Ardoise vs the Alternatives: The Evaluation Framework

The training platform market has structured itself at speed. Almost every vendor now displays “AI-powered” on its homepage, and buyers sit through demos that all start to look alike. Three weeks in, the decision often comes down to list price or logo recognition. That is not carelessness. It is what happens when nobody supplies an evaluation framework.

This article supplies one. It explains where Ardoise sits, on which criteria, and in which situations another solution will be the better choice. You do not need a pitch. You need to know what to test.

Artificial intelligence has become a checkbox. It tells you nothing about what the platform actually does, or at what level it intervenes.

The useful distinction is architectural. An AI-native learning platform is one designed around artificial intelligence from the start: the AI reaches the learning data and acts inside the learning flow. An AI-augmented platform has integrated AI functions into an existing architecture, where the AI is called from outside the flow. Both approaches are legitimate and serve different needs. They behave very differently the moment an internal process changes.

Four questions separate them in a demo. We call them the four demo questions, and they work on any vendor, including us.

Recommending a module because the learner ticked “beginner” at enrolment and rebuilding a learning path according to their role, actual level and field context are two distinct mechanics.

Take two people going through the same negotiation training. Thomas sells on the shop floor, in luxury retail, to a customer who has already decided to buy and is hesitating over the model. Léa sells a three-year contract to a buying committee of six, four of whom will never meet her. Same topic, same skill objective, two realities with nothing in common. A platform that serves them the same module has satisfied a requirement, not built a skill.

Test in the demo: ask to see two paths generated for two real roles in your organisation, on the same subject. Compare them side by side. If the difference comes down to module order, you are looking at catalogue filtering.

This is the most discriminating of the four questions, and the least often asked.

In a conventional setup, a procedure updated in internal documentation has to be manually worked back into the training: identifying affected modules, rewriting, validation, republication. The delay runs into weeks. Meanwhile, teams are training on a stale version.

Test in the demo: ask for a live walkthrough of that cycle, not a general answer about AI. How long between a source document being updated and the learning path being updated? Which sources can be synced? What human validation is still required?

If every change to a scenario goes through an external vendor or an internal technical resource, total cost of ownership no longer bears any relation to the licence price. It is the most systematically underestimated line item in tenders.

Test in the demo: ask to modify a piece of content yourself, during the session, without help from the consultant in the room. The response to that request is already information.

Completion rates measure attendance. They do not measure whether anyone works differently. The gap is not marginal: studies put the share of training that gets applied back on the job at around 12%.

The indicators that count are skills retention at 90 days, transfer observed in the work situation, and movement in the business KPIs tied to the programme. Ask for client cases with real numbers in contexts comparable to yours, including the measurement methodology. A named testimonial without a figure is not proof, and this sector produces a lot of them. For the full measurement protocol, including how to set a baseline and what to track at 30, 60 and 90 days, see our guide to measuring the operational impact of training.

Ardoise is an AI-native learning platform, built in Paris and designed around artificial intelligence from the ground up rather than retrofitted with it. Four distinct functional layers, working together.

Ardoise Intelligence® syncs learning paths with the company's internal processes and documentation, so that each business change requires less editorial work. Each course is built on a structured knowledge base assembled from your existing materials: documents, presentations, PDFs, legacy SCORM packages and approved web content. Replace a source document once and every learning path built on it reflects the change, with no per-course rework. Human control sits at the gate: your team decides what enters the knowledge base, and once a source is approved and replaced, propagation is automatic. Pre-publication testing and regression testing are available as additional safety layers.

The ASPECT® methodology structures each experience around skills acquisition in context: simulations, contextualised scenarios, expert conversations. It is the pedagogical framework that ties every sequence to an identified work situation.

Ardoise Studio® is the no-code authoring tool. Training teams create, update and manage AI-native courses in minutes rather than months, without coding or design skills.

Ardoise Collection® and Ardoise Bespoke are two entry points. Collection offers curated, ready-to-deploy AI courses contextualised to your organisation and personalised to your people. Bespoke is fully custom creation, aligned with your content, teams and deadlines. Speed of deployment on one side, deep strategic alignment on the other, under one architecture.

The SCORM Connector is the newest addition to the architecture, and it exists because of a pattern we kept hearing from L&D teams: they wanted Ardoise experiences inside the LMS they already run, but every integration route carried a toll. An API project means months of IT paperwork and validation queues. Flat-file user management means permanent maintenance. The SCORM Connector removes the toll. Ardoise courses deploy into your existing LMS as standard SCORM packages, the format virtually every LMS already accepts, while the experience itself stays live, contextualised and updated on the Ardoise side. Completion and scores report back into your LMS as they always have. Your LMS remains the system of record, and your L&D team gains an AI-native capability on its own authority, without opening an IT project to get it.

Established LMS platforms that have added AI functions. Mature platforms with real, documented strengths: proven integrations, broad catalogues, stability in large-enterprise environments, established partner ecosystems. The question to put to them is architectural. On the four points above, and especially on the update cycle, ask for the live demo rather than the product sheet.

Generalist content libraries. Low entry price and considerable catalogue depth. For personal development, cross-functional upskilling or broad acculturation, the value for money is hard to beat. Their scope ends where business specificity begins: no integrated authoring, no personalisation by internal role, no sync with your processes. These are libraries, and they fill that role very well.

Platforms designed natively around AI. A recent segment, which includes Ardoise. The argument is architectural: the AI reaches the learning data and acts inside the learning flow, rather than being called from outside. The segment's drawback is symmetrical. These are younger companies, with fewer years of production track record and less extensive integration ecosystems than platforms that have been in market for fifteen years. On a large-enterprise project, that is a point to investigate, not to ignore.

No comparison table published by a vendor will help you decide. Here instead is the grid to fill in during your demos, putting the same question to every candidate.

On the last four rows, a response time longer than a week is itself an answer.

Our own answers, since we are asking you to demand them from everyone. SCORM 1.2 and 2004 import, xAPI, SAML 2.0 and OpenID Connect single sign-on, and HRIS connectors via middleware alongside a REST API are live today. A SCORM connector for delivering Ardoise courses into your own LMS is available on request. Data location, transfers, reuse and reversibility are set out in our Data Processing and Protection Policy, downloadable from ardoise.ai.

The moment a platform processes learner data, it acts as a processor. For UK organisations that means the UK GDPR with the ICO as single supervisory authority; for Nordic organisations, the EU GDPR (with Norway and Iceland covered through the EEA) plus consultation duties with employee representatives that most vendors have never heard of. Either way: a clear data processing agreement, minimisation of the data collected, and traceability of automated processing.

One point is specific to AI and often poorly anticipated: if your learners' data feeds the training of the vendor's models, your company carries a liability it has generally not assessed. Require the contractual commitment not to reuse. Ardoise never uses customer data to train foundation models.

A learning platform that profiles, recommends and can influence decisions about learners triggers a data protection impact assessment in cases of large-scale profiling. Automated decisions must remain under effective human control. In the UK, the Data (Use and Access) Act 2025 replaced the old Article 22 with a regime carrying explicit safeguards: telling data subjects the decision was automated, a route to challenge it, and a human review mechanism. Ardoise meets all three: learners always know when they are interacting with AI, they can contest an automated outcome and request human review, and no significant decision about a person is made without human involvement.

On hosting, the UK line is worth stating precisely: the EEA sits on the UK adequacy list, so transferring UK learner data to a French or other EEA data centre needs no additional mechanism. Transfers to countries without UK adequacy require the International Data Transfer Agreement or the UK Addendum to the EU standard contractual clauses, plus a documented transfer risk assessment.

Ardoise hosts customer data in EU data centres, France primary, with encrypted backups held in EU regions. Some sub-processors, AI inference among them, process minimised and pseudonymised data outside the EEA under EU standard contractual clauses, with the UK Addendum applied to UK data. The full sub-processor list and transfer safeguards are in our Data Processing and Protection Policy, and Ardoise has a designated Data Protection Officer.

For regulated sectors, ask which specific regime applies to your deployment: in UK healthcare the NHS Data Security and Protection Toolkit, in financial services the FCA's outsourcing and operational resilience requirements, or your sector equivalent. ISO 27001, SOC 2 Type II and Cyber Essentials Plus do not constitute GDPR compliance, but they signal a level of audit you are entitled to ask about. Ask any vendor which certifications it holds itself and which it inherits from its infrastructure providers, because the two are routinely conflated.

Ardoise is built for organisations training large frontline teams, continuously, on complex and shifting processes. Luxury, retail, manufacturing, financial services: sectors where content evolves faster than conventional production cycles and where on-the-job performance is measurable.

Three situations in which another solution will serve you better.

Your main need is a low-cost generic catalogue for standard compliance training. A content library will be cheaper and sufficient.

Your catalogue is stable and your content changes rarely. The production cost differential, which is the main return-on-investment lever for an architecture like ours, will not materialise.

Your organisation is not in a position to invest time in contextualising content. The architecture amplifies what is well designed upstream. It does not fix what is not.

We would rather say this before the pilot than six months into it.

What is an AI-native learning platform? An AI-native learning platform is designed around artificial intelligence from the ground up: the AI reaches the learning data and acts inside the learning flow, adapting paths, scenarios and assessments to each learner's role and context. This differs from an AI-augmented platform, where AI functions such as content generation or recommendations are added onto an existing architecture and called from outside the flow.

What should you test in an AI learning platform demo? Four things, live: two distinct learning paths generated for two real roles on the same subject; the full cycle from a source document changing to the learning path updating; modifying a piece of content yourself without help from the vendor's consultant; and a written client case with real numbers and the measurement methodology behind them.

What data protection questions apply to AI learning platforms? Where the data is hosted and under which legal entity, whether learner data is used to train the vendor's models, and the exit procedure at contract end, all in writing. Platforms that profile learners at scale trigger a data protection impact assessment, and automated decisions must remain under effective human control, with a route for the learner to challenge them.

This grid does not replace a pilot. It stops you choosing by default, on brand recognition or list price.

Then test on a real scope: one team, one role, one internal process currently changing. Ask for a demo built around a business case from your own sector, with your content and your constraints. That is the point at which the criteria in this article become measurable.