The learning file reaches for John Kay’s idea of obliquity: the claim that complex goals are often achieved indirectly. Happiness is not usually found by monitoring happiness. Wealth is not reliably built by grabbing at every immediate profit. Great companies often become profitable by serving customers, building craft, and protecting the capabilities that make future profit possible.

This matters because modern work is full of direct measures. Teams count tickets closed, hours saved, leads generated, prompts answered, pages shipped, and costs reduced. Measurement is useful. The trap begins when the measure becomes the target and the target starts damaging the system it was supposed to improve.

The digest cites Boeing and ICI as examples from Kay’s argument: organizations that were strongest when they pursued engineering or research excellence, then weakened when financial targets took primacy over the conditions that had produced performance. The broader lesson travels beyond those cases. In adaptive systems, direct pressure changes behavior, incentives, and culture. The second-order effects can overwhelm the intended gain.

AI adoption is a fresh test of the same principle. A company that pursues “replace workers with agents” directly may optimize for visible headcount reduction while losing tacit knowledge, accountability, and customer trust. A company that pursues better service, faster learning loops, safer automation, and stronger tools may end up with lower costs by a more durable path.

Personal productivity follows the same map. The direct pursuit of perfect optimization often creates anxiety and tool churn. The oblique path may be simpler: sleep enough, write clearly, keep promises, review work, build skill, reduce avoidable commitments, and leave room for recovery. Those practices are less glamorous than a new system, but they compound.

Obliquity is not anti-metric. It is anti-naive-metric. The right question is whether a target strengthens or corrodes the system beneath it. A good KPI directs attention without becoming a substitute for judgment. A bad KPI invites people to satisfy the number while hollowing out the work.

Friday’s workbench note is therefore practical. When a goal is complex, ask what adjacent capability would make the goal more likely. Build that capability, then measure carefully. The path may bend, but the bend is the point.