A customer receives an AI-assisted claim denial, but the organization cannot identify the model version, documents used, or human review behind it. Transparency makes information about an AI system’s operation available and understandable, so people can assess its use and organizations can trace decisions. Without it, trust and recourse suffer. section 6.4 section 5.5.3
Transparency and explainability are related but distinct. Transparency supports reconstruction of what happened; explainability addresses why a particular outcome occurred. A complete trace may not explain the outcome, while a plausible rationale may lack records that verify it. A model’s written reasoning can also be unreliable evidence of what influenced its answer. section 6.4 section 2.6.4
Put transparency into the decision workflow:
- Disclose when people interact with AI rather than a human, and explain in plain language what happens to their data. section 5.5.3 section 5.1
- For consequential outcomes, explain the main factors, their influence, and the options for challenging the result. section 5.5.3
- Assemble a decision record from the execution pipeline, retrieval layer, and review interface. Capture the model and configuration, inputs and outputs, retrieved sources, tool use, and human review as appropriate to the use case. section 6.4
- Match controls to deployment: check vendor retention, deletion, export, and opt-out capabilities for SaaS; trace the application’s own logs and stored data for API integrations; document provenance and unlearning limits for in-house models. section 5.5.4
Table 6.3 The Decision Record (conceptual schema). Each row is one automated or AI-assisted decision. Fields are grouped by transparency layer. In practical scenarios, operational and regulatory information may live in separate tables or collections.
|
Layer
|
Field Name
|
Description
|
|
Provider
|
model
|
Identifier
or alias for the base model version. Contains {provider, model_name,
model_tag, api_release_date, model_card_url}.
|
|
System
|
system_prompt_version
|
Hash or ID
for the current prompt or system configuration
|
|
System
|
retrieval_sources
|
IDs or
hashes of documents shown to the model. Contains {doc_id, snippet_offset,
source_version}
|
|
System
|
input
|
The raw user
input that triggered the decision. Contains {user_prompt_redacted_id,
user_prompt_hash}
|
|
System
|
output
|
The raw
model output before post-processing. Contains {outcome (one of:
“approve”,“deny”,“request_info”,“escalate”), structured_key_facts:[],
policy_refs:[]}
|
|
System
|
postprocessing_id
|
List of
transformations applied after model response
|
|
Explainability
|
rationale_explanation
|
Human-readable
justification, constructed via template or retrieval
|
|
Explainability
|
evidence
|
List of
factors that influenced the decision, derived from business rules, retrieval
sources, or structured extraction. Contains {factor_id, factor_value, source,
actionable_guidance}
|
|
Explainability
|
confidence_flag
|
Optional
confidence or evidence score
|
|
Reviewer
|
review_id
|
Identifier
for the human review. Contains {reviewer_id, outcome (one of:
“approve”,“deny”,“request_info”,“override”), review_timestamp,
override_explanation}
|
|
Customer
|
access_log:[]
|
Logs access
requests. Contains {user_id (pseudonymized), timestamp, request_id}
|
|
Regulator
|
bundle_id
|
ID tying
this record to quarterly export bundles
|
|
Regulator
|
lawful_basis
|
GDPR basis
for retention and processing
|
|
Regulator
|
record_hash
|
Cryptographic
hash of the full record, logged for audit integrity
|
The decision-record schema groups the information needed to trace an AI-assisted decision across transparency layers. section 6.4
Disclosure and explanation duties depend on context. For example, the EU AI Act requires notice of AI involvement in conversational systems unless it is obvious; US adverse-action notices must be specific and intelligible even when AI produced the decision. section 5.5.3




















