The Operating Model for the AI Era

Automating a broken process doubles the problem

We frequently see companies automate a workflow and discover they've automated its inefficiency. A process with fourteen hand-offs, three sign-offs and constant rework is faster when automated — but it's still a bad process, running at a higher speed and with more confidence in its output.

The AI-era operating model starts from the question: what should this process be, given what machines can now do? It's a redesign question, not a tooling question.

The new division of labor

The most durable shift we're implementing with clients is a rebalancing of work: machines handle the repetitive, high-volume, rule-adjacent tasks; humans handle judgment, exceptions and relationships. The practical consequence is that a surprisingly large share of 'knowledge work' is actually repetitive work wearing a disguise.

The organizations getting this right are re-scoping roles before they deploy technology. They define what a human-plus-machine worker looks like — and what training gets them there — as a first-class deliverable, not an afterthought.

• Work is sorted into three buckets: automate, augment, elevate.
• Roles are redesigned around judgment and exceptions, not volume.
• Managers are measured on outcomes, not activity.

Hierarchy follows the workflow

When work becomes visible in real time — status, quality, velocity — the traditional reporting layers lose their purpose. The organizations we're working with are flattening decision rights around workflows: people who see the information are the people who make the decisions.

This is uncomfortable for middle management, and the honest ones acknowledge it. The most successful transformations we've run treat this as a leadership development problem, not a structural one. Managers become coaches and exception-owners rather than information couriers.

A design principle, not a project

An operating model is never finished. The AI-era version has a built-in review cycle: every quarter, one process in every major function is redesigned for the newest capabilities. Not automated — redesigned. That cadence is the difference between a transformation that happened once and a company that continuously transforms.