Build the Program — Atelier Apprend's Transfer Test

A four-round, intermediate executive-education simulation set inside Atelier Apprend, the custom-programs unit of a Lyon business school (EUR 14M annual billings, 22 designers, 60 associate faculty, 80+ tailor-made programs a year). On 9 June 2026, three days before design lock, Nordpharm Industries — a EUR 4.2B European pharma-manufacturing group and one of your largest accounts — raises the bar mid-procurement: the flagship program for 240 mid-level operations managers across 9 sites in 6 countries must now demonstrably improve on-the-job performance, not just satisfaction, and the board wants the unproven AI personalization engine used as a visible differentiator. The EUR 3.6M, three-year framework renews ONLY if this first cohort proves transfer-to-job. The budget is fixed and tight: EUR 1.08M total, EUR 4,500 per learner, no contingency; launch in 10 weeks. Three constraints bite: a fuzzy brief (the sponsor describes the problem three contradictory ways), a poor track record (4.3/5 satisfaction but only 22% measured transfer across the last 12 programs), an AI tool that is unproven and a works-council/GDPR flashpoint, and a real format trade-off (high-touch action-learning ~EUR 6,200/learner vs scalable digital ~EUR 2,800/learner). Across four rounds you (1) SCOPE the real challenge — converge three conflicting problem statements into one sharply measurable behavior change with a success metric and an explicit out-of-scope list, discovering that a program scoped to a fog cannot prove transfer; (2) SELECT and blend the action-learning format against the EUR 4,500 ceiling so learners work on the real challenge, engineering completion without padding; (3) EMBED the AI personalization engine — adaptive pacing, coaching prompts, and the politically charged skills-diagnostic — with a GDPR/works-council consent and data design and a fallback, absorbing a mid-round event (the works council files a formal data-privacy query and the vendor reveals the diagnostic adds EUR 320/learner); and (4) PROVE the impact to a facilitator playing the CHRO and a works-council rep, with a measurement plan that evidences behavior change, a KPI target scorecard against the 22% baseline, and a one-paragraph renewal case. The math rewards a sharply scoped challenge, a transfer-driving blended format inside budget, governed AI that enhances rather than replaces action-learning, and a measurement plan that evidences behavior — and punishes the five classic errors: designing on a fuzzy brief, optimizing cost over impact, gold-plating over budget, switching on AI without governance, and measuring only reaction. Final KPIs track Transfer-to-Job rate (%), Skill Gain, Completion, and Cost per Learner (EUR).