the exercises, and there is no licence on either.
Tell us what your team does all day.
That is genuinely the whole brief. We will tell you on the call which parts AI helps with and which parts it does not, including when the answer is that you do not need us.
Write directly: saleem.najjar@umbrella500.com
Opens your mail app with the details filled in. Nothing is stored on this page.