The question we need to ask is simpler: What business problem are we trying to solve?

The companies getting the most value from AI are moving beyond AI assistants to agentic applications that can reason, take action, and do the work rather than simply recommend the next step. 

Here are four of the wrong questions leaders are asking today and what they should be asking instead.

Don’t ask: How many AI agents should we deploy? 

Instead ask: What business outcome are we trying to improve? 

Much of the AI market has been focused on helping users work faster. Copilots answer questions, and AI assistants summarize calls and help employees generate ideas. 

Those advances matter, but they also encourage leaders to focus too much on the number of AI agents vs. what they can do. Enterprises don’t create value by accumulating agents. They create value by solving business problems, including reducing payroll issues, accelerating workforce scheduling, lowering supplier sourcing costs, increasing sales win rates, or improving cash collection.

That’s where the distinction between agents and agentic applications becomes important. A single agent improves a task, whereas an agentic application achieves business outcomes. 

Don’t ask: Can AI help make better recommendations?

Instead ask: Can AI actually help get the work done?

One of the most common misconceptions in the market is treating recommendations and execution as if they’re the same thing.

Many AI systems are already capable of analyzing information, identifying patterns, and recommending next actions. While helpful, this still leaves much of the execution work undone. 

Consider a procurement scenario. An AI assistant might identify a potential supplier and recommend issuing a sourcing event. An agentic application goes further. It can create the sourcing event, evaluate responses against business criteria, route approvals according to policy, update contracts, and move the process forward while maintaining oversight and governance.

Ultimately, AI that can execute improves outcomes. 

Don’t ask: Which AI model performs best on benchmarks?

Instead ask: Does AI understand how our business actually operates?

Much of the AI conversation focuses on model performance, reasoning capabilities, and benchmark scores. But in enterprise environments, intelligence alone rarely creates value.

AI that drives outcomes needs to understand the context of an enterprise in its entirety, from policies, workflows, and approvals to historical transactions, organizational structures, and the current state of business processes.

While a model may know how procurement works in theory, that doesn’t mean it understands your specific supplier policies, approval hierarchy, contract obligations, or current inventory position.

Enterprise context is the differentiator. Without that context, even sophisticated models are often just making educated guesses.

Don’t ask: How do we get a single AI system that can do everything?

Instead ask: How should these specialized AI agents work together?

A common misconception is that enterprise AI will eventually converge around a single all-knowing digital assistant. Business processes don’t work that way.

Organizations are complex ecosystems of specialized functions, each with unique expertise, responsibilities, and objectives across finance, HR, supply chain, customer service, and sales. The same principle applies to AI. Rather than relying on one agent to do everything, organizations need coordinated teams of specialized agents that work together toward a shared objective.

Consider hiring a new employee. A recruiting agent identifies candidates, an HR agent prepares onboarding, a finance agent validates budget availability, and a manager-facing agent coordinates approvals. Individually, each agent contributes specialized expertise. Together, they move a business process toward a shared outcome.

The real value comes from the agents’ ability to coordinate work across business processes.

Today, we’re entering the era of systems of outcomes, where agentic applications continuously monitor signals, reason, coordinate actions across systems, and help move business processes toward defined objectives.

That’s why leaders need to start asking different questions. Instead of focusing on the number of AI agents they deploy or the benchmark scores of the underlying models, they should ask whether AI improves business outcomes, executes meaningful work, understands how their business operates, and coordinates effectively across functions.

That’s the future of enterprise software: applications that don’t just support work, but help get it done.

The opinions expressed in Fortune.com commentary pieces are solely the views of their authors and do not necessarily reflect the opinions and beliefs of Fortune.