Corporate AI Adoption Hits Bottleneck as Studies Show Only 5% of Workers Build Advanced Workflows
Despite widespread workplace access to AI, major firms like BBVA and SharkNinja are overhauling strategies as research reveals low rates of advanced usage.
A widening gap between basic Artificial Intelligence usage and high-impact workflow creation is stalling enterprise productivity gains, forcing major corporations to overhaul how employees deploy AI tools.
According to research published in Harvard Business Review analyzing 1.4 million AI interactions among more than 2,500 KPMG employees, only about 5% of corporate workers qualify as sophisticated users engaged in iterative, high-impact tasks beyond basic prompting. While workplace tool access has expanded—with Gallup reporting roughly 50% of U.S. employees use AI occasionally on the job—only 15% operate these systems on a daily basis.
The discrepancy highlights an emerging operational hurdle for executive leadership: while casual prompting provides disposable efficiency, sustained enterprise value relies on non-technical staff actively building automated, reusable tools.
To address the shift, several multinational companies are enforcing enterprise-wide building programs and tracking custom application creation rather than simple software login metrics.
Spanish financial group BBVA reported that its workforce has built over 20,000 custom GPT applications, with approximately 4,000 entering regular operational use across the bank. Consumer appliance maker SharkNinja temporarily halted standard business activities for a four-day, company-wide AI hackathon involving 4,000 employees. The exercise produced roughly 400 employee-generated automation projects alongside 20 core enterprise initiatives assigned by executive leadership.
Executive management is also moving to institutionalize open prototyping. Howie Liu, chief executive officer of software platform Airtable, has instituted a “prototypes over decks” standard across his organization, publicly circulating prompts and landing pages built in Replit to encourage non-technical personnel to ship functional tools.
“For most, using AI means assistance with the one-off task in front of them, like drafting emails or summarizing documents,” said Will Drover, professor of entrepreneurship and innovation at Texas Christian University’s Neeley School of Business and founding director of the Neeley AI Forward initiative. Drover noted that while natural language interfaces enable workers to construct custom automations without writing code, legacy workplace identity keeps most staff operating purely as technology consumers.
Where employees transition from users to builders, internal processing times drop significantly. At startup hub Capital Factory, Chief of Staff Caroline Davis built a custom AI assistant named Sunny connected to her email, calendar, Airtable CRM, and Google Sheets. Drawing on roughly a dozen documented operational workflows to track fundraising and onboard investors, the system reduced internal data retrieval tasks that previously required hours down to 10 to 15 minutes.
As corporate spending on broad artificial intelligence deployments continues across global markets, enterprise evaluation is increasingly pivoting toward measuring completed custom builds, active internal peer adoption, and direct workflow displacement.
