Unit of analysis: One material AI-shaped decision
Can the institution reconstruct and defend how the decision was made?
For one disputed, consequential or high-risk decision, one AI-shaped judgement or one material decision pathway requiring reconstruction.
Client outcome:
A governed finding on whether the decision can be reconstructed and defended, including material governance gaps and priority remediation.
Unit of analysis: One active, pilot-stage or imminent AI use case
Can this use case be governed in practice?
For one bounded AI-enabled workflow or material use case being designed, piloted, deployed, scaled or scrutinised.
Client outcome:
A governed finding on whether the use case can operate under credible authority, oversight, safeguards, intervention and accountability.
Unit of analysis: Multiple AI use cases, functions or institutional systems
Can the institution govern AI coherently at scale?
For enterprise portfolios, multiple use cases, board and executive oversight, group-level governance and institutional accountability architecture.
Client outcome:
An enterprise-level finding on governance coherence, material risks, critical dependencies and priority remediation.
A clear conclusion on what the available evidence supports, what remains uncertain and which conditions are materially significant.
Identified weaknesses in decision traceability, human judgement, authority, accountability, challenge, intervention, information quality or institutional learning.
Clarity on where responsibility exists without sufficient decision rights, practical authority, ownership of judgement or power to intervene.
Focused actions addressing the most material conditions rather than an undifferentiated list of recommendations.
A concise, client-safe report suitable for governance, assurance and decision-making forums.
Where agreed, an executive readout or facilitated workshop to test findings, clarify implications and establish remediation priorities.
LSI identifies what the institution can responsibly conclude, where the evidence is insufficient, which conditions remain materially significant and what must change for the decision, use case or portfolio to be governed credibly.
These diagnostics do not constitute certification, a legal opinion, a regulatory determination or an assurance opinion. They provide decision-makers with an evidence-based governance finding, bounded to the agreed scope and the evidence made available.
For oversight of material AI use, decision accountability, risk exposure and enterprise governance conditions.
For leaders responsible for deploying, scaling, overseeing or defending AI-enabled decisions, workflows and institutional outcomes.
For functions assessing accountability, regulatory exposure, governance adequacy, legal defensibility and institutional risk.
For assurance, oversight and scrutiny of whether AI-shaped decisions are governed, evidenced and traceable in practice.
For institutions where AI affects customers, employees, access, pricing, eligibility, risk, compliance or other consequential outcomes.
For providers and advisers who need to understand whether the institutional environment around an AI use case can support credible authority, oversight, intervention and accountability.
LSI identifies the institutional question, confirms the appropriate unit of analysis and selects the relevant diagnostic route.
The unit of analysis, boundaries, stakeholders, evidence requirements and intended output are agreed before the assessment begins.
The client provides the relevant governance, decision, technical and operational materials, together with access to appropriate stakeholders.
LSI examines the decision pathway, the institution’s representation of reality, the quality of the available evidence, the exercise of authority and human judgement, and the conditions governing challenge, intervention, accountability and traceability.
Material factual findings are tested with appropriate client stakeholders before conclusions are finalised.
The client receives a governed finding setting out supported conclusions, material conditions, priority risks and recommended remediation.
The engagement remains bounded to the agreed unit of analysis. LSI’s underlying assessment criteria, evidence standards, scoring logic and diagnostic methodology remain proprietary.
It is not enough to explain what the technology was designed to do.
The institution must also be able to explain how AI shaped the decision, where accountable human judgement was exercised, who held practical authority, what safeguards were applied and who remained accountable for the consequences.
Which decision, use case or institutional system requires closer scrutiny?
Language Strategy Institute
Copyright © 2026 Language Strategy Institute - All Rights Reserved.