The questions that decide whether an AI pilot belongs in production.

They usually stall because ownership, data boundaries, authority, evaluation evidence or operating controls remain unresolved. A successful pilot proves a capability under limited conditions; it does not prove that the organisation can own and control the workflow in production.

It is ready only when the workflow has an accountable business owner, representative evaluation evidence, bounded system authority, observable controls, a tested fallback and an acceptable business case. A single strong score cannot compensate for a critical gap in security, privacy or operating responsibility.

The legitimate outcomes are proceed, proceed with conditions, redesign, defer or stop. Treating all five as valid prevents a team from converting missing evidence into an automatic launch.

It should contain the workflow boundary, accountable owners, baseline and value case, evaluation record, authority limits, data and vendor dependencies, control evidence, monitoring, rollback conditions and the signed decision. The pack should make assumptions and unresolved evidence visible.

It is a fixed-scope advisory architecture engagement for one priority workflow and one production decision. It maps the workflow, authority, dependencies, material risks, controls and evidence before implementation is presumed.

It delivers a workflow and system map, authority and responsibility model, evidence assessment, control requirements, target operating architecture and an executive decision record. The close records one of the five decision states and any conditions or evidence gaps.

The fixed fee is €22,000 excluding VAT. Delivery is planned for 15-20 business days after the agreed readiness conditions are met, including access to evidence, decision-makers and relevant specialists.

No. It does not include model development, integration, migration, deployment, managed operations or specialist legal and security assessments. It does not certify legal, privacy, security, regulatory, technical or production outcomes.

The architecture review records role, intended purpose, possible risk classification, transparency, human oversight and evidence questions that may affect the workflow. Legal applicability and compliance conclusions remain with qualified legal and compliance specialists.

An accountable business executive should own the decision, supported by the people who hold material operating or risk authority. Depending on the workflow, that may include IT, information security, data protection, legal, procurement and employee representation.

ERP and CRM platforms are mapped only as dependencies, including interfaces, data boundaries, ownership and control implications. The Architecture Mandate does not sell ERP or CRM integration and does not promise to replace or connect a platform.

The strongest fit is an organisation facing a material decision about one AI workflow, with an executive sponsor and access to the relevant evidence and stakeholders. It is not designed for general AI education, tool shopping or an unbounded transformation programme.