How managers use AI to make decisions

Across industries, leaders are now testing the interplay between AI and human judgment—and developing the processes that let the two work together.

AI as a strategic partner

Spikes embeds AI into his executive workflow. He likens it to how large firms use management consultants to map scenarios and risks as well as act as a sounding board. He uses AI to help with complex decisions across people dynamics, situational gray areas, and selecting external partners or service teams. ​​It could, for example, offer advice on handling disagreements among colleagues or partners, or offer alternate perspectives that challenge someone’s initial point of view.

“I’m constantly having conversations” with different AI tools, says Spikes. “I’ll give them information and have stand-up conversations with them—almost like a full research team, the way you would use McKinsey or PwC” consultants. He says he’ll come to “a fork in the road of decisions” and use AI “to decide this pathway or that pathway.”

He’ll run scenarios related to ambiguous judgment calls through multiple models to compare perspectives before stepping in himself. He says no sensitive data is shared with LLMs; when he’s working with his team or vendors, he often asks for ideas on handling “challenging milestone situations,” including when the company has set goals or KPIs and misses them. The AI doesn’t replace his decision-making; rather, it gives him more insight with which to make a decision.

He points to a recent case with a contractor he let go. The work ended in the first week of the month, but the contractor insisted on being paid for the full month. Spikes ran the scenario through two different AI models. One gave a firm, black-and-white answer—prorate the work and move on. Another tool framed the issue more gently, emphasizing the person’s past contributions. While Spikes ultimately held to his earlier decision—prorating the payment—he says the AI conversations influenced the tone, leading him to approach the discussion with more empathy.

He thanked the vendor for their earlier work but explained that prorating was necessary to maintain fairness across the team. But had he not consulted AI, he may not have been nudged toward that balance. Asked whether AI changed the underlying decision, Spikes says no, but it influenced his tone. “It made me a little bit kinder than I would have been,” he admits.

Supporting day-to-day decisions 

Elsewhere, companies are weaving AI into operational decisions to give employees clearer visibility and speed up decision making. 

Dave Glick, SVP of enterprise business services at Walmart, says corporate teams use an internal AI tool called the “associate super agent.” It works like a single front door: Employees ask a question, and the system quietly hands it off to small, task-specific tools in the background.

One use case is when employees want to understand what went wrong with a shipment or delivery. A shipment might arrive without a corresponding purchase order or end up at the wrong building; the AI system gathers data from multiple sources to piece together what likely happened. 

“Many of these tasks are sort of detective work,” Glick says, emphasizing that the human remains in control and can override any conclusion the AI suggests. What used to require digging through multiple databases is now compressed into a much faster preliminary review, with the AI assembling the data before the employee makes the call.

Marne Martin, CEO of expense-management software firm Emburse, notes that AI works best when the decision is repeatable and the data feeding it is clean. “If you have more than 3.5% of inaccurate or highly biased data in your model, you will not get to the accuracy that you can just trust AI,” she says.

Similarly, Infosys CTO Rafee Tarafdar says the IT services firm ties AI reliance to risk: The higher the stakes and the shakier their confidence in the model for a given use case, the more a human needs to step in.

Is overreliance on AI risky?

The efficiency gains from using AI are early wins, but researchers caution that exposure to AI can change how people act, prompting them to defer to either AI’s judgment too much or default to more control-oriented responses.

University of Massachusetts Lowell associate professor of management José-Mauricio Galli Geleilate says his research shows that consulting AI “turns your framing of the problem and how you see the problem,” nudging leaders “more towards control,” like punitive or surveillance-oriented solutions. 

His coauthor, Beth Humberd, also an associate professor of management at UMass Lowell, describes the effect as a kind of psychological distancing: When managers turn to a machine instead of a colleague, they don’t have the human cues that they would have in asking another person for their thoughts. It’s those cues that “make you pause and consider the person on the other side,” she says.

Léonard Boussioux, an assistant professor of information systems at the University of Washington’s Foster School of Business, says his research shows people can quickly fall in line with AI because the models are “really good at crafting sound arguments,” and humans tend to trust anything that feels logical and well-articulated. 

To curb these effects, researchers say organizations need to build in friction—by forcing people to slow down, questioning the output, and bringing in human context that AI can’t capture.

Companies say they’re using AI to augment but not replace human judgment. And as adoption grows, many are still figuring out where the handoff will be. For many, the hurdle may be more cultural than technical: forcing employees to question AI’s output while getting comfortable with its integration into daily workflows.

AI is “a level up from where we normally are,” says Spikes. “A CEO now has another counselor that is limitless in its ability to pull in data and information. It’s informing me, and it’s giving me a wider point of view.”