Junyr.eu
Open-enrolment session · Day 2 · Five dates 2026-2027
Enterprise Agentic Processes training
Steer autonomous execution across every function of the company, with software engineering as the guide. From autonomous coding to automated prospecting: the same operating grammar. For technical and non-technical profiles alike. This day runs on five dates between September 2026 and February 2027: it always runs, whatever the number of registrations.
Open-enrolment calendar 2026-2027
Every session always runs, whatever the number of registrations. This Day 2 can be booked on its own or as a two-day course with Day 1, AI Strategy & Transformation, held the day before each date above.
Your trainer: Paul-Antoine Tual
Founder-trainer, Croissance & Transitions · Founder of Junyr
Mines Nantes engineer, qualified lawyer, developer since 1993. AI Transformation Leader and turnaround manager specialising in AI-led transformation for SMEs and mid-sized firms, he forged the Junyr Method™ in the field, building and running Junyr Suite, a business platform hosted in Europe and independent of the major AI labs.
AI Decrypted (his blog, in French): 33 essays sorted by theme, read here ↗
The thesis of the course. When AI agents produce artefacts (code, sales proposals, campaigns, reporting), value concentrates upstream and downstream of production: framing, review, sign-off, decisions. Software engineering has fifty years of experience delegating execution to systems: specification, review, versioning, testing, reversible deployment, observability. This course teaches how to transpose that grammar to any function in the business.
Guiding principle of the Junyr Method™: the agent proposes, the human commits.
The pedagogy
4 stages, in every sequence
Anchor
A concrete demonstration on real tooling (MCP agents, the Junyr roadmap module): you see the method work before hearing about it.
Experiment
Workshops work exclusively on the real processes participants bring with them.
Formalise
The debrief turns the experience into a transposable principle, the grid fills in discipline by discipline.
Transfer
Every sequence produces a written record (plan, checks, register) you can replay in-house the next day.
The framework
The 3 tiers of agentic maturity
Assisted
AI produces on instruction, the human steers every step: proposals, analysis, first drafts.
Supervised
The agent executes inside the information system's own tools (MCP), with human sign-off before any committing action.
Guided autonomous
Workflows, skills and scheduled tasks run unattended, under guardrails and a log.
Moving from one tier to the next is never a question of tooling: it is a question of conditions being met.
The mechanics
Steering the machine: context, tokens, orchestration
Context management
The context window is the agent's working memory: limited, costly, decisive. Compact plans (1 to 3 screens), only the context a step actually needs, durable memory externalised (logs, registers), detection of a saturated context.
Token management (FinOps)
The token is a production cost: budget per mission, cost per useful result, model choice matched to how critical a step is, economic stop rules: an agent stuck in a loop is a cost that keeps running.
Orchestrating long-running tasks
A multi-hour mission breaks down into coordinated sub-agents: parallel decisions, serialised commits, human checkpoints, a log, incident recovery without replaying everything.
The transposition
From autonomous coding to every autonomous mission
Autonomous coding is the founding case: the method was forged on it. But the same grammar (decision-complete plan, human gate, orchestration, checks, log) steers any autonomous mission in the business.
| Autonomous mission | What the agents do | Where the human gate sits |
|---|---|---|
| Autonomous coding (founding case) | Multi-step implementation, tests, cross-review between sub-agents | Plan approval, promotion to production |
| Automated prospecting | Sourcing, qualification, CRM enrichment, personalised sequences, follow-ups | Target and message sign-off, review on the first sends then sampling |
| Market and competitive watch | Scheduled multi-source collection, summaries, alerts on signals | Defining the signals, review before distribution |
| Collections and follow-up | Deadline detection, graduated reminders, tracking commitments | Approved reminder plan, escalation past a threshold |
| Tier-1 customer support | Qualification, knowledge-base answers, routing | Sign-off on committing replies, escalation outside the playbook |
| Content and SEO/GEO | Channel adaptation, optimisation, scheduling | Signed brief, batch promotion |
| Reporting and management control | Collection, consolidation, anomaly detection, summaries | Defining the checks, review of anomalies |
| Recruitment sourcing | Candidate search, shortlisting, outreach messages | Sign-off on criteria and every outreach |
The method
The roadmap-driven flow, generalised
Inbox
Everything comes in, everything is explicitly triaged
Decision-complete plan
1 to 3 screens, actionable without re-exploring
Human gate
No committing execution without approval
Serialised execution
Only one "writer" commits at a time
Tooled verification
Executable checks, independent of the agent
Log
Append-only: the log is the record of truth
Learning objectives
7 objectives, matched to deliverables
Transpose the disciplines of software engineering to steering an agentic process in any function.
Position every process on the 3 tiers and define the conditions to move up.
Design a roadmap-driven flow: inbox, plans, human gate, serialised execution, checks, log.
Steer a long autonomous mission: context, token budget, sub-agent orchestration with incident recovery.
Diagnose process and data debt, and sequence it by criticality.
Steer a portfolio: value × effort prioritisation (ROI under 12 months), net gain, cost control.
Set up governance: agent register (AI Act art. 50, GDPR), the shift from producer to orchestrator.
Going further, in-house
The reference course (4 days / 28 h, in-house)
Learning by doing: software engineering as the grammar of delegation
- The productivity paradox: why artefact review is dying, and what replaces it
- Anchoring workshop: have AI produce a complete artefact and experience the assisted tier
- The transposition grid, worked through discipline by discipline
Designing the flow: roadmap-driven applied to a business function
- The 4 Junyr rules: plan by default, human gate is absolute, single writer, the log is the record of truth
- Through-line workshop: 2 decision-complete plans on your own process
- Live demonstration from inbox to log, on real tooling
Cleaning up the foundations: debt, checks, and the economics of the machine
- Process and data debt: surfacing it, sequencing it, paying it down with AI
- Executable gate checks (the business's own "regression tests")
- Context and token management: compact plans, externalised memory, mission budget
Orchestrating and steering: autonomous missions, portfolio, governance
- Orchestrating long-running tasks: demonstrated on autonomous coding and on automated prospecting
- The 5 non-negotiable conditions for autonomy: auditing honestly, knowing how to say no
- Roadmap: prioritised portfolio, orchestration plan, 30/90-day milestones, agent register
Practical information
Audience, follow-up, pricing
Audience & prerequisites
AI transformation project leads, transformation directors, PMOs, CIOs, function heads. No programming skills required. A 1-hour positioning interview beforehand.
Assessment & follow-up
Continuous assessment through deliverables, final multiple-choice test (66% pass mark), certificate of attendance. 90-day access to the lab.junyr.app lab and the resource library, individual coaching at day 90.
Pricing & funding
€690 (excl. VAT) per participant per day in an open-enrolment session, 8 participants maximum, the session always runs. OPCO funding possible: quote on request. Training is accessible to people with disabilities, contact the referent beforehand. Access timeframes, full accessibility details and cancellation terms: see the practical information page.
This course can also be run on your premises.
Either the taster day or the full 4-day course, tailored to your real processes, up to 12 participants, on a date of your choice.
Further reading
Articles that set the scene
Croissance & Transitions is a training organisation certified Qualiopi, certificate n° ATA 1966 2026 issued by ATALIA Certification, valid from 28/07/2026 to 27/07/2029. The quality certification was awarded for the following category of action: training actions. OPCO funding possible: quote on request.
Croissance & Transitions, SASU, French training-activity registration (NDA) 93830888583 (PACA region). Trainer: Paul-Antoine Tual. Quality and disability referent: Paul-Antoine Tual, paul@junyr.eu, +33 7 60 93 50 20. Programme sheet last updated 26/09/2026, open-enrolment calendar 2026-2027 above. Delivered in person, eligible for OPCO funding, quote on request for funding arrangements.
Book your seat at the next session.
20 November 2026, or one of the three dates that follow through to February 2027. Quotes within 48 working hours.
Want to frame the strategy at board level first? See Day 1, AI Strategy & Transformation →