Posts, research and upcoming book on delivering AI.

AI often makes individual tasks much faster without improving the process they belong to. To capture the value, leaders need to look at where work waits, how exceptions are handled and what happens to the time that AI saves.

AI programs usually add cost and complexity before they reduce them, and returns tend to arrive in stages over two to three years. Leaders who plan for this pattern are less likely to cut strong programs too early or to protect weak ones for too long.

As AI systems move from making recommendations to taking actions, having a person somewhere in the process is no longer an adequate control. Organizations need to define what each system may do, who can stop it and who is accountable for its results.

Access to AI is becoming universal, so it will not differentiate companies for long. The advantage will go to organizations that use signals from daily operations, such as overrides and exceptions, to improve their decisions and workflows.