Roughly 25 USD per user per month on an annual commitment, about 250 MAD at today’s rate: that is the entry price of ChatGPT Team, the plan most Moroccan SMEs look at first when they decide to move off personal accounts. For thirty staff, that is around 7,500 MAD a month and 90,000 MAD a year, before VAT and before any currency movement.
The figure is a fair starting point, but it does not answer the question managers are actually asking. It prices shared access to a general-purpose model that knows nothing about your price list, your contracts or the contents of your ERP. So the useful comparison is not Plus versus Team. It is a general subscription versus an assistant wired into your own data.
The decision comes down to concrete lines: what you really pay, what the tool can do without a human in the middle, and what the CNDP expects from you in either case.
250 MAD per user per month: what ChatGPT Team actually bills
Published pricing for ChatGPT Team sits around 25 USD per user per month billed annually, a few dollars more on monthly billing, with a two-seat minimum. ChatGPT Plus stays near 20 USD per person per month. Plan names and amounts have changed several times since 2024, so check OpenAI’s pricing page on the day you build your budget rather than copying a number from an article.
What escapes the first budget is everything around the headline price. Billing is in dollars, so a plan built at 9.8 MAD to the dollar does not hold at 10.4. Digital services supplied by a non-resident provider fall within Moroccan VAT rules, and the exact treatment, including reverse charge, should be confirmed by your accountant instead of assumed. Payment is its own friction: foreign-currency allowances on online payments and Office des Changes rules make a recurring USD charge less trivial than a local subscription. Paying with the manager’s personal card solves it for three months and then becomes an accounting problem at year end.
One more thing gets underestimated. An annual commitment is billed per seat, not per use. Open thirty accounts and have twelve people use them seriously, and you still pay for thirty. A six to eight week pilot with ten volunteers, measured on named tasks (tender responses, meeting notes, translating sales material), costs far less than a full rollout followed by a quiet downsizing at renewal.
Plus, Team, Enterprise: what really changes behind each invoice
The gap between Plus and Team is not model quality, it is governance. Plus is a personal account: no admin console, no single invoice, no way to recover the history of someone who leaves, and privacy settings left to each individual. Team adds a shared workspace, central seat management, internal assistants shared between colleagues, and the commitment that your business content is not used by default to train models.
Enterprise is a different category: enterprise authentication, account provisioning, audit logs, data residency options, a firmer contract. The price is not public, it is negotiated with a salesperson, usually with a minimum seat volume and an annual commitment. For a forty-person Moroccan company that conversation rarely lands, and this is exactly where the comparison shifts: you are not choosing between Team and Enterprise, you are choosing between Team and something else entirely.
- Plus: individual use, no visibility for IT, fine for testing, not enough to roll out.
- Team: managed seats and internal sharing, but no enterprise single sign-on and no full audit trail.
- Enterprise: identity controls and traceability, negotiated price, sized for headcounts larger than the average SME.
What a subscription never solves inside a Moroccan SME
A subscription gives you an excellent writer and a solid translator. It does not give you your company’s memory. The available connectors assume your documents already live in a tidy cloud workspace with clean permissions. In a typical Moroccan SME, the useful information is spread across a file server, email attachments, an ageing ERP and three spreadsheets maintained by one person.
Take a deliberately hypothetical but ordinary case. A sales admin manager is preparing a tender response and needs the warranty terms applied to that client over the past three years. The answer is scattered across eight contract PDFs, two email threads and a spreadsheet. With a subscription, someone must find the files, upload them, then check the output. The gain is real, but it lands on the writing, not on the search, which was where the time went. With an assistant connected to the contract repository, that flips: the search is automated and the answer cites its sources.
Three further limits show up in almost every rollout. A subscription cannot write into your systems: no quote created, no ticket opened, no CRM record updated. It leaves no usable record of what was told to a client, which matters the moment an answer commits a price or a lead time. And on your internal facts it produces answers that sound right and are wrong, because it invents what it does not have. That is a difference in kind from an internal AI assistant grounded in your own documents, where the answer is constrained by the source.
Where your data goes, and what the CNDP expects from you
With any ChatGPT plan, processing happens on infrastructure outside Morocco. Business plans state that your content is not used for training by default, which settles one worry but not the legal one: personal data is transferred abroad, and Law 09-08 governs that transfer.
In practice compliance sits with you, not with OpenAI. You are the data controller: declaration of the processing, a legal basis, informing the people concerned, and the transfer formalities where they apply. Procedures and forms change, so verify the current obligations on the CNDP website before committing. The most commonly skipped step is internal. Without a written usage policy, your teams will paste CVs, medical files, bank statements and named client lists into a public chat and nobody will know. A one-page note listing prohibited data categories, naming an owner and explaining what to do after a mistake does more for your compliance posture than moving from Plus to Team. The obligations are laid out in our piece on CNDP compliance for ChatGPT and Copilot at work.
The real cost of a custom assistant, next to the subscription
A custom assistant is not a more expensive subscription, it is a project with a different cost shape. You prepare and clean the document base, build the retrieval index, reproduce your access rights (a salesperson should not be able to query HR files), ship an interface where people already work, host all of it, then supervise answer quality over time.
On the Moroccan market, a serious first scope (one business function, one document base, 20 to 50 users, sourced answers, no writing back into the ERP) generally lands between 120,000 and 300,000 MAD to build, over 6 to 12 weeks. Running it adds a recurring line: hosting, model consumption, supervision and fixes, often 2,500 to 10,000 MAD a month depending on volume and service level. Ranges by project type are broken down in our AI project cost guide for Morocco.
The arithmetic helps, but it misleads when read alone. Thirty Team seats cost about 90,000 MAD a year and nothing else. A 200,000 MAD build with 5,000 MAD a month of running costs is 260,000 MAD in year one, then 60,000 MAD after that. On cost alone, custom only wins if the time recovered or the mistake avoided exceeds the gap, which you work out the same way as any software investment: see our method for estimating how long an AI project takes to pay back. And in most cases we see, the two coexist: subscriptions for general writing, an internal assistant for anything touching your data.
Behind the “ChatGPT Morocco business” search, three separate decisions
Before signing anything, answer three questions honestly, with one real piece of work in mind rather than the general idea of using AI.
- Is the answer you need inside your own documents? If the value depends on your contracts, procedures, order history or catalogues, a general model cannot reach them reliably or repeatably.
- Does the assistant need to act in a system? Create a quote, open a ticket, update a client record: a subscription does none of that, it produces text someone retypes.
- Would a wrong answer leaving your company commit the business? Price, lead time, warranty, regulatory advice: if so you need cited sources, access rights and a record, not an anonymous chat.
No yes answers: stay on the subscription, train the teams, write the usage policy, stop there. One yes: the subscription plus one or two targeted automations is usually enough. Two or three: scope a custom assistant on a single measurable perimeter and keep the subscription for everything else. If you are still unsure whether your organisation can absorb the project, the AI readiness test before the quote is the right first step.
Frequently asked questions
Does ChatGPT Team make my company compliant with Law 09-08?
No. The plan improves internal governance (managed seats, no training on your content by default), but declaring the processing, informing the people concerned and covering the transfer outside Morocco remain your obligations as data controller. Check the applicable formalities on the CNDP website.
Can a Moroccan corporate card pay for a ChatGPT subscription?
Often yes, but it depends on your bank, the card type and the foreign-currency limits applied to online payments. Confirm the setup with your bank and accountant before opening thirty seats, and do not treat a manager’s personal card as a lasting answer.
Is it a choice between ChatGPT and a custom AI assistant?
Rarely. They cover different needs: the subscription handles generic text production, the custom assistant handles questions whose answer lives in your data. Plenty of companies keep a dozen paid seats and reserve the custom build for one business function.
How long until a first internal assistant is usable?
Plan on 6 to 12 weeks for a single scope, provided the document base exists and is reasonably clean. If documents are scattered, undated or duplicated, most of that time goes into data preparation rather than the model.
Is ChatGPT Enterprise worth negotiating at 50 users?
Ask for a proposal, it costs one call, but do not build your plan around it. Volume and commitment terms target larger organisations, and at 50 users that same budget often funds a custom scope that actually knows your documents.
Deciding with a priced scope
The right call does not come from a plan comparison, it comes from one costed use case: which function, which document base, how many users, which mistake is expensive today. Arrowlancer builds exactly this kind of AI work, from internal assistants and document analysis to sovereign AI hosted in Morocco, always starting from the narrowest scope that produces a measurable result. The details are on our services page, and if you want to compare your subscription budget against a real quote on a defined scope, send us your use case and the headcount involved.