ПРИЕМКА КОРПОРАТИВНОГО ИИ-ПОИСКА ОБОСНОВАНИЕ РЕШЕНИЯ

Стандарты и качество. 2026. № 10. С. 105–109.

Белов Б. П., Kocharian R., Самойлов И. А., Panfilov P.

Acceptance of enterprise AI search requires checking answer correctness and access controls, defining acceptable errors, and assessing whether testing is sufficient. A record for individual acceptance criteria was developed. Its eight fields link requirements and risks to system configuration, testing, and the acceptance decision. The fields are mapped to provisions in four Russian standards. The analysis draws on 30 published accounts of enterprise search and related workflows. Twenty-four describe at least one control mechanism; four are examined in detail against the record’s fields. Missing acceptance evidence is identified. A hypothetical access-control example shows how to document a decision not to release the system. The procedure distinguishes acceptance, refusal, and deferral. The template is intended for requirements specifications, test programmes, and acceptance reports.

Keywords: управление рисками, искусственный интеллект, risk management, управление качеством, artificial intelligence, quality management, retrieval-augmented generation, acceptance testing, test evidence, приемочные испытания, свидетельства испытания