Training providers: adding AI to your offer, where to start

The same question comes up in almost every sales meeting: „do you do AI?". Saying no costs you contracts, and improvising a one-day „discover ChatGPT" session won't hold up three months from now in front of a client who has already tried the tool on their own. Good news: adding AI to your catalog doesn't require hiring a €60,000-a-year expert, and it doesn't require rebuilding your offer either. What it does require is three decisions, taken in the right order: which audience you're targeting, which certification you attach to it, and who produces the content.

This guide gives you the order in which to make those decisions, the three possible methods with their real costs and timelines, a comparison table, and the risk areas to sort out before your next Qualiopi audit. The goal is simple: have an AI offer you can sell within a few weeks, not in a year.

Where to start: the four decisions, in order

Start with certification and funding, not with content. An AI module with no certification attached sells like any other one-day training course, and gets compared on a single criterion: price. That exact same program, tied to a certification registered with France Compétences, becomes fundable — and suddenly you're playing against a different set of competitors.

The four decisions, in this order:

1. Audience and use case. An executive assistant, a salesperson and a communications officer don't need the same skills. Pick two profiles maximum to get started, ideally ones you already train on other topics. You already have the prospect base, you know their constraints, and you'll sell faster.
2. The certification. Two certifications are directly relevant here today: RS7439 (AI-assisted content marketing) and RS6792 (AI and sales), both registered in the Répertoire Spécifique and therefore fundable. You can check each record, its certifying body and its assessment methods on the France Compétences Répertoire Spécifique. Important point: to deliver a certification from the Répertoire Spécifique, you have to be accredited by the certifying body. That's not a formality to handle at the end of the project — it's the first door to open.
3. Content production. This is what determines your launch timeline and your margin. Three options, detailed below.
4. The evidence. Attendance, individualization, assessment, ongoing monitoring. On a remote AI course, this is the topic that derails audits, because classroom habits simply don't transfer.

One last piece of context that's useful for your sales pitch: since February 2, 2025, the European AI Act has required organizations deploying AI systems to ensure a sufficient level of AI literacy among their staff. So your clients now have a regulatory reason to train their employees, not just an appetite for it.

Method 1: plug in a turnkey, certifying AI catalog

This is the fastest route: you get ready-made programs, fundable certifications and access to the tools, without hiring anyone or producing a single video.

That's exactly what Skilzy does — a French e-learning platform built to learn and actually use AI. The catalog includes more than 15 programs: creating images, producing video (UGC, ads, faceless channels, avatars, voice-over), generating music, automating tasks with n8n, writing good prompts, coding with AI. The two state-recognized certifications mentioned above, RS7439 and RS6792, are attached to the setup.

The difference that really matters for a training provider is the built-in AI Lab: learners use the real tools directly from the platform, with credits included. In practice, that means you don't have to make every participant buy six subscriptions at €20 or €30 a month, and you're not juggling trial accounts that expire halfway through the session. Better still, the work produced in the Lab becomes activity evidence you can use in your compliance file.

What you gain:

• A launch timeline of a few weeks, just the time needed for accreditation and your sales ramp-up.
• Zero instructional design to produce, zero filming, zero editing.
• Content that stays maintained when tools change their interface or their model, which happens several times a year.
• A fundable offer, and therefore a sales pitch that doesn't rest on your price alone.

For the catalog, certifications and multi-learner deployment side of things, the Skilzy offer for training providers is what to look at; if you want the detail of what a ready-to-sell catalog actually contains, the article on the turnkey AI training catalog gets concrete, and the one on how to offer AI training without hiring a trainer covers how to organize learner support.

The honest limitation: you don't control the pedagogical roadmap. If your core business is a very specific niche (AI applied to property surveys, say), you'll need to layer your own sector modules on top of the foundation.

Method 2: build your AI training in-house

This is the longest and most expensive method, but it's the only one that gives you content that's 100% yours, on sector expertise nobody else covers.

The rough numbers, so you can decide with your eyes open:

• Instructional design. For classroom delivery, budget 3 to 5 days of design for 1 day of training. For scripted e-learning, the usual market range is 20 to 40 hours of production per hour of finished content.
• Tool subscriptions. A serious program covering image, video and automation quickly pulls in four to six paid services, meaning €100 to €200 a month on the trainer's side — plus the separate question of learner access.
• Maintenance. Interfaces and models change every quarter. A module filmed six months ago often shows screens that no longer exist, and a participant who can't find the button checks out within three minutes.
• Certification. Putting together your own registration file for the Répertoire Spécifique is a project in its own right, with a skills framework, an assessment framework, and a review process that takes several months.

There's one case where this method holds up: you have rare sector expertise, an in-house trainer who's already comfortable with the tools, and enough session volume to amortize the production. Otherwise you pay twice — in cash and in months of delay.

Method 3: work with a freelance trainer or partner provider

This is the most flexible way to answer a one-off request, but it's also the one that crushes your margin and your ability to scale.

A freelance AI trainer typically charges between €700 and €1,200 per day excluding VAT for in-company delivery, more for highly sought-after profiles. You front nothing, you produce nothing, you invoice with a markup.

What it gives you:

• A fast response to a tender or a request from a long-standing client.
• The credibility of a practitioner who works with these tools every day.
• Fine-grained adaptation to the client's context.

What it costs you:

• A thinner margin on every session, with no economies of scale.
• Scheduling dependency: your best trainer is unavailable exactly when your client can free up their teams.
• Responsibility that stays yours. When you subcontract, you have to make sure your provider meets the requirements of the national quality framework, and document that check.
• No asset: the day the trainer stops, your offer stops with them.

Plenty of providers combine approaches: a shared asynchronous foundation for the fundamentals, plus a specialist for a high-value half-day sector workshop.

Comparing the three methods on the criteria that decide it

On timeline, start-up cost and ability to scale, method 1 wins in the majority of cases; method 2 is only justified for a sector niche.

The pragmatic reading: start with method 1 so you have something to sell this quarter, add method 3 for the sector workshops your clients keep asking for, and only move to method 2 once a need comes up often enough to justify the production.

What a Qualiopi auditor will look at in your AI offer

The four checkpoints are remote attendance, individualization, trainer competence, and monitoring of technological change.

• Attendance and completion. In asynchronous delivery, a scanned sign-in sheet isn't enough. You need activity traces: logins, progress, submitted work. On an AI course, the participant's output (an image, a video, an automation workflow) is legible, time-stamped evidence. What the auditor expects is covered in detail in the article on Qualiopi evidence for e-learning.
• Individualization. You have to show that the path for a complete beginner and the path for someone already self-sufficient aren't identical. An entry-level assessment plus differentiated tracks is enough — as long as it's written down and actually applied.
• Trainer competence. Keep evidence up to date: certificates, portfolio work, training completed. „He's been using ChatGPT for two years" doesn't count as evidence.
• Ongoing monitoring. Indicators 23 to 25 of the framework cover legal and regulatory monitoring, changes in occupations, and pedagogical and technological innovation. On AI, this is where you can genuinely shine: document dated monitoring, with sources and the impact on your content. The national quality framework published by the French Ministry of Labour gives the exact wording of each indicator.

One more habit worth building: state in writing, in your program and your terms, which tools are used, what data learners are not allowed to enter into them, and how you handle generated content. Your clients' procurement teams and data protection officers are now asking.

The shortest path to launching this quarter

Adding AI to your offer comes down to three moves: pick an audience you already know, attach a fundable certification, and produce no content you can get ready to sell. Everything else — sector workshops, in-house modules, upskilling your trainers — gets built afterwards, on top of a catalog that's already running.

If you'd rather get yourself up to speed first, the guide to learning AI as a beginner is a good place to start.