AI at the Crossroads: Agentic Ambitions Meet Operational Realities

Enterprise adoption of generative AI (GenAI) has reached a saturation point. Virtually every large organization is now leveraging the technology to enhance productivity, streamline decision-making and drive innovation. Notably, companies across diverse sectors are rapidly adopting AI tools to innovate and improve the services and goods they offer to customers. Amazon, for example, utilizes AI throughout its operations, from eCommerce product recommendations to logistics optimization to cashierless physical stores, where customers pay via the Amazon app in “Just Walk Out” shopping.

While this rapid proliferation reflects a broader confidence in the technology’s potential, it also masks the complexity of its current limitations. Nonetheless, discussions increasingly focus on the next evolution of AI: agentic systems capable of fully autonomous decision-making and action, without human oversight. While Microsoft Copilot, OpenAI and others are already developing and deploying agentic AI tools, the reality on the ground remains far more measured.

Across key industries, most chief operating officers (COOs) indicate that the bulk of tasks they perform using GenAI, including product innovation, customer engagement and cybersecurity management, still require heavy human involvement. More than half of those surveyed worry about the accuracy of AI-generated outputs. Only narrow functions, such as software code generation and fraud detection, show signs of meaningful autonomy. Even those require some level of manual review or human-led system control, making agentic AI a more distant prospect rather than a reality.

Reaping the Financial Rewards

At the same time, the financial rewards are becoming clear for firms that have moved aggressively into using GenAI technology. Among high-automation enterprises, none report concern about recouping their investment. But that confidence comes with a tradeoff: Eight in 10 firms that are highly automated through their use of GenAI tools now cite data security and privacy as top concerns, compared to just 39% of those that are less automated. By contrast, half of lower-automation firms still question whether GenAI will deliver meaningful value.

These are just some of the findings detailed in “AI at the Crossroads: Agentic Ambitions Meet Operational Realities,” a PYMNTS Intelligence exclusive report. This edition examines the opportunities and challenges that large companies face when utilizing AI to support and optimize their business operations. It draws on insights from a survey of 60 COOs working at U.S. firms that provide goods, technology or services and generated at least $1 billion in revenue last year. The survey was conducted from April 1, 2025, to April 11, 2025.

• GenAI Is Still a Supervised Tool
• Automation Is Narrow and Context-Dependent
• Automation Boosts ROI—But Raises Security Flags
• Read More
• Methodology

GenAI Is Still a Supervised Tool

GenAI still requires human operators for prompting and assessing the outcomes of most tasks.

Despite a near-universal embrace of GenAI among enterprises, widespread use of agentic AI—a next-generation technology that holds the promise of allowing autonomous software systems to operate completely independent of humans—is far from a reality. Data shows that human intervention remains a core component of most AI applications across the goods, technology and services industries. While GenAI can support ideation and offer data-driven suggestions, it falls short of producing breakthrough innovations independently that COOs feel comfortable putting into motion.

For instance, critical functions including generating feedback on product processes, cybersecurity management and product innovation still depend heavily on human guidance. This is especially true for technology firms, where most, if not all, COOs say these functions require a human operator. For services and goods firms, these tasks require human oversight 60% to 100% of the time. Even common uses—like assisting employees and customers with accessing information or generating summaries of emails and reports—require human supervision in 50% to 100% of cases, depending on the industry.

These findings underscore a fundamental reality: Most GenAI tools remain tethered to humans. The reason: Most enterprise functions are complex, interdependent and context-rich—conditions that challenge today’s GenAI capabilities. While GenAI tools enhance speed and productivity, they do not replace creativity, judgment or decision-making. In industries where context, ethics or regulatory implications are significant, such as healthcare, finance and logistics, manual oversight is non-negotiable.

GenAI is best understood today as a high-performance co-pilot. It drafts, flags and synthesizes, but it does not autonomously steer. Enterprises planning for the long term must build processes that enable collaboration between the technology and humans, rather than assume full delegation is possible or safe. The extent of GenAI’s reliance on human oversight suggests that while agentic AI has the potential to disrupt GenAI, it may also fail to replace human operators.

Automation Is Narrow and Context-Dependent

Full automation through GenAI is currently limited to narrow and industry-specific use cases: Services firms use the software to automate code, while technology companies automate fraud detection.

Automation through GenAI is progressing, but only within narrow and clearly defined contexts. Data shows that true autonomy is limited to select functions such as software code generation and fraud detection. These tasks are repeatable, logic-driven and low-risk—ideal for AI execution. In fact, the technology and services industries report slightly higher automation rates. For instance, COOs at technology firms say that tools to identify fraudulent behavior, errors or inconsistencies are mostly automated 100% of the time. Services firms report that generating code is automated in 100% of all cases.

The biggest barrier to broader automation across other functions is complexity. Strategic functions require context, nuanced understanding or compliance—all areas where GenAI currently falls short. This explains why automation remains siloed within companies, even in digitally mature organizations.

Adoption patterns also reflect enterprise risk tolerance. While IT and development teams are experimenting more readily, COOs are more cautious. They are avoiding full autonomy for tasks that could impact brand trust, present legal exposure or negatively affect customer experience.

The implication is clear: Automation success depends on matching the tool to the task. Overgeneralization—i.e., a one-size-fits-all approach—can lead to failed deployments or unintended consequences. Enterprises must continue to focus on targeted use cases where GenAI’s capabilities align with operational needs, while also investing in the necessary infrastructure and talent for gradual expansion.

Automation Boosts ROI—But Raises Security Flags

As automation advances, doubts about return on investment fade, even as security tops concerns for eight in 10 high-automation firms.

As enterprises increase their use of GenAI, concerns about pouring dollars into the technology diminish, but new risks emerge. According to PYMNTS Intelligence research, none of the high-automation firms surveyed cited return on investment as a concern. By contrast, half of low-automation firms still worry about whether GenAI is worth the investment. Automation, it seems, validates itself financially only over time.

At the same time, increased adoption brings greater exposure. Among high-automation enterprises, 80% cite data security and privacy as their top concern, more than double the 39% reported by low-automation firms. As GenAI systems increasingly touch more sensitive workflows and datasets, the risk surface expands significantly.

Other top concerns include integration with other software, implementation issues (62%) and the accuracy of AI-generated outputs (57%). While these hurdles may not block adoption, they require serious attention. Enterprises must ensure their GenAI use is auditable, transparent and ethically sound. Interestingly, regulatory compliance is less of a concern—just 12% of firms list it as a top drawback. This likely reflects a lag in regulations governing the rapidly evolving technology, rather than an absence of risk. As rules and standards evolve, compliance pressure will intensify, particularly around explainability and accountability.

The overall message is twofold: GenAI delivers measurable business value, but it introduces new enterprise vulnerabilities. To scale successfully, enterprises must match automation initiatives with investments in cybersecurity technologies, data governance and workforce training. In the later phases of GenAI adoption, the challenge shifts from proving value to managing risk. Enterprise leaders must not mistake maturity for safety—GenAI’s operational power demands equally powerful safeguards.

Read More

PYMNTS Intelligence is the leading provider of information on the consumer trends driving innovation in consumer finance, digital payments and financial inclusion. To stay up to date, subscribe to our newsletters and read our in-depth reports.

• From Spark to Strategy: How Product Leaders Are Using GenAI to Gain a Competitive Edge
• Smart Spending: How AI Is Transforming Financial Decision Making
• Since March, Triple the CFOs Report Very Positive ROI From GenAI
• GenAI’s ROI Divide: How Leading Industries Innovate While Others Lag

Methodology

“AI at the Crossroads: Agentic Ambitions Meet Operational Realities,” a PYMNTS Intelligence exclusive report, examines the opportunities and challenges that chief operating officers face when utilizing GenAI to support and optimize their business operations. It draws on insights from a survey of 60 COOs working at U.S. firms that provide goods, technology or services and generated at least $1 billion in revenue last year. The survey was conducted from April 1, 2025, to April 11, 2025.