The AI methodology for French SMEs and mid-caps
Échelle Junyr™ — 5 levels. 5-phase framework. 12 months.
↑ INDUSTEC case study: +182 % documented ROI in 9 months, 1.7 FTE freed.
By late 2025, 55 % of French small businesses use generative AI. Only 11 % extract measured value. The window is still open — but it is closing.
of French small businesses use generative AI (Bpifrance Le Lab, 2025).
extract real, measured value, visible in P&L (McKinsey).
median 12-month ROI for well-scoped AI missions in SMEs (50–500 employee panel).
of AI projects fail before deployment — it's a management issue, not a technology one (Gartner).
Méthode Junyr™ is the structured framework we operate as transition managers for French SMEs and mid-caps of 50 to 500 employees. It strings together diagnostic, scoping, foundations, pilot deployments and consolidation over 12 months for a total budget of €30k to €80k.
Échelle Junyr™ situates your company on a proprietary 5-level scale in under 10 minutes. None of the levels can be skipped. The golden rule: skipping the Orchestra stage costs 9 failures out of 10. The method traces a workable, measured, reproducible path.
European hosting, sovereign models (Mistral, OVHcloud, Scaleway), AI Act registry kept from phase 3. Sovereignty is no longer a marketing claim: it is an operational and legal risk topic.
Five levels, from Spectator to Pioneer. None can be skipped — skipping foundations leads to 90 % failure rates.
"We watch AI from a distance — are we missing something?"
The company is aware generative AI exists — the leader has heard about it, a few staff may have tried ChatGPT once. But no official tool, no identified business use case, no budget line. It is not a failure: it is a starting point. The danger is standing still while competitors move.
~50 % of SMEs"Are my staff using AI each in their own corner?"
Shadow-AI territory: uncoordinated individual usage, everyone subscribed to their own tool, sometimes on a personal account, with no framework. Gains exist but stay private — they vanish with the employee. The risks: data exfiltration, undetected hallucinations, personal dependency.
~30 % of SMEs"Is AI embedded in our processes, piloted and measured?"
2 to 5 priority business use cases deployed with a real project approach and measured. Usage charter validated by the exec committee, 2 to 4 business champions, data policy, quarterly AI committee, AI Act registry. This is where the documented 159 % median ROI sits.
~13–15 % of SMEs"Is AI a structural competitive advantage?"
AI becomes an attribute of the company: proprietary knowledge base (5 to 10 years of cleaned business archives), agents under supervision on defined perimeters, real-time AI Act registry, ISO 42001 governance under way, LLM gateway, mature FinOps. Competitive replication takes 18 to 24 months.
~2–3 % of SMEs"Are we defining the standard for our sector?"
AI-first workflows redesigned around agent capabilities. Autonomous agents measured in hours then days of operation, under asynchronous supervision. ISO 42001 certified, mature AI FinOps, AI literacy across all three tiers — leaders, managers, front line. The 2026 pioneers become the 2030 sector benchmarks.
< 0.5 % of SMEsAn SME at the Artisan level investing €200k in autonomous agents fails in 9 cases out of 10 — not for lack of technology, but for lack of foundations. The POC graveyard is filled with ambitious projects that skipped the Orchestra stage.
Designed for SMEs and mid-caps of 50 to 500 employees, operated by an external transition manager. Total budget: €30k to €80k.
Mapping of critical processes, data audit, leadership and exec posture, identification of resistances and champions. 12-point analysis grid (0 to 5). Deliverable: 15-to-25-page report presented to the exec committee, concluding on 3 to 5 priority use cases by impact/effort ratio.
For each selected case, a 2-to-4-page document: scope, named users, mobilised data, candidate tools, costed budget, expected ROI, risks, transition plan, success indicators. One hour of scoping saves twenty hours of deployment.
AI usage charter validated by the exec committee, documented data policy, three-tier training plan, target architecture defined, priority data structured. This is the invisible investment that separates the 11 % who reap from the 89 % who sprinkle.
Sprints of 4 to 6 weeks, one use case per sprint. Named user, pre-defined indicators, mid-point review, production transition review, monthly aggregated exec review. Strict rule: no transition without indicators met. Unprofitable case = stopped.
Industrialisation of cases in production, user upskilling, continuous measurement. Lasting governance: AI Act registry up to date, quarterly AI committee piloted by exec, year-2 roadmap. At 12 months the company sits solidly at the Orchestra level, with the Architect trajectory mapped.
The two disciplines that turn a POC into a production system.
Technology accounts for only 10% of an AI project's success (the BCG 10-20-70 rule); people and processes account for 70%. The Méthode Junyr™ industrializes this human factor through five levers: mandate and narrative, business champions, three-tier training, minimum viable governance + AgentOps, and adoption measurement. This is the move from Artisan to Orchestra — in other words, from POC to industrialization.
Every Junyr™ agent follows the “Plan → Execute → Verify/Test” pattern (three LLM requests per step): verification is a native step, not an after-the-fact check. Combined with the control plane (observability, human validation, auditable log), this is what makes an agent ready for industrialization.
None is technological. All can be neutralised — full chapter-by-chapter detail in the white paper.
The gadget syndrome — choosing a tool before formulating the problem.
The perpetual POC trap — 70 % of pilots stay dead-letter for lack of pre-defined go-live criteria.
The lone leader illusion — carrying the topic alone without 2 to 4 operational internal champions.
The instant-ROI mirage — bolting an assistant onto a poorly defined process amplifies the noise.
Data blindness — audit data quality before the use case, never after.
Case study — 9 months, documented ROI
Anonymised industrial SME, 78 employees, €23.5M revenue. Mission April 2025 → January 2026.
At diagnostic, AI existed only as individual usage (ChatGPT for follow-ups, LinkedIn, Mistral tests). Artisan level. Priority use case identified: technical quote drafting — 600 to 800 quotes/year, 2 to 4 hours per application engineer, totalling 1,800 to 2,400 hours annually.
275 hours saved per month on quote drafting.
1.7 FTE freed, redeployed onto post-sales follow-up and pre-sales site visits.
+18 % revenue on the commercial perimeter concerned; quote → order conversion went from 38 % to 44 %.
Documented 182 % ROI over 12 months. Total mission cost (€32k engagement + €8k licences + €4k internal) paid back by the 5th month.
Anonymised study. Results from a specific mission. ROI depends on sector context, initial maturity, team engagement and execution quality.
Regulation (EU) 2024/1689 phases in across 2026. For SMEs using AI systems: registry of AI systems, transparency vis-à-vis end users, traceability of AI decisions. Low cost if governance was put in place from phase 3; heavy otherwise.
Junyr™ brings together Méthode Junyr™ — the AI transformation framework for SMEs — Junyr Suite — the sovereign AI operating system for SMEs (email, calendar, documents, CRM and finance) with native MCP integration and RFC 3161 cryptographic timestamping (formerly Junyr Mail) — and Junyr Agents™ — the AI agent recruitment and management platform.
From diagnostic to deployment: discover how AI agents under mandate, supervised and auditable, take on your SME processes.
Discover Junyr Agents™60 minutes by video call, no commitment. Outcome: your position on the Échelle Junyr™ scale, 3 priority use cases identified, a 90-day roadmap drawn. The calibrated first step that separates started projects from intentions.