The Most Expensive Part of a Workflow May Be the Handoff Nobody Owns

Adopting AI is no longer the hard part. Nearly 70% of companies across four major economies already use AI, according to the National Bureau of Economic Research. Yet the impact on productivity remains small. That gap raises a practical question: If companies use more AI and do individual tasks faster, why doesn’t more of that productivity translate into business performance?

We are looking in the wrong place.

A business doesn’t operate as a collection of individual tasks. Work moves between people, functions, and systems, and every time it moves there is an opportunity to wait, reconcile, re-enter information, verify something, or figure out who needs to act next. Companies can make each part of a process more efficient and still lose significant time between those parts.

“Instead of only asking where AI can make a task faster, we should also be asking where the business is waiting.”

McKinsey calls this the “coordination tax.” Its research estimates that coordination activities can consume 35 to 60% of work time in knowledge-intensive organizations.

That changes the question. Instead of only asking where AI can make a task faster, we should also be asking where the business is waiting.

Value hides in the handoffs

This is particularly relevant in refined fuels. Buying, selling, and moving fuel requires decisions across commercial and operational functions. Pricing, supply, allocation, credit, ordering, scheduling, terminal activity, transportation, and customer service can each work well on their own. But the business and the customer experience the process across all of them.

A market change becomes a pricing or purchasing decision. An order must align with supply, allocation, and credit. A terminal exception can affect transportation and the customer. At each point, information needs to move with enough speed and context for someone to act.

AI sped up the decision, exactly as promised. But faster decisions haven’t turned into savings because the workflow has not kept up. One energy customer recently gave me an example: a decision took five minutes, but execution took hours because broken processes held it up.

Others name the same disconnect as their own daily pain points. “Physically the fuel is there, but digitally it’s not.” Or, “Our daily reality is spending more time fixing spreadsheets than serving customers.”

These are not complaints about any single task. They are complaints about the space between tasks, where information should move but doesn’t.

McKinsey found a striking example in a demand-to-production workflow at a large industrial manufacturer. Processing across seven steps took 12 to 24 hours, but delays between those steps added 9 to 18 days. Most of the time went into waiting, and that waiting carried a cost.

In refined fuels, unnecessary delay means slower customer response, more manual work, decisions made later than they should be, and experienced people spending time resolving preventable friction instead of applying their judgment where it creates more value.

Making a task faster is not enough

If someone can complete a task in five minutes instead of 20, that is useful. But if the output then waits three hours for somebody to check it, reconcile it, or decide what happens next, the business has optimized a task without improving the workflow. It has simply moved the bottleneck.

The larger opportunity is to reduce the friction around the work. Can information move without somebody entering it again? Can routine discrepancies be reconciled before they require manual intervention? Can an exception reach the right person with the information needed to decide? Can we shorten the time between seeing a change and responding to it?

AI can play a role in those improvements, as can automation, with better-connected data and simpler processes. The technology is less important than what changes as a result. Experienced people should spend more time making decisions, serving customers, and handling exceptions where judgment matters, and less of it searching for information or compensating for disconnected processes.

Modernization without interrupting business

There is another constraint that matters in refined fuels and across the physical economy. The operation has to keep running. You cannot stop moving fuel while you redesign how the business works. Customers still need products, terminals still operate, supply must be managed, and trucks have to load.

“The value is in the space between them, where information moves, decisions wait, and ownership becomes less clear.”

That makes modernization different from changing a purely digital workflow. Improvements must remove friction without creating new operational risks, which is one reason I believe leaders should start with the workflow rather than the technology.

Look at where work stops, where information gets entered more than once, where people switch between systems, and where teams repeatedly reconcile competing versions of the same information. Look at the exceptions that require someone to reconstruct what happened before they can make a decision. Then ask what that friction is costing the business in time, customer experience, productivity, and missed opportunities.

Not every handoff should disappear. Controls, accountability, and human judgment matter. But a handoff should exist because it adds value or manages risk, not because the organization has learned to work around it.

Look at the time between the boxes

Companies have spent years optimizing the individual boxes on workflow diagrams. The value is in the space between them, where information moves, decisions wait, and ownership becomes less clear. Small delays at those interfaces compound across an organization even when every individual function appears to perform well.

AI gives businesses new ways to address some of that friction, but deploying more AI is not the objective. A better test is whether the business can move from information to action with less waiting, less reconciliation, and less unnecessary work.