See if this feels familiar: Your organization is rolling out new AI tools, leadership is encouraging everyone to use them, and your LinkedIn feed is full of tips and tricks for getting more out of AI at work. So you lean in. You start using AI to research, brainstorm, analyze, and draft.
At first, it feels like a productivity breakthrough. But over time, you start noticing a few things. Emails and presentations across your organization are beginning to sound remarkably similar. The time you save generating content is sometimes offset by the time you spend fixing or reworking it. And occasionally, you find yourself putting your name on work you're not entirely confident you could explain or defend.
You're using AI more than ever. But are you actually getting better at your work?
The answer isn't another prompt engineering trick or learning the latest features of your AI tools. It's about fundamentally rethinking how you work with AI: knowing when and how to leverage its capabilities, engaging critically with what it produces, and applying your own expertise and judgment to turn outputs into meaningful results.
Your goal should be to use AI in ways that improve the quality of your work, strengthen your thinking, and build rather than diminish the capabilities that make you valuable. The P.R.O. framework—Purpose, Refine, Own—offers a practical approach to doing just that.
Download the Critical Thinking with AI toolkit
Why Critical Thinking Matters
Technology has long been used to offload cognitive work with great effect. With calculators, we don’t have to calculate long formulas with pen and paper. We no longer need to own encyclopedias now that search engines can retrieve information for us.
But the capacity of generative AI presents risks and opportunities unlike anything we’ve seen before. Whereas other technology can support memory and understanding, AI can offload human skills like reasoning, synthesis, prioritization, evaluation, and judgment.
When cognitive offloading leads to over-reliance, we begin to defer our own judgment to the technology and engage in cognitive surrender. Three risk patterns are worth paying attention to:
- Heavier AI use has been associated with weaker critical thinking. A 2025 study found that greater AI tool use correlated with greater cognitive offloading and lower critical-thinking measures.
- People tend to trust AI even when it is wrong. In a 2026 study, participants who consulted AI followed its recommendations 93% of the time when correct, and 80% of the time when incorrect.
- AI assistance can improve short-term performance while hurting later independent performance. In a 2026 randomized control trial, participants performed better when using AI, but subsequently scored 6–16 percentage points lower on math tasks and 13 points lower on reading comprehension than those who worked without AI. They were also more likely to skip questions rather than attempt them.
To avoid becoming over-reliant on AI, we need a new mental model for using AI that works with our critical thinking skills, not replacing them. The P.R.O. framework is a practical approach for using generative AI while keeping human judgment engaged before, during, and after AI use.
The P.R.O. Framework for Critical Thinking with AI
The P.R.O. framework provides a simple discipline for using AI in ways that keep us actively engaged in the work. P.R.O. has three steps:
[PRO framework graphic]
Purpose: Define What You Want to Do with AI
Effective AI use begins before you even open your AI tool or write a prompt. Start by clarifying what the purpose of your task is, where AI can add the most value, and which parts of the work require your own expertise and judgment. Then consider what you want AI to do, what context it needs, and what a useful output would look like.
Ask yourself the following:
- What am I trying to accomplish?
- What inputs or context does the AI need?
- What output(s) do I want?
- What should the AI not do?
You may be tempted to skip this step and just get started. But defining your purpose actually saves time because it prevents generic outputs.
For example, say you want to prepare for an important meeting with a senior customer. If you asked AI, “Help me prepare for my meeting,” that prompt actually contains a number of different potential tasks. The AI tool will have to decide whether you want to research the customer, synthesize notes from previous conversations, identify issues worth exploring, plan your meeting strategy, or develop thoughtful questions. Each of those tasks may require different data inputs that take more or less time to access and prepare. You may end up receiving a response that’s unhelpful, too broad, or too comprehensive.
Narrowing down your request ahead of time will clarify your thinking and improve your AI output.
Refine: Challenge and Improve AI's Output
Once you’re inside the tool, be deliberate and don’t stop at the first answer. Refining is where we challenge and improve the outputs to suit our needs.
A useful initial prompt generally provides three things: data, context, and instructions. Give AI the source material or data it needs, explain the situation, and be clear about what you want it to do.
[Pie chart graphic of a good prompt]
Treat the first output as a starting point, not the destination. Read it critically and diagnose what it needs. Does the analysis lack depth? Is it too generic? Does it reflect only one perspective? Are important assumptions going untested? Are claims supported by evidence?
Try one of these common refinement moves for iterating with AI:
- Deepen it by asking for more specificity or nuance.
- Clarify if the response is difficult to understand.
- Reframe the output for a different audience or stakeholder.
- Challenge it by asking for an argument against the conclusion.
- Verify the claims by asking for sources and looking at evidence.
- Apply it to next steps or recommendations.
Returning to the customer meeting example, suppose AI recommends opening with an opportunity to expand the relationship. Rather than accepting that suggestion, you might challenge its assumptions: What evidence suggests the customer is ready for that conversation? What if their immediate priority is resolving an existing issue? What alternative approaches might be more effective?
Refining makes the output better, but it also keeps you intellectually engaged. Instead of simply consuming AI-generated content, you are using AI as a thought partner to explore the problem from multiple angles.
Own: Apply Your Judgment to the Final Work
Own is where the work becomes yours, not AI's. Ensure that the final product reflects your judgment and expertise. Ultimately, you should understand the work well enough to explain the reasoning behind it, defend your conclusions, and adapt your approach when circumstances change. Your name is on the work, and you should be prepared to stand behind it.
Start by curating what AI produced. Keep the insights that are useful and relevant, and remove what is generic, repetitive, or inconsistent with what you know about the situation. Next, verify what matters. Check important facts, numbers, names, dates, and sources, and watch out for hallucinations.
Make the final work your own by integrating your expertise and perspective. Ensure you understand every part well enough to explain and defend it. You should be able to take responsibility for the work.
For that same customer meeting, you ultimately decide whether to lead with growth opportunities, address existing challenges, or take a different approach entirely. You make that judgment based on your understanding of the customer and the relationship, and you need to be prepared to explain your reasoning and adapt your strategy as the conversation unfolds—because once the conversation begins, your AI tool can’t help you anymore.
A useful accountability test is to ask yourself these questions:
- Can I explain it?
- Can I defend it?
- Can I adapt it?
- Can I stand behind it?
Using the Right AI Diligence for the Task
Not every use of AI deserves the same level of scrutiny. Using AI to brainstorm names for a team meeting is very different from using it to recommend a pricing strategy or interpret a contract.
So how do you know how much diligence is enough? We can judge how diligent we need to be in using the PRO model based on two questions:
- How much would it matter if this were wrong?
- Am I qualified to judge whether it is right?
[AI diligence matrix graphic]
When the stakes are low and you know the subject well, AI can simply help you move faster. When the stakes are high but your expertise is strong, AI can help you pressure-test your thinking and identify blind spots. When you are less familiar with a subject and the outcome is less consequential , AI can be a valuable tool for learning and exploration. If the stakes are high but your knowledge is low, tread cautiously, seek expert guidance, and avoid using AI output as the basis for a decision.
Better AI Use Requires Better Human Thinking
To realize AI’s full potential, organizations must resist the temptation to offload all thinking onto AI. Instead, the PRO framework encourages us to use our human judgment and critical thinking skills to know where AI can add value and where it can be improved.
At Vantage Partners, we help organizations build these capabilities through training and coaching with practical approaches and frameworks that change behavior. Our Critical Thinking with AI workshop helps professionals put these principles into practice. Rather than focusing on specific AI tools or features, the workshop helps participants rethink how they approach AI-assisted work, apply the P.R.O. framework to real business challenges, and develop the habits to use AI effectively while strengthening their own critical thinking and judgment.
Ready to put P.R.O. into practice? Download our Critical Thinking with AI Toolkit for practical resources to get more value from AI while keeping human judgment at the forefront.
