McKinsey's 3 Horizons of AI Transformation - Which One Is Your Company In?

What it looks like: Employees have ChatGPT or Copilot accounts. They use them to draft emails, summarize documents, and answer questions. There is no central AI strategy. No redesigned workflows. No governance.

Reported value: 13% of leaders at this stage report meaningful enterprise value.

Common trap: Mistaking usage for adoption. If 70% of employees are using ChatGPT but no workflow has been redesigned, you are in horizon 1 regardless of how many seats you have.

What it looks like: AI is deployed in cross-functional workflows. AP invoice processing, lead qualification, customer support triage, and reporting are automated end-to-end. There is a governance structure and a measurement framework.

Reported value: 24% of leaders report meaningful enterprise value.

What makes it different from horizon 1: The workflow is owned. There is someone accountable for the agent's performance, the error rate, and the ROI.

What it looks like: Roles, workflows, and the operating model have been rebuilt around AI. Planners are exception strategists. Sales reps handle only qualified opportunities. AI does the mechanical work; humans do the judgment work.

Reported value: 48% of leaders report meaningful enterprise value.

What makes it different from horizon 2: The org chart changes. Job descriptions change. AI is not a tool the company uses. It is the architecture the company runs on.

McKinsey found that companies at horizon 1 that redesigned workflows were 5.3x more likely to report value than those that did not. At horizon 2, companies with AI-fluent leadership teams were 3.9x more likely to capture value.

Translation: redesign + leadership fluency is the actual unlock. Tool deployment alone is a multiplier on zero.

Ask three questions:

• Can you name the workflows AI runs in production, not pilot? If the answer is "we're testing a few things," you are in horizon 1.
• Is there a senior leader whose full-time job is AI governance and ROI? If the answer is no, you are in horizon 1.
• Have you changed a job description in the last 6 months to add "AI workflow management" or "exception handling"? If the answer is no, you are in horizon 1 or early horizon 2.

Pick one workflow. Any workflow. The criteria: high volume, rule-based, low risk of catastrophic error. Map it. Redesign it. Deploy an agent. Measure before and after time, error rate, and cost. If the numbers move, you are now in horizon 2 for that workflow.

Then pick the next one. That is how reinvention actually starts - not with a strategy deck, but with a redesigned workflow that produced measurable value.