The ROI Business Case for AI Literacy: Thin but Provable
ROI for AI-literacy training is thin but provable. The causal chain training → behavior change → productivity is supported at every link but rarely measured end-to-end in a single study. That is your opportunity.
1. The evidence (Guide 2, §4)
| Finding | Design | Strength |
|---|---|---|
| Training complements AI adoption (adoption pays off with training) | Panel econometrics (establishments) | (B) |
| Workplace training steepens the exposure→adoption gradient | 35-country survey + shift-share | (B) |
| AI-integrated training sustains performance/readiness | N=174 IT professionals | (C) |
| AI-supported training upskilled residents on a real task (and could deskill) | Within-subjects, 8 residents/150 CXRs | (D) |
| Direct per-practitioner ROI ($12-25K → $150-300K) | Practitioner economic analysis | (E, verify) |
2. The weakest link
Training → firm ROI is quantified only at (E) level. That is the evidence gap your evaluation designs should fill (Guide 2 §5).
3. How to build a defensible business case
- Quasi-experimental design (Guide 2 §5.1): pre/post with comparison group, real tasks, objective scoring.
- Kirkpatrick L1-L4 table + ROI where cost data allows.
- Document the evidence chain: role → task → level → stored record → report (Art. 4).
Further reading
Sources: Guide 2 (Evaluation), §4/§5.1; corpus of 16,459 papers (as of 2026-10). EOF echo "en10 done"