Where does AI change how work gets done?

• Where to startRepetitive work that crosses systems versus the loudest request

• Augment or redesignLower risk versus bigger return
• Build or buy agentsDifferentiation versus speed

• Where people stay in the loopControl versus throughput

• How far autonomy goes, and what stays humanThroughput versus the cost of an unsupervised mistake

• Use-case lists without process owners

• Automating a broken process

• Pilots with no path to production

• Autonomy widened without the error handling to match

The starting level. Nothing precedes it.

• Assistants embedded in a process still need a platform to draw on.AI-Ready Data Estate at Establishing

• Agents act correctly only on shared business definitions.Enterprise Context Layer at Scaling
• Agents in production must be inventoried and controlled.Governance & Trust at Scaling
• End-to-end agents need reliable data products.AI-Ready Data Estate at Scaling

• Agent-first processes need business domains that own them.Organization, Skills & Operating Model at Scaling
• Redesign at scale is funded by proven value.AI Value Realization at Scaling
• Agent-first processes are what an enterprise ontology makes possible.Enterprise Context Layer at Compounding

Only the next level in this pillar.

Only the next level in this pillar.

• P&L-level value shows up once processes change.AI Value Realization at Compounding
• AI fluency becomes a role expectation once agents are part of daily work.Organization, Skills & Operating Model at Compounding

Top of this pillar. Nothing else depends on it.