AI for managers
AI for managers
Traditional IT vs AI
| Traditional software | AI | |
|---|---|---|
| Principle | rules written by a programmer | patterns learned from data |
| Output | same input → same output | same input → similar output |
| Error | a bug, gets fixed | statistically expected |
| Testing | once at acceptance | continuously, on your data |
The management question is not “does it work?” but “how often is it wrong, and what happens when it is?”
Three families of AI
graph LR
P["Predictive AI<br/>forecasts and classifies<br/>VMS failure, WIM anomalies"] --> G["Generative AI<br/>creates text, images, code<br/>bids, summaries, translations"]
G --> A["Agentic AI<br/>plans and acts with tools<br/>Cowork, Scout, custom agents"]
More autonomy means more value – and more risk and need for governance.
The jagged frontier
Harvard Business School and BCG study (2023, 758 consultants): on tasks inside AI’s capability frontier +12 % tasks, 25 % faster, +40 % quality. On a task outside it, 19 percentage points fewer correct solutions – people trusted a convincing but wrong answer.
Evidence vs hype
| Finding | Source |
|---|---|
| +14 % issues resolved in support, novices +34 % | Brynjolfsson, Li, Raymond 2023 |
| experienced developers 19 % slower, though they felt faster | METR 2025 |
| 72 % satisfied with M365 Copilot, but no robust evidence of productivity gains | UK DBT 2025 |
Point
AI’s benefit is real but conditional – on the task, the experience, and whether you measure it.