Expert Perspective: Lessons on Productivity and Learning

Two teams, same tools, opposite trajectories

Last year I reviewed the AI adoption results for a functional team six months into a deployment that had looked like a clear success at the three-month mark. Time saved: up. Tasks completed per hour: up. Adoption metrics: strong. The review presentation showed green across every measure.

By month six, the productivity curve had flattened. The team had optimised within their existing workflows. They had not changed how they worked.

A second team — in the same organisation, with the same tools — had shown weaker early metrics. They were spending time on configuration, building prompt practices, understanding where AI judgment was reliable and where their own oversight was essential. The dashboard had interpreted this as lower adoption. It was higher capability investment.

Productivity is a lagging indicator, not a leading one

Most organisations measure AI tool adoption wrong. They track time saved, tasks completed per hour, and throughput metrics. These numbers improve in the first quarter. They look convincing in review presentations. And they are measuring the wrong dimension of what is actually happening.

By the time a productivity gap appears in the metrics, the capability decision was made twelve months ago.

The Digital Worker dimension of transformation thinking (D5 of the 6xD framework — covering how people deliver change and how they work alongside AI) treats AI-augmented capability as a practice development question, not a tool deployment question. Tool deployment asks: have workers adopted the tool? Practice development asks: have workers built the discipline? Workflow intelligence — the durable capacity to work with AI effectively that continues to develop as tools improve — is the construct that answers the second question.

Workers who redesign workflows pull ahead within a year

The pattern is consistent across sectors, team sizes, and tool categories.

Workers who optimise for immediate productivity gains — using AI to complete familiar tasks faster — show strong early metrics. Six to twelve months later, their productivity curve flattens. The tools have accelerated their existing workflows. They have not changed how they work.

Workers who invest time in workflow redesign show different numbers in the early months. In the first three to six months, their measured productivity is often lower than peers who took the shortcut. Twelve months in, the trajectories separate visibly. Workers who redesigned their workflows have built capability that continues to compound. Workers who optimised have reached a ceiling.

The measurement system is not capturing a behaviour. It is reinforcing one.

Change what you measure before you change anything else

Change what you measure before you change anything else. The leading indicator for AI-augmented capability is not productivity. It is how deliberately and systematically workers are redesigning their workflows — the frequency of workflow reviews, the depth of prompt practice development, the degree to which workers document and share what works.

Start with one team. Define three workflow intelligence indicators. Measure them alongside your current productivity metrics for one quarter. The picture that emerges will tell you far more about where your organisation is going than any throughput dashboard you have run to date.

If your highest-performing AI users are the workers who found the fastest route to a metric gain rather than the workers most actively redesigning their workflows, you have a leading indicator problem. The capability gap will appear in your metrics — but later than you want it, and after the compounding advantage has already built up elsewhere.