The MATIA Method™ Scale: where do you really stand?
Most AI maturity models borrow their framework from large corporates: they measure technological sophistication, model governance, data culture (concepts that do not carry the same meaning in an SME with 80 staff as they do at Saint-Gobain). The MATIA Method™ Scale corrects this flaw. It was designed for SME business leaders, with a logic of rapid self-diagnosis and action. The 5-level framework is consistent with the 2025-2026 “golden standard” frameworks (Gartner 5 levels, PwC AI-Native, Microsoft Agentic L100→L500) while remaining specific to French SMEs.
Five levels, five metaphors, five questions:
The Spectateur
“AI: we're watching from a distance. Are we missing something?”
No structured AI use. The business leader knows generative AI exists; a few staff may have tried ChatGPT once in a personal capacity; no official tool, no business use, no budget line.
Around 50% of French SMEs
The Artisan
“Are my people using AI, each in their own corner?”
Individual Shadow AI. ChatGPT, Claude or Mistral used on personal accounts, without a framework, without measurement. Gains exist but are private, fragile, invisible on the P&L. Risk of data exfiltration.
Around 30% of French SMEs
The Orchestre
“Is AI embedded in our processes, steered and measured?”
2 to 5 business use cases in production, measured monthly by the executive committee. Usage charter, internal champions, documented data policy. This is where the documented 159% median ROI on AI engagements resides.
13 to 15% of SMEs today
The Architecte
“Is AI a structural competitive advantage?”
AI is no longer a project: it is an attribute of the business. A proprietary knowledge base, autonomous agents under supervision, a real-time AI Act register, an LLM gateway, ISO 42001 initiated. A structural competitive advantage.
2 to 3% of French SMEs
The Pionnier
“Are we setting the standard for our sector?”
AI-first workflows and hybrid human-plus-agent teams in production, ~100% of eligible staff augmented, AI oversight at senior-management level. Demonstrated value on the P&L, “Owned Intelligence”, certified ISO 42001. Data sovereignty and agentic interoperability (A2A/MCP) as constraints require. The “golden standard 2030” target detailed below: the Pionniers of 2026 are the benchmarks of 2030.
Fewer than 0.5% of French SMEs
The 9 dimensions of maturity, level by level
Overall positioning breaks down into nine dimensions. Level 5 describes the “golden standard 2030” target of the Pionnier, achieved today by fewer than 0.5% of SMEs. The numerical thresholds (≈100%, < 1%, > 99.5%) are apex targets, not market averages.
| Dimension | Spectateur | Artisan | Orchestre | Architecte | Pionnier: 2030 target |
|---|---|---|---|---|---|
| Sovereignty & control of models | Consumer tools, data not controlled | Personal accounts, exposed data | Classified data, EU/SecNumCloud for sensitive data | Sovereign cloud + RAG, targeted fine-tuning if justified | Sovereign AI: RAG + targeted fine-tuning, local/on-premise option, reversibility |
| Resilience & imperviousness to the splinternet | Total dependence, risk ignored | Single provider, no plan | Basic DR/BCP, data replicated in EU | Multiple providers, proven reversibility | Multi-cloud redundancy + on-premise failover: continuity even in a splinternet event |
| Augmented staff | 0% official | ~10-30%, unsupervised | ~30-50%, priority roles | ~70% of eligible roles | ~100% of eligible staff, core orchestrating agents |
| Cultural adoption & AI literacy | None | Self-taught, uneven | 3-tier training (> 30%) | Rolled out, role-specific | ~100% culturally adopted, continuous learning, leaders augmented |
| AI governance | None | None (Shadow AI) | Charter, quarterly committee, AI Act register, prior CSE consultation for internal AI (≥50 staff, Art. L.2312-8) | Monthly executive committee, ISO 42001 initiated, structured staff-rep dialogue | Golden standard: certified ISO 42001, Responsible AI, senior oversight, permanent staff-rep dialogue |
| Technical debt | n/a | Hidden (ad hoc scripts) | Tracked, backlog identified | Controlled (< 10%) | < 1%, continuous purge via ultracoding |
| Security (security score) | Attack surface not controlled | Possible leaks (Shadow AI) | Access control, secret management | NIS2-ready, > 95% | > 99.5%, PQC-ready, zero untracked incidents |
| Workflows & agents | None | One-off assistants | 2-5 measured use cases | Agents supervised (defined scopes) | AI-first, multi-agent orchestration in production, agent-ready IS (A2A/MCP interop) |
| Value & FinOps | None | Costs not measured | ROI tracked, budget in P&L | Mature FinOps (gateway, cost/task) | Value on the P&L, “Owned Intelligence” |
Full detail of the 5 levels and the 12-point self-diagnostic grid: see the dedicated article or chapter 2 of the white paper.
The 5 fatal mistakes that doom AI projects
80% of AI projects were failing (RAND, 2024); by mid-2026, 43% of major initiatives are still judged doomed to fail (HCLTech). Five mistakes are responsible:
- The gadget syndrome: choosing a tool before formulating a problem.
- The perpetual-POC trap: 70% of AI POCs remain dead letters.
- The lone-leader illusion: carrying the matter alone means nobody else uses the tool.
- The mirage of instant ROI: AI does not create value straight off the production line, but within redesigned processes.
- Data blindness: without reliable fuel, the most sophisticated engine will not start.
None is technological. All are mistakes of execution discipline. This is exactly what the MATIA Method™ approach locks down.
The 5-phase approach
Phase 1: 360° diagnostic (weeks 1 to 3)
Process mapping, data audit, the stance of the business leader and executive committee. Deliverable: a 15- to 25-page report, presented to the executive committee, concluding with 3 to 5 priority use cases ranked by impact/effort ratio.
Phase 2: Scoping the use cases (weeks 4 to 6)
For each case retained: scope, target users, data involved, candidate tools, budget, expected ROI, risks. One hour of scoping saves twenty hours of deployment.
Phase 3: Preparing the foundations (weeks 7 to 12)
Usage charter, data policy, three-tier training plan, target architecture, structuring of priority data. By the end of this phase, the company has become capable of making its deployments succeed.
Phase 4: Pilot deployments (weeks 13 to 26)
Sprints of 4 to 6 weeks, one use case per sprint, an end user identified from the outset, success indicators defined before launch, a monthly executive committee review.
Phase 5: Consolidation and governance (months 7 to 12)
Industrialisation, lasting governance (AI Act register, quarterly AI committee), year 2 roadmap.
A case study, in figures: INDUSTEC
An industrial SME, 78 staff, €23.5 million in revenue. At Artisan level (MATIA-2) at the April 2025 diagnostic. At 9 months: 275 hours saved per month on drafting quotes (1.7 FTE freed up), +18% in revenue across the commercial scope concerned, a documented ROI of 182% over 12 months. Confirmed shift to a stabilised Orchestre (MATIA-3).