AI Won’t Fix What Management Hasn’t Defined

By Neil S. Kohrn

A company announces that it is implementing artificial intelligence.

The board hears progress. Management sees potential savings. Employees wonder whether they are training their replacements.

But one question deserves an answer before the announcement:

What, specifically, is AI being asked to improve?

“Implement AI” is not a business objective. Neither is keeping up with competitors or demonstrating that the company is embracing innovation. Without a defined purpose, an AI initiative can consume money, management attention, and employee confidence without establishing whether anything useful has been accomplished.

AI is not a panacea. Its value depends on where it is applied, the information it receives, the limits placed on its use, and the judgment exercised over its output.

Those are management responsibilities.

Before selecting a tool, leadership should identify the current condition, the desired outcome, and the obstacles between them. Is the problem slow customer response? Inconsistent purchasing decisions? Time spent searching internal records? Or a process that nobody has questioned in fifteen years?

Each requires a different approach. Some may benefit from AI. Others may require a clearer policy, better information, or a manager willing to make a decision.

Applying AI to a poorly understood process can make its weaknesses harder to detect. Faster output does not establish better performance. A polished answer does not establish accuracy.

Consider a company seeking to improve customer service. AI might help employees locate information and prepare responses for review. Success could be measured through response time, accuracy, and whether the customer’s problem was resolved.

If management measures only how many inquiries are processed, it may reward speed while overlooking repeated contacts, incorrect answers, and frustrated customers.

The technology can appear successful against a poorly chosen measure while the business outcome deteriorates.

The same lack of definition can damage employee confidence.

When leadership announces AI primarily in terms of labor savings, employees have reason to question what comes next. Asking them to embrace an “additional resource” while leaving its effect on their jobs unexplained invites suspicion.

Those concerns deserve an honest response. Some applications may change staffing requirements. Others may reduce repetitive work and allow employees to devote more attention to customers, exceptions, and decisions requiring experience. Leadership should explain what has been decided, what remains uncertain, and how those decisions will be made.

Enthusiasm cannot substitute for candor.

Employees who perform the work should also help evaluate the application. They know where information is incomplete, which exceptions occur, and where a seemingly simple task requires judgment. Excluding that knowledge can undermine the implementation before it begins.

A disciplined approach starts with a defined problem and a limited application. Establish the existing performance, test the proposed use, review errors and exceptions, and determine whether the benefit justifies the full cost—including training, oversight, correction, and maintenance.

Someone must remain accountable for the result. “The AI recommended it” is an explanation of how a decision was reached. It does not discharge responsibility for making it.

Acceptance must be earned through demonstrated usefulness and credible leadership. Employees are more likely to value a resource when they understand its purpose, can influence its use, and see it helping them accomplish worthwhile work.

The uncomfortable question for a board is whether management understands the business problem well enough to direct the technology.

If leadership cannot explain what should improve, how improvement will be measured, and who remains responsible, it has more work to do before implementation.

AI can be a valuable resource. Defining its job remains a human obligation.