Perspectives

On AI, operations, and the work that scales.

Field notes from the Radiant Work practice. What the research says about why AI pilots fail, what the businesses making AI pay off do differently, and how the operations underneath the model decide whether the model ever ships.

Jump to a section

01 The P2P Gap Why most pilots never reach production. 02 Context The data layer that makes agents work. 03 Operations Redesign the work, then add the AI. 04 Governance Shadow AI, readiness, and the operating model.
01 / The P2P Gap

The Pilot-to-Production Gap

Why most AI prototypes never reach sustained operational use, what the studies actually find, and what the small group of companies converting pilots into production do differently.

02 / Context

Context as the Foundation

Foundation models are trained on the public internet. They do not know your clients, your pricing, your scope, or your playbook. The hardest part of an AI program is giving the model the situational awareness a competent new hire would have after their first month.

03 / Operations

Operations First, AI Second

McKinsey's 2025 research finds that fundamental workflow redesign is the strongest predictor of AI's impact on EBIT, not model choice. The work begins by mapping how work actually moves, not by selecting a tool.

04 / Governance

Governance and AI Readiness

63% of AI users at work operate without policy. Shadow AI is the default state of most organizations. Building an AI program means building the operating model around the model: ownership, metrics, feedback loops, and what happens when the AI is wrong.

Working through these questions
in your own business?
Let's talk.

Take the first step toward a business that runs with clarity and momentum.

Schedule a Conversation →