Prototype And Questions

A narrow first prototype should avoid trying to map every standard or automate every answer. The goal is to prove whether a profile-linked datapoint layer makes suggested responses more accurate, reusable, and reviewable.

First Prototype Shape

Recommended scope:

• choose one disclosure area, likely climate or another part of the current document-processing MVP;
• choose one discloser or a small redacted sample of disclosers;
• use CDP questionnaire references plus one or two external mappings, such as ISSB / IFRS S2 and ESRS E1;
• create 10-20 structured datapoints from approved source evidence;
• let the suggestion service retrieve those datapoints alongside uploaded documents;
• compare suggested responses with and without the datapoint layer;
• evaluate citation accuracy, reviewer usefulness, missing-evidence behavior, stale-data handling, and cross-standard clarity.

Questions To Resolve

• Where do Framework Alignment Tags live technically, and are they available as structured product metadata?
• Which team owns standards mappings as living data: Disclosure, Product, Data, or a cross-functional group?
• Are current mapping assets machine-usable, or are they primarily documents for humans to read?
• Which standards matter most for a first build: ISSB / IFRS S2, ESRS E1, TNFD, GRI, GHG Protocol, or jurisdiction-specific requirements?
• How should the system distinguish "helps answer this standard" from "compliant with this standard"?
• Who validates a datapoint before it can be reused in a formal disclosure response?
• How should parent-child relationships affect reuse across subsidiaries, city-region hierarchies, or group-level disclosure?
• What consent, data-protection, licensing, or customer-agreement boundaries apply before reusing uploaded evidence?
• Should Discloser Profiles become the identity anchor for reusable disclosure datapoints, or should datapoints live in a separate service that only references profile IDs?
• What minimum profile coverage is required before a datapoint can safely be suggested for reuse?
• Which CDP, Briink, or internal suggestion-service components should own extraction, retrieval, review state, and response drafting?

Commercial And Strategic Notes

This fits the broader CDP Digital Brain / AI operating-layer opportunity. The value is not a standalone feature ticket; it is a reusable governed layer for discloser identity, evidence, standards mapping, retrieval, review, and handoff.

Possible Baker Street work package:

• map the current interoperability workflow and owners;
• inventory the existing mapping assets and metadata;
• define the reusable datapoint model;
• prototype retrieval over uploaded documents plus structured datapoints;
• create evaluation criteria and reviewer workflow;
• preserve the decisions, prompts, evidence rules, and open questions in the CDP Index so future teams can extend the pattern.