Writing a business plan that holds up in Morocco: method and AI assistant

Most of the business plans we receive fail at the same point, and it isn't the point their authors worry about.

Key points

• The Investment Charter can cover up to 30% of a project's total amount as a direct grant.
• The SME scheme targets projects between MAD 1 million and MAD 50 million, from companies turning over MAD 1-200 million.
• The portal's AI assistant writes the structure and the market framing. It replaces neither your numbers nor your knowledge of the ground.

Where files actually stall

Not the vision, not the market, not even the product. It's the handover from narrative to numbers.

A file describes an opportunity for twelve pages, then presents a funding need without showing how it was calculated. The investor can do nothing with that. They aren't arguing about the amount: they have no handle to argue with.

The rule is simple. Every number in the plan must answer "where does this come from?" in one sentence. A MAD 2.4 million requirement breaks down into line items, each item into a quote or an explicit assumption. If the breakdown doesn't exist, neither does the number.

What the Investment Charter obliges you to know

The Moroccan framework has changed, and it rewards precise files. Framework law 03.22 introduced a direct grant system reaching up to 30% of the investment amount, something the 1995 charter never offered.

A scheme specific to micro, small and medium enterprises was adopted by decree no. 2.25.342, published in the Official Bulletin on 3 July 2025, then completed by four further decrees in November 2025 (CRI Fès-Meknès, 2025).

The eligibility criteria shape your plan:

• Annual turnover between MAD 1 million and MAD 200 million.
• Investment project between MAD 1 million and MAD 50 million.
• Less than 25% of capital held by a company turning over more than MAD 200 million.
• No public shareholder.

If your project sits inside those bounds, your business plan should make that legible on page one. An institutional reader who has to hunt for your forecast turnover in an annex will file the dossier elsewhere.

The scheme also provides support from the Regional Investment Centres, and it was designed to stay reachable for young companies without an extended trading history.

What the AI assistant writes, and what it never will

The drafting assistant in the portal runs on Mistral models. It does three things well.

The structure. A complete plan, in the expected order, with no missing section. That's mechanical, and it's exactly what a machine does better than a founder at 11pm.

The activity description. From a few inputs it produces a clear, sober presentation that avoids the promotional register investors distrust.

Market framing. It organises the analysis, sets up the right categories, and flags missing angles.

What it won't do:

• Invent your numbers. Any it offers are placeholders to replace. A generated forecast left as-is is spotted immediately, and it destroys the credibility of everything around it.
• Know your ground. Your customers' payment terms, the real seasonality of your area, the availability of skilled labour in your region: none of that can be inferred.
• Excuse you from proofreading. Each generation costs credits, which is also an incentive to think before running it again.

Personal data handling follows the Moroccan framework; our commitments are set out on the CNDP compliance page.

A five-pass method

Writing a business plan in one go doesn't work. Build it in layers.

1. Structure, generated by the assistant. Twenty minutes.
2. Numbers, by hand, in a separate spreadsheet. It's the longest pass and the only one that really counts.
3. Narrative, rewritten to fit the numbers rather than the other way round. When the two contradict each other, the numbers are almost always right.
4. Risks, stated plainly. A file with no risk section isn't reassuring. It's incomplete, and the reader knows it.
5. A read by someone who doesn't know the project. Whatever they don't understand, an investor won't either.

Once the plan exists, the pitch deck follows in an hour. And the pitch deck counts directly towards your trust score.

The four costing mistakes we see most

They turn up whatever the sector, and they all take half a day to fix.

Revenue as a share of market. "The market is worth 400 million, we're targeting 2%, so 8 million." That reasoning describes no actual sale. Start again from the number of customers, the average basket and the purchase frequency, in that order.

Forgotten working capital. Plenty of plans finance equipment and nothing else. But weeks pass between buying materials and getting paid, and those weeks have to be funded. It's the leading cause of an emergency raise six months after launch.

Understated payroll. Social contributions, holidays and the time it takes someone to learn a role almost never appear in the forecasts we receive. Count employer cost, not net salary.

Optimistic payment terms. A Moroccan institutional client rarely pays in thirty days. Build the plan on your sector's real terms, then check the cash position holds. If it only holds under ideal terms, it doesn't hold.

What an institutional reader looks for first

A relationship manager at a Regional Investment Centre or a bank doesn't read a file front to back. They look for four things, in this order, and give up if they can't find them.

1. The amount requested and what it pays for, on one page.
2. The founder's own contribution, in value and as a percentage. A file with no personal stake calls for an explanation, not silence.
3. The break-even point, expressed in volume and in date, not only in dirhams.
4. The security offered, or its absence, stated plainly.

Putting those four in the first two pages doesn't make a file stronger. It makes it legible, which is the condition for being judged on its strength.