Automating training compliance with n8n: enrolments, attendance, follow-ups

Administrative load is eating your margin

The French National Quality Framework, annexed to decree no. 2019-565 of 6 June 2019, comprises 7 criteria and 32 indicators — 23 of which apply to a standard continuing-education provider. Each one requires observable, dated evidence.

The difficulty is not intellectual. It is repetition. For every session: send enrolment letters, confirm, chase inactive learners, collect attendance records, send post-session evaluations, chase non-respondents, issue certificates, archive evidence, aggregate satisfaction metrics.

Multiply by your annual session count. This is where the profitability of a mid-sized training organisation disappears — into low-value tasks, performed manually, with a non-zero omission rate that becomes a non-conformity at audit.

n8n can automate this chain. Here is how, concretely.

Why n8n rather than Make or Zapier

All three do the same job: connect applications and chain actions. Three differences matter for a training organisation.

Self-hosting. n8n can run on your own server. Your learner data — names, emails, sometimes health-related information for accessibility accommodations — then passes through no third party. Make and Zapier are hosted services, predominantly outside the European Union. For an organisation subject to GDPR and handling participant data, this is not a cosmetic difference.

Cost model. Make and Zapier bill per operation. An organisation that automates seriously generates tens of thousands of operations monthly, and the invoice scales with volume. Self-hosted n8n costs the price of a server — in the order of a few tens of euros per month, regardless of volume.

Technical freedom. n8n lets you insert code, call any API and manipulate data without connector constraints. When your LMS or industry-specific CRM has no native integration — common in the training sector — this is decisive.

The honest counter-argument: self-hosted n8n assumes someone can administer a server, manage backups and apply security updates. If nobody in your organisation can, Make is preferable to a poorly maintained n8n.

The five automations with the highest return

1. Enrolment letters and confirmations

Trigger: an enrolment is created in your CRM, or a row appears in your session spreadsheet.

Chain: n8n retrieves the participant and session data, generates the enrolment letter from a template, sends it by email with the programme and internal rules attached, creates the calendar event for the trainer, and logs the dispatch with a timestamp.

What it proves at audit: that the participant received information on objectives, methods and delivery conditions before entering training, with a defensible date.

2. Attendance records

Trigger: the start of each half-day session, on a schedule.

Chain: n8n sends each participant an individual attendance link with a single-use token, collects signatures into a database, automatically chases non-signatories after thirty minutes, and alerts the trainer if anyone remains missing.

The gain: reconstructing attendance sheets after the fact is one of the most painful exercises in audit preparation. Automated, it disappears.

3. Chasing inactive e-learning participants

Trigger: a daily progress check in the LMS.

Chain: n8n queries the LMS API, identifies learners with no login beyond a defined threshold, sends a personalised message referencing their actual progress, and escalates to the learning lead after repeated unsuccessful attempts.

What it proves at audit: active monitoring of engagement and adaptation of support — a structural requirement of the framework for distance learning.

4. Evaluations and satisfaction metrics

Trigger: end of session, then a three-month delay.

Chain: automatic dispatch of the immediate feedback questionnaire, chase of non-respondents after forty-eight hours, aggregation into a dashboard, dispatch of the deferred questionnaire three months later, and continuous calculation of satisfaction and completion rates.

The gain: evaluation response rate is the weak point of most organisations. Automated chasing typically moves it from around 40% to over 70%, with no human intervention.

5. Certificates and evidence archiving

Trigger: session closure.

Chain: generation of completion certificates as PDFs, individual dispatch, deposit into a storage area organised by session, and automatic assembly of the evidence file — enrolment letter, attendance, materials, evaluations, certificate — into a structure usable at audit.

What it proves: on audit day, you open a folder instead of reconstructing a session.

A reference architecture

A simple, robust setup, proven in mid-sized organisations:

CRM or enrolment database
        ↓
     n8n (self-hosted, VPS in France)
        ↓
   ┌────┴────┬──────────┬─────────────┐
   ↓         ↓          ↓             ↓
 Email     LMS       Storage      Dashboard
 (SMTP)   (API)    (S3 / WebDAV)  (PostgreSQL)

Technical points to watch.

Containerise. Run n8n in an isolated Docker container with capped resources. An application error must not take down the server.

Never version your secrets. API keys belong in n8n credentials or environment variables, never in plain text inside a node or a Git repository.

Back up your workflows. n8n exports workflows as JSON. Version them. A lost workflow is a week of rebuilding.

Protect the admin interface. Access to n8n must be restricted by IP allowlist and placed behind authentication. A publicly exposed n8n interface grants access to every connector you hold — and therefore to your learner data.

Log executions. If a dispatch is disputed, the timestamped execution trace is your evidence.

The next step: a RAG teaching assistant

Once the administrative layer is automated, the same infrastructure supports a conversational assistant that queries your own content: course materials, frameworks, internal FAQs.

The principle is retrieval-augmented generation. Documents are chunked, converted into vectors and stored in a database. For each question, the relevant passages are retrieved and supplied to the language model, which answers from them rather than from memory.

Two practical consequences. The assistant answers on your content, not on generalities. And it can cite its source, which limits fabricated statements.

With European models such as Mistral, or local inference via Ollama, the entire chain stays on your infrastructure. No teaching data leaves the European Union.

Where to start

Do not automate everything at once. The sequence that works:

1. Enrolment letters — the simplest workflow, the most immediate gain, and how you learn the tool
2. Evaluations and follow-ups — strong impact on your satisfaction metrics
3. Attendance records — more technical, but this is where the audit pain lives
4. Evidence archiving — once the first three run reliably
5. The RAG assistant — a separate project, to be treated as one

Allow one to two days for the first two, longer for attendance depending on your LMS.

Training on this architecture

MyRex trains training organisations on this full chain, in a three-day masterclass: AI learning design, automation and RS6776 preparation. Day two is devoted entirely to automating compliance processes with n8n. Your team leaves with three working workflows deployed on your own infrastructure.

We are not presenting a theoretical architecture. MyRex publishes and operates Next Call AI, a voice-agent platform running 24/7 in production, whose business orchestration runs precisely on self-hosted n8n.

MyRex SAS is registered as a training provider under activity declaration number 11757647975 with the prefect of the Île-de-France region. This registration does not constitute State approval.

See the programme →