Abstract
We describe the architecture of a deployed AI coaching platform for cohort-based executive education. A large-language-model coach conducts longitudinal, curriculum-grounded coaching conversations with program participants, while the platform converts those conversations into structured, persistent learning records — commitments, insights, confidence self-assessments, points of resistance, application plans — without forms or explicit data entry. We present the platform's design philosophy (dialogue over dashboards; measuring learning outputs rather than content consumption), its memory and context architecture, the curriculum grounding and pacing engine, peer-learning mechanisms including AI-staged consent-gated introductions and conversational peer review with reviewer-side coaching, a learner-owned synthesized "playbook" artifact, and a privacy architecture providing default-private records, consent-gated sharing, tiered audit retention, and pseudonymized LLM processing. We note throughout where conventional alternatives (tool-calling, retrieval variants, other caching schemes) would serve equivalently. The intent is a faithful implementation record of the platform as deployed.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Hirst, Peter, "Architecture of an AI Coaching Platform for Cohort-Based Executive Education", Technical Disclosure Commons, (June 24, 2026)
https://www.tdcommons.org/dpubs_series/10559