The Maturity Mountain
Your company is the climber. The two sliders decide its altitude: how much you invest in tools vs. how much you invest in redesigning the work itself. Notice which one moves you past the plateau. Drag to orbit.
drag to rotate
Estimated productivity gain
—
Base camps of the mountain
| Camp | What it looks like | Typical gain |
|---|---|---|
| 1. Task assist | Individuals use chatbots for emails, summaries, drafts. Adoption is bottom-up and invisible to process metrics. | 5–15% personal time savings that rarely reach the P&L |
| 2. Embedded tools | AI inside existing software (CRM copilots, code assistants). Same processes, faster keystrokes. | 10–25% on specific tasks; studies show coding tasks ~55% faster |
| 3. Redesigned workflows | Processes rebuilt around AI: claims triaged by models with human review, support resolved end-to-end. Roles change. | 30–50% cycle-time cuts on the workflow |
| 4. New operating model | Products, pricing, and org structure assume AI. Headcount grows in oversight and taste, shrinks in throughput work. | Category-level advantage — the gains competitors can't copy by buying licenses |
Why dashboards lie
- Adoption ≠ impact. "80% of employees use AI weekly" measures logins, not outcomes. MIT-affiliated research in 2025 found the vast majority of enterprise GenAI pilots produced no measurable P&L effect.
- Saved minutes evaporate. Ten minutes saved per email doesn't compound unless the workflow downstream absorbs it — otherwise it becomes slack, not output.
- The wrong-thing metric test: if your AI KPIs would look identical for a company using AI brilliantly and one using it as a toy, they measure the wrong thing. Better: cycle time per process, cost per resolved case, revenue per employee.
- Redesign is scary, so it's skipped. Changing a workflow means changing jobs, approvals, and power. Buying licenses upsets nobody — which is exactly why it's the default and why it plateaus.