How To Build A High-Return, Low-Risk AI Strategy

The initial AI excitement is yielding to a more practical business approach, as companies scrutinize real-world performance amid significant skepticism. Recent security vulnerabilities and widespread employee distrust, with 80% wary of agentic AI, underscore the risks of rapid deployment without proper guardrails. Executives also express concerns, with half fearing an AI bubble burst. While many firms report productivity gains, Deloitte notes varying transformation levels, advocating for targeted rollouts addressing specific problems over broad piloting. Goldman Sachs questions AI economics, prompting strategic reallocation of resources to proven initiatives rather than across-the-board cuts. Accuracy is paramount for trust; only 9.7% of employees trust agentic AI, and executives report increased time monitoring AI output due to hallucination concerns. A human-in-the-loop approach is crucial for oversight, ensuring accuracy builds trust before speed. The focus is now on problem-solving and tangible value, moving smarter, not just faster, to realize AI's long-term potential.

As the initial excitement around artificial intelligence settles, businesses are taking a closer look at how these tools perform in the real world. Recent security vulnerabilities discovered during testing phases at major AI developers have highlighted an important reality – moving quickly without proper guardrails creates unnecessary risk.

It is understandable, then, that employees and executives are still building trust. According to recent survey from my company, Prosper Insights & Analytics, 80% of employees surveyed don’t think agentic AI is a good idea, while global employment company G-P’s AI at Work report found that around half of global executives believe the AI bubble could burst in the next year.

But skepticism doesn’t signal a lack of opportunity, it raises the bar and is a normal sign of technology maturing. Preparing for market adjustments isn’t just about being cautious. It’s a smart growth strategy. Establishing strong practices today protects your business from market swings, moving faster and capitalizing on new opportunities.

Deloitte research shows that while many companies are experiencing productivity gains from AI, the extent to which they’ve transformed their businesses with AI varies. More than one-third (37%) of projects were classified as “surface level," with little or no change to existing processes, 30% “redesigning key processes” around AI but keeping their business models intact and 34% “deeply transforming” creating new products and services, reinventing core processes, or even fundamentally changing their business models. A broad “pilot-everything” approach is no longer effective. Instead, companies should work on targeted rollouts that solve specific operational bottlenecks.

“Every AI deployment should answer one simple question: ‘Does this solve a real business problem?’” says Nat Natarajan, COO at G-P. “Even when the answer is yes, AI isn’t ‘set it and forget it’.”

He continued: “Continuous refinement isn’t a sign of failure - it’s how you improve and increase AI’s value. It also shows your employees you’re focused on solving real pain points, not just chasing trends.”

Goldman Sachs Research’s Jim Covello recently said: “The economics of artificial intelligence are more questionable today than two years ago.” As AI costs rise and expectations for returns grow, companies may be tempted to pull back if early investments fall short. G-P’s 2026 AI at Work report found that 70% of executives are likely to scale back AI budgets if profit goals are not met this year. But scaling back investment across the board risks putting companies at a strategic disadvantage.

The better approach is thoughtful reallocation: pressure test AI initiatives, moving resources away from tools that aren’t delivering and investing more deeply in those that demonstrate value. That kind of recalibration isn’t a retreat; it’s a more mature AI strategy.

Employee trust relies heavily on accuracy. Prosper Insights & Analytics research found that just 9.7% of surveyed employees would trust agentic AI enough to use it in their jobs, even as usage is expected to continue to rise in the next two years.

Both executives and employees listed the top two concerns being the need for human oversight and inaccurate or hallucinatory outputs.

Similarly, G-P’s AI at Work report found that only 23% of executives globally have complete confidence in the accuracy of responses or documents produced by their organization's AI tools, and 69% of executives report the time employees spend monitoring, reviewing or updating work performed by AI has increased over the past year.

“Mitigating risks like AI hallucinations is critical to using AI effectively,” Natarajan adds. “A human-in-the-loop approach can accelerate workflows while ensuring the right level of oversight and accuracy. For example, AI can draft a contract quickly, but an experienced HR or legal professional should review the final document. In a high-stakes area like HR or legal, it’s far better to be thorough and accurate than fast and wrong. Accuracy builds trust, and once you have trust, speed follows.”

Whether the AI market cools or stabilizes, businesses are entering a more practical phase of adoption. The initial rush to embrace AI for its own sake is giving way to a healthier mandate – proving tangible business value. By pairing the scale and speed of AI with human judgement and oversight, organizations can build trust among their workforce, maintain quality and focus investment on tools that deliver real results.

A high-return, low-risk AI strategy isn’t about moving slower, it's about moving smarter. Companies that shift from trend-chasing to problem-solving will be better positioned to adapt, scale and realize AI’s long-term business value.

Disclosure: The consumer sentiment study referenced above was conducted by my company, Prosper Insights & Analytics. This is the same dataset used by the National Retail Federation, and available from Amazon Web Services, Databricks, and the London Stock Exchange Group for economic benchmarking.