Study: AI Returns Depend on Business Workflow Overhaul
A PYMNTS Intelligence report finds that deeper AI adoption yields higher returns, but only when companies fix underlying organizational issues like data
Enterprises investing heavily in artificial intelligence are finding that further spending on advanced models may be less valuable than fixing the internal operational problems that stop current AI from doing real work. This is the core finding from the PYMNTS Intelligence report "AI at Work: Why Deeper Enterprise Use Produces Stronger Returns," published in September.
The study revealed a significant gap between how much AI companies use and the value they extract from it. Companies with AI embedded in one or two business functions reported using the technology across an average of 41 out of 75 measured tasks. Interestingly, firms with AI embedded in three or more functions used it across 40 tasks, showing virtually no difference in breadth of use.
However, the return on investment told a different story. Only 55% of the lighter adopters reported generating returns from their AI investments. Among the companies with AI embedded in three or more functions, that figure jumped to 93%.
The Shift from Technology to Operations
AI appears to be hitting a familiar technology bottleneck. The obstacle is no longer access to the technology itself but the enterprise's readiness to use it. An AI assistant that drafts a memo can function atop a disorganized company. But an AI agent tasked with approving an invoice, modifying a customer account, starting a procurement workflow, or recommending major financial moves cannot. These are not merely technology requirements; they are operating model requirements.
This shift helps explain another counterintuitive data point from the report. Companies with three or more embedded AI functions reported facing an average of 5.6 barriers to adoption. Companies with no embedded AI reported just three barriers. The firms furthest along with AI are not discovering fewer problems. They are uncovering more fundamental issues within the company itself.
Uncovering Organizational Debt
Moving AI from pilot projects to full production exposes accumulated organizational debt that smaller tests can ignore. This debt includes incompatible databases, unclear ownership of processes, inconsistent policies, disconnected workflows, and approval structures designed for humans manually moving information between systems. Consequently, the next wave of enterprise spending on AI may be classified under different budget lines.
The coming investment boom may technically be counted as spending on cybersecurity, identity management, cloud infrastructure, data management, consulting, integration, or workflow software. Economically, however, much of this spending will essentially be for AI enablement.
The New Competitive Advantage
If access to powerful AI models continues to become a commodity, merely possessing AI will offer little competitive edge. Rivals can buy access to roughly the same intelligence. What they cannot quickly purchase is the organizational architecture needed to exploit it effectively. This turns clean data, interoperable systems, clear decision rights, and mature governance from routine corporate housekeeping into critical productive assets.
The central question for businesses is no longer which company has the smartest AI model, as everyone can rent one. The decisive question is which company can actually put that intelligence to work within its operations.
