From Harvard Business Review · · 3m
Map your workflows first to unlock AI's real value
Companies that study how others redesigned around AI identify far more use cases and generate double the revenue, new research shows.
In brief
Companies that study how others redesigned around AI identify far more use cases and generate double the revenue, new research shows. Research shows companies that learn from others' AI workflow redesigns identify 44% more use cases and generate nearly double the revenue. Originally reported by Harvard Business Review.
The AI adoption paradox
Nearly four years after large language models became widespread, most companies have adopted them but few have turned that investment into real productivity or performance gains.
The mapping problem
Researchers from INSEAD and Harvard Business School identified the core challenge: managers know AI can perform useful tasks but struggle to identify where and how it can reshape their businesses.
The field experiment

Five hundred fifteen early-stage companies worldwide received AI tools, training, and mentorship. Only half saw examples of how other organizations had redesigned workflows around AI.
Dramatic results from examples
Companies exposed to real-life examples identified 44% more AI use cases than the control group, completed 12% more tasks, were 18% more likely to acquire paying customers, and generated nearly double the revenue.
Map existing workflows end to end
Examine how changes in one activity affect every step that follows. Look for handoffs, approvals, information translations, and decision stalls. Those transition points often offer greater opportunities than the tasks themselves.
Find the real bottleneck
Ask what is really slowing you down.The answer usually is not the most obvious task but a handoff or decision that holds up everything that comes after.
“Finding the bottleneck that's closest to your customers or closest to the outcome you care about is often the first step.”
Bottlenecks reappear elsewhere
When one activity accelerates, another becomes an unexpected bottleneck. Faster product development increases customer feedback volume. Automated reporting exposes data quality weaknesses. Organizations discover yesterday's constraint disappeared only to reveal another one farther downstream.
Treat AI as ongoing process
Many AI projects initially appear successful before business performance plateaus. Organizations solve one problem but leave the next one untouched. Continuously identify and remove new constraints to capture value.
Redesign the entire process
The highest-performing companies did not stop at identifying bottlenecks. They redesigned the work surrounding them. AI changes the economics of a process only when organizations rethink how activities fit together.
Measure impact, not activity
Managers should ask whether AI-enabled workflows allow them to reach more customers, produce better products or services, or require fewer resources. Otherwise they are measuring activity rather than impact.
“They should ask themselves whether the AI-enabled workflow allows them to reach more customers, produce better products or services, or require fewer resources than before.”
Allow people to experiment
The best-performing companies learned how others reorganized around AI but did not merely copy their ideas. Managers considered what was possible inside their own businesses, leading to experimentation and new processes.
International Motors' approach
Robert Oh, chief digital and information officer at International Motors, a one hundred twenty-four-year-old Illinois-based manufacturer, initially saw dozens of AI experiments improve individual tasks but not the business itself.
From tasks to business value
International Motors mapped cross-functional workflows, identified twenty-one opportunities spanning sales, manufacturing, supply chain, and service, and prioritized five with the biggest impact.
What it means
Research shows companies that learn from others' AI workflow redesigns identify 44% more use cases and generate nearly double the revenue.
Read the TL;DR — this story on one card.
