My role and contributions
Integration, training and evaluation statuses describe the end of my participation; subsequent business progress has not been verified.
As digital R&D BP support, I advanced data-source integration, agent role design, training scheduling and compliance review. Actual business adoption has not been verified.
Background
The planning team needed AI assistance for product planning and business analysis — combining internal product data, external public data and agent workflows.
Challenge
Internal data carries security and permission constraints; external data needs RPA acquisition; and agents must reach enterprise data sources (北极星) reliably.
Red line
Internal data fed into personal AI accounts requires explicit security and permission review first — a precondition, not an afterthought.
Solution
A Feishu Aily-based agent (connected to GPT-5 and 北极星 data) handles structured execution, while Codex is used for review and strategic reasoning. RPA feasibility is evaluated for external data acquisition.
Data access
Aily is connected to GPT-5 and 北极星 data, wiring internal and external sources.
Training and rollout
Training had been scheduled when my participation ended; subsequent adoption remained to be verified by the business team.
Selection and licensing
Supplier selection and data-license evaluation were still underway when my participation ended.
Technical details
- Knowledge extraction from internal and external data sources.
- Business reasoning via an Aily + Codex split (execution vs. strategy).
- Skill design for reusable planning workflows.
Results
Aily had been connected to GPT-5 and 北极星 data; training was scheduled, and supplier selection and data-license evaluation were still underway when my participation ended. Subsequent completion has not been verified.
Why it matters
The intended workflow lets agents help find data while people retain strategic judgment. Whether it reached stable business use has not been verified.
Lessons learned
Internal data fed into personal AI accounts requires explicit security and permission review before ingestion — it cannot be an afterthought.