The Junyr Method™ Scale: where do you really stand?
The Junyr Method™ is a proprietary decision framework for SMEs that assesses observable capabilities across five cumulative levels, helping leaders connect individual AI use to processes with clear responsibilities, controls and measured results.
- Capabilities must be supported by evidence.
- Budgets and delivery dates are scoped to each organisation.
The Spectateur
“AI: we're watching from a distance. Are we missing something?”
The Spectateur level describes an organisation where AI use is absent or insufficiently visible to support a shared decision about tools, data and responsibilities.
- Inventory the tools and data already in use.
- Agree initial rules and assign responsibility.
The Artisan
“Are my people using AI, each in their own corner?”
The Artisan level describes individual or local AI practices whose usefulness needs to be assessed on a defined business process before wider adoption.
- Select a recurring process with a business owner.
- Establish a baseline and evaluate a bounded pilot.
The Orchestre
“Is AI embedded in our processes, steered and measured?”
The Orchestre level describes coordinated uses whose adoption, quality, costs and business impact are reviewed together, with an accountable owner and a documented fallback.
- Standardise approved uses, data and validation.
- Review adoption, performance and incidents regularly.
The Architecte
“Is AI a structural competitive advantage?”
The Architecte level describes an organisation with shared foundations that allow supervised AI uses to operate across processes without losing control of access, changes or dependencies.
- Test access controls, logging, recovery and supplier exit.
- Manage changes, total cost and risk across the shared platform.
The Pionnier
“Are we setting the standard for our sector?”
The Pionnier level describes several redesigned processes where cooperation between people and AI produces sustained results and remains subject to supervision, review and improvement.
- Demonstrate durable value across the agreed scope.
- Arrange independent scrutiny proportionate to risk and use it to improve operations.
The 9 dimensions of maturity, level by level
The nine dimensions describe separate capabilities, so the overall level must reflect the highest stage whose essential requirements are demonstrated across the agreed scope; tool sophistication or a strong average score cannot compensate for a critical gap.
| Dimension | Spectateur | Artisan | Orchestre | Architecte | Pionnier |
|---|---|---|---|---|---|
| Sovereignty and control | Uses are invisible | One supplier selected locally | Classified data and approved suppliers | Governed architecture, tested portability | Dynamic choices, substitutable components |
| Resilience and continuity | Dependencies are unknown | Informal manual fallback | Documented backup and degraded mode | Tested recovery and exit | Exercised continuity across scenarios |
| Augmented staff | No official use | Individual practices | Priority roles equipped and supported | Assistance embedded in eligible roles | Human-AI cooperation redesigned at scale |
| AI literacy and culture | Limitations are poorly understood | Self-directed learning | Role-based pathways and understood rules | Verification and supervision are mastered | Continuous learning linked to incidents |
| AI governance | No owner | Isolated sponsor | Owners, policy and periodic review | Operable register and decisions | Proportionate independent review and continuous improvement |
| Technical debt | Unobserved | Dispersed scripts and connections | Backlog and minimum standards | Controlled versions, tests and documentation | Debt managed against maintainability objectives |
| Security | Unknown surface | Uneven local controls | Governed identities, secrets and logs | Tested threats, incidents and suppliers | Controls reassessed as risks evolve |
| Workflows and agents | Outside processes | Occasional assistants | Production uses with approval | Supervised agents in defined perimeters | Governed, reversible multi-agent processes |
| Value and FinOps | No starting point | Reported gains, dispersed costs | Baseline, budget and impact reviewed | Cost per outcome and attributable ROI | Continuous allocation by value, risk and learning |
Full detail of the 5 levels and the 12-point self-diagnostic grid: see the dedicated article or chapter 2 of the white paper.
Five obstacles to reliable AI deployment
An AI project can stall for several different reasons, and a useful diagnosis separates the business objective, delivery decisions, team adoption, economics and data readiness before deciding what to change.
- An unclear objective makes it difficult to choose a useful process or judge the result.
- A pilot without acceptance criteria, an owner or an exit decision can remain an experiment indefinitely.
- Insufficient user involvement leaves daily practice, training and responsibilities unresolved.
- An incomplete cost model overstates gains by omitting integration, verification, support or the time needed to realise savings.
- Poorly defined data quality and access controls undermine both performance and trust.
The five-phase approach
The five phases organise a deployment from diagnosis to ongoing governance, while the calendar remains a planning hypothesis that depends on scope, data readiness, integrations and the availability of the people responsible.
1. Diagnosis
Identify the processes, existing uses and constraints that should guide the investment decision.
- Map tools, users, data and dependencies.
- Record a baseline and shortlist useful cases.
2. Scope the use cases
Define a pilot that can answer a business question through explicit acceptance criteria and a realistic cost and risk assessment.
- Assign the business owner, users and responsibilities.
- Specify quality, cost, benefit, risk and stop criteria.
3. Prepare the foundations
Make the data, controls and working practices ready for a bounded pilot, with documented decisions about access and human intervention.
- Prepare the required data and integrations.
- Agree usage rules, training, validation and fallback procedures.
4. Run and evaluate pilots
Observe a limited deployment with real users and compare its measured results with the baseline before deciding whether to expand it.
- Track adoption, quality, incidents and total cost.
- Keep, revise or stop the use case against the agreed criteria.
5. Consolidate and govern
Extend uses that have demonstrated their value while maintaining operational ownership and a review process for changes, incidents and supplier dependencies.
- Standardise validated processes and maintain documentation.
- Review value, risks and the next investment decisions regularly.
INDUSTEC: a protocol for evaluating quotation assistance
The illustrative INDUSTEC quotation example explains how to evaluate an AI assistant on a defined commercial process, using a consistent baseline and separating gross time savings from economic benefits that can actually be attributed to the change.
- Measure the same quotation population before and after the pilot, including review and correction time.
- Track quality, adoption and exceptions alongside processing time.
- Calculate benefits and full costs on the same period, then test the sensitivity of the result.