OpenAI’s latest report on AI and workplace skills highlights a clear divide in how artificial intelligence is changing jobs: AI is becoming increasingly capable of handling execution, while humans continue to play the central role in deciding what work should be done and why.

The company analyzed how its own researchers use AI and examined which tasks have increasingly been delegated to AI systems. Coding, technical support, monitoring computer runs, and analyzing results emerged as areas where automation has advanced significantly.

But when it comes to choosing research priorities, allocating resources, and deciding which ideas deserve further investment, human judgment remains central.

What workplace tasks is AI automating?

OpenAI’s research points to growing automation in several areas that traditionally required substantial employee time.

These include:

• Coding: AI can generate, modify and debug software, reducing the amount of routine programming work people need to perform manually.
• Technical support: AI can help troubleshoot problems and provide solutions to common technical issues.
• Monitoring computer runs: AI can oversee computational processes and flag issues without requiring constant human attention.
• Analyzing results: AI can process large amounts of information and help researchers identify patterns or interpret outputs.

These developments suggest that the impact of AI may be better understood at the task level rather than simply by asking whether an entire job will disappear.

A single job can contain dozens of tasks. AI may take over some of them while leaving other responsibilities firmly with the employee.

Why is strategic decision-making still a human domain?

AI can increasingly help people execute a plan, but determining which plan to pursue remains a different challenge.

Questions such as these require context and judgment:

• Which project should receive funding?
• Which research idea is worth pursuing?
• How should limited resources be allocated?
• When should an experiment be stopped?
• Is a promising result actually meaningful?
• Should a system be deployed, paused or abandoned?

These decisions often involve competing priorities rather than a single objectively correct answer.

They can also require understanding organizational goals, risk tolerance, ethical considerations and long-term consequences—factors that cannot necessarily be resolved simply by processing more data.

Human judgment may become more valuable

OpenAI’s findings point toward a potential change in the skills employers value.

If AI handles more of the execution involved in writing code, troubleshooting problems or processing results, workers may spend more time setting direction, evaluating options and making decisions.

That could make skills such as these increasingly important:

• Strategic thinking
• Problem framing
• Critical reasoning
• Prioritization
• Risk assessment
• Communication
• Domain expertise
• Judgment under uncertainty

The shift does not mean technical skills will become irrelevant. Instead, technical knowledge may increasingly be combined with the ability to determine how and where AI should be used.

OpenAI says people still set research priorities

OpenAI emphasized that its researchers continue to make the decisions that determine the direction of its work.

The company says people “still set our research priorities, judge which ideas and results to pursue, and decide whether to scale, pause, or deploy systems.”

That distinction is important.

AI can help generate an idea, analyze an experiment, or write the code required to test a hypothesis. But someone still needs to determine whether the result matters and what should happen next.

In other words, automation of execution does not necessarily mean automation of responsibility.

The workplace may shift from doing to deciding

The broader implication is that AI could change the composition of jobs without eliminating the jobs themselves.

A researcher, for example, might spend less time writing routine code and more time designing experiments. A software engineer could spend less time fixing straightforward bugs and more time making architectural decisions. A manager might receive AI-generated analysis but still need to decide which recommendation to act on.

This creates a different definition of productivity.

The valuable employee may not simply be the person who can complete the most tasks. Increasingly, it could be the person who can identify the right tasks to complete in the first place.

The key takeaway

OpenAI’s research suggests that the dividing line between humans and AI is not simply about what each can do.

AI is making rapid progress on execution-heavy tasks, including coding, technical assistance and data analysis. Humans, meanwhile, remain responsible for many of the decisions that determine priorities, direction and consequences.

That could make human judgment—not just technical ability—one of the most important workplace skills as AI becomes more deeply embedded in everyday work.