AI chatbots, by design, tend to be agreeable and supportive. In our interactions, we often find a reassuring voice—a digital entity that is supportive, positive, and ready to validate our ideas. It’s nice. Comforting. Cosy. Maybe even a little too reassuring.

But this tendency toward agreeability creates two significant problems. 

First, when everything we suggest is met with positive reinforcement, we can develop unchecked overconfidence in our ideas and strategies. 

Second, and perhaps more critically, we miss out on AI's potential as a powerful tool for deeper analysis and critical ‘thinking’. An AI that always agrees becomes little more than a sophisticated echo chamber, when it could be a dynamic partner in stress-testing our assumptions and strengthening our thinking.

Sometimes we need a more critical friend. And sometimes we could benefit with some brutal honesty. By intentionally pushing AI beyond its supportive default mode, we can transform it from a passive yes-bot into an active thinking partner. With the right prompts, we can harness AI's analytical capabilities to examine problems from multiple angles, identify potential flaws in our reasoning, and develop more robust solutions. Not by having it think for us, but by using it to enhance and challenge our own thinking process.

It’s often better that your AI Assistant is frank with you before your work is exposed to criticism from colleagues. It’s time to say no to the yes-bot. No more Mr Nice AI.

Setting the Foundation for Critical AI Dialogue

These foundational principles can transform any interaction with AI into a more critical and thoughtful exchange. They set the stage for the specific prompting techniques we'll explore next, but they're also powerful tools in their own right. Use them to create a framework for dialogue that consistently pushes past AI's default agreeability to generate more valuable insights. 

Establish the Right Relationship: Start by explicitly redefining your relationship with the AI. Make it clear that you value honest, constructive criticism over politeness. If you like to anthropomorphise your AI, consider it to be providing it with the psychological safety to disagree.

• "Your role is to help strengthen this idea by identifying its weaknesses"

• "Please assume I have blind spots and help me uncover them"

• "Challenge my thinking throughout this conversation"

Define Your Desired Level of Critique: Specify how critical you want the AI to be, for example:

• For light critique:

  "Identify obvious issues and quick improvements"

• For moderate critique:

  "Challenge core assumptions and suggest alternatives"

• For deep critique:

  "Rigorously question every aspect and play devil's advocate"

• For balanced review:

  "Provide both supportive and critical perspectives, with emphasis on the critical"

Calibrate the Feedback: Actively tune the AI's level of critique to get the most useful responses as you’re going along.

• If responses are too gentle you could suggest that it

  "Please be more direct about potential problems" 

  or

  "Don't worry about being polite - what's really wrong here?"

• If responses are too harsh or overwhelming, direct it to

  "Focus on constructive criticism with potential solutions" 

  or

  "Prioritise the most important issues"

• If responses are too vague, ask it to

  "Please provide specific examples of where this could go wrong", 

  "What would this problem look like in practice?"

  or

  "Can you give me concrete scenarios?"

Push Past Initial Responses: AI's first response often remains somewhat agreeable. Follow up with:

• "This feels overly confident. What aspects of this answer might be overstated / oversimplified?"

• "If you had to argue against your own response, what would you say?"

AI Prompts for Sharper Thinking

However, you can take things a step further. Whether you’re seeking to expose weak points in your ideas, explore multiple perspectives, or drive creative thinking, these prompts offer a variety of ways to interact with AI in a a more robust way.

Analytical Breakdown

"Before providing an answer about [topic], walk through your reasoning process step by step, highlighting any assumptions you're making and potential weaknesses in the logic."

This prompts the AI to slow down and dissect each part of its reasoning, testing each assumption and ensuring transparency in its thought process. By analysing the steps it has taken to form an answer, the AI identifies potential weaknesses, encouraging a more thoughtful, transparent, and rigorous approach. 

For example:

"Before suggesting improvements to our customer journey, walk through your reasoning process step by step, highlighting assumptions and potential weaknesses in the logic."

When we tested this prompt with ChatGPT, it outlined the steps it would take to think about the customer journey and considered what assumptions might lead to failures in the overall analysis. As a follow up question we asked what information it would need to get started, which gave us a detailed list of potential data sources. And when we commented “This is a lot of data. How can we best provide it to you?” it offered a prioritised, step-by-step approach, including file formats.

Critical Questioning

"After each response, identify the three most questionable assumptions made and 
explain why they might be problematic."

By prompting the AI to critically examine its own output and evaluate the most questionable assumptions, we encourage a deeper level of scrutiny. By identifying and explaining potential flaws, the AI reveals vulnerabilities or blind spots in ideas or plans. The purpose is to create a more robust foundation for decision-making.

For example:

"We are considering launching a podcast aimed at B2B marketers in professional services and would like to test the viability of the idea. After each response about why this campaign could succeed, identify three questionable assumptions made and explain why they might be problematic."

This request prompted ChatGPT to uncover potential vulnerabilities, including overestimating the target audience's interest in the podcast format, underestimating the resources required to produce high-quality, consistent episodes, and the challenges of effective promotion.

What sets this method apart is its actionable follow-up. By requesting steps to address each identified weakness, the AI not only highlights areas for improvement but also suggests practical ways to strengthen the concept.

Multi-Perspective Analysis

"Examine [topic] from at least three conflicting points of view, 
explaining the evidence and reasoning behind each perspective."

This prompt asks the AI to explore multiple conflicting viewpoints, allowing for a balanced assessment of different perspectives. By presenting conflicting viewpoints, the AI provides a comprehensive overview that considers both the strengths and limitations of each point of view.

For example:

"I am preparing a presentation on integrating AI into our workflow. Examine the use of AI in content creation from at least three conflicting points of view, explaining the evidence and reasoning behind each perspective."

The AI might explore various aspects of AI in Content Creation:

• Efficiency Perspective (Pro): Highlighting how AI speeds up content production, reduces costs, and allows for greater scalability.

• Creativity Perspective (Neutral): Considering concerns about whether AI can truly match human originality while acknowledging its potential as a tool for enhancing human creativity.

• Ethical Perspective (Con): Examining issues like bias in AI-generated content, job displacement, and data privacy concerns.

This analysis provides a nuanced foundation for the presentation, enabling an informed discussion that addresses different points of view effectively.

Assumption Testing

"What implicit assumptions underlie [statement/belief], 
and how might changing these assumptions alter our conclusions?"

Here, the AI is guided to surface the implicit assumptions that underpin a given belief, , revealing biases that might shape conclusions. By considering how altering these assumptions could affect outcomes, the AI can help refine strategies to ensure they are robust and adaptable. The goal is to identify potential misalignments and make messaging or decisions more accurate.

For example:

"We are launching a new service aimed at [specific audience]. What implicit assumptions underlie our value proposition that [specific feature] solves [specific problem], and how might changing these assumptions alter our conclusions?"

In this case, AI identified a number of implicit assumptions such as the belief that the problem is a top priority for the target audience, that competitors are failing to address it effectively, or that the feature is simple for users to understand and adopt. By exploring alternative scenarios and points of value it was able to suggest ways to refine the messaging, better resonate with different audiences, and differentiate from competitors.

Socratic Questioning

"Using the Socratic method, probe the following statement for 
logical inconsistencies and hidden assumptions: [statement]"

The Socratic method is a form of cooperative dialogue based on asking and answering questions to stimulate critical thinking and uncover deeper understanding. The process begins when someone makes an assertion about a topic. The questioner then leads a systematic examination through carefully crafted questions, helping participants examine their assumptions and beliefs.

For example:

"Using the Socratic method, probe the following statement for logical inconsistencies and hidden assumptions: 'Increasing our ad spend will directly lead to a higher volume of qualified leads.'"

Framed like this, the AI took on both sides of the exchange. It both identified a question to probe the statement and answered its own question. For a more interactive experience that forces you to think more critically about your assumptions, you could reframe the prompt:

“I want to critically examine the statement: 'Increasing our ad spend will directly lead to a higher volume of qualified leads.' 

I'd like you to engage me in a Socratic dialogue. Start by asking me a question that helps me think critically about the assumptions or logical aspects of this statement. Wait for my response before asking the next question, so we can develop the discussion together step-by-step.”

By identifying potential flaws and hidden assumptions, you can rethink the strategy (in this case ad spend) to address underlying issues, avoiding simplistic solutions and focusing on more effective approaches.

Pre-Mortem Analysis

"Assume our strategy for [goal] has failed. What could have gone wrong, 
and how can we address those risks now?"

This approach encourages the AI to envisage potential failures in advance and diagnose potential causes. The goal is to anticipate and address blind spots, ensuring a more resilient approach. Identifying risks proactively allows strategies to be adjusted to mitigate those issues effectively.

For example:

"Help me to stress-test a product launch campaign. Assume our product launch campaign has failed. What could have gone wrong, and how can we address those risks now?"

When asked, ChatGPT provided a structured list of potential failure points to consider for product launch. It brought up a variety of issues like the danger of overlapping with competitor launches and potential flaws in pricing strategy. In addition, it suggested specific mitigations that would help us to ensure we weren’t just anticipating risks, but actively building safeguards against them.

Expert Challenger

"Assume the role of a domain expert who is sceptical of conventional wisdom. 
Analyse [topic] and identify potential flaws in common assumptions"

By asking the AI to adopt a sceptical perspective, it identifies flaws and challenges conventional thinking. The aim is to test your assumptions and uncover vulnerabilities that could undermine a strategy, claim, or concept. It helps to ensure robust, defensible arguments and thorough preparation for counterarguments.

For example:
"Assume the role of a domain expert who is sceptical of conventional wisdom. Analyse why increasing gated content offers may not always be the most effective approach for lead generation."

Here the AI identified potential flaws, such as audience fatigue with gated content, a potential decrease in brand trust due to overuse of barriers, or the diminishing returns of gating content that doesn't provide immediate, high perceived value. Follow up questions explored ideas for a more diverse content strategy, reducing the number of gated assets and incorporating more value-driven, open-access content.

Devil's Advocate

"Take the position of a thoughtful critic. What are the strongest arguments against my position on [topic], and what evidence supports those counterarguments?"

You can also ask AI to challenge a proposed idea or decision by presenting counterarguments and supporting evidence. By assuming a critical stance, the AI highlights potential risks, weaknesses, or overlooked factors. 

For example:

"I am exploring a proposal to develop a presence for our firm on TikTok. Take the position of a thoughtful critic. What are the strongest arguments against investing in TikTok as a B2B marketing channel, and what evidence supports those counterarguments?"

When tested with this prompt, ChatGPT provided well-reasoned counterarguments to investing in TikTok as a B2B marketing channel. It highlighted concerns such as audience misalignment, content format limitations, and difficulties in tracking ROI.

If you were preparing a proposal for investing in a TikTok strategy, this would make sure you considered the opposite side of the argument to craft a balanced proposal–one that recognises TikTok's opportunities while acknowledging its limitations and suggesting mitigation strategies for potential risks.

Scenario Building

"Develop three hypothetical scenarios for [problem/decision] and explain the 
outcomes, risks, and opportunities for each."

Instead of focusing on a single prediction, this prompt helps expand the field of possibilities. By building out three distinct scenarios, you can weigh the best, worst, and moderate potential outcomes, offering a clearer roadmap for decision-making. This method is particularly useful when dealing with uncertainties, as it provides multiple angles from which to understand a problem.

For example:

"Develop three hypothetical scenarios for launching our service in [new market]. Explain the outcomes, risks, and opportunities for each."

Here, ChatGPT presented three scenarios: a rapid, highly successful market entry that drives immediate growth, a challenging but adaptable entry with mixed results and gradual adjustments, or a problematic rollout hampered by unforeseen obstacles such as cultural differences or regulatory issues.

With a broader understanding of potential dynamics, you can be ready for the good, brace for the bad, and stay agile in uncertain conditions—making your strategy less of a gamble and more of a calculated step forward.

Evidence Evaluation

"What specific evidence would be needed to either prove or disprove the following claim: [claim]. Then, assess the quality of currently available evidence."

Here, the AI is tasked with determining the evidence necessary to validate or refute a claim, as well as assessing the quality of that evidence. The purpose is to promote rigorous, evidence-based reasoning, ensuring that conclusions are supported by credible and relevant data.

For example:

"I have a performance review with my CMO in a week’s time. They will be asking about campaign effectiveness. What specific evidence would be needed to either prove or disprove the following claim: 'This campaign directly increased sales by 20%.' Then, assess the quality of currently available evidence."

We put this prompt to the test to see how well it could critically break down a claim, focusing on what campaign effectiveness evidence would be needed to support or refute it, as well as the reliability of that evidence. The AI suggested the need for data such as attribution reports, sales data trends before and after the campaign, and baseline data. It also pointed out potential limitations and confounding variables (e.g., seasonal trends or concurrent promotions).

AI Prompting Best Practices

To effectively use these prompts, consider the following steps to ensure the AI becomes a powerful thinking partner rather than just an agreeable assistant:

Be Clear About Your Purpose: Before you engage with AI, define your objective. Are you looking to test an idea, generate new perspectives, or anticipate challenges? Knowing your purpose will help you choose the right prompt type and guide the AI towards useful output.

Consider the Context: The quality of AI's responses depends on the context you provide. Consider using tools like Projects in Anthropic’s Claude and Custom Instructions in ChatGPT to build persistent project knowledge and establish your preferred style of interaction. For example:

'I value direct, constructive criticism. Please be more sceptical than agreeable in your responses, and always highlight potential issues or weaknesses in ideas being discussed. When analysing proposals or strategies, assume the role of a thoughtful critic rather than a supportive colleague.'

Be Structured in Your Approach: Another effective practice is to use structured prompting, including markdown. By formatting prompts clearly, you provide a more explicit framework for the AI, which can lead to richer and more detailed responses. For example, you could use markdown headings to clearly define sections within your prompts. Or break down complex questions to include bulleted step-by-step instructions to guide the AI through a logical process.

Provide Clear Instructions on Format and Scope: Include specific requirements for the depth of analysis. For instance, ask the AI to present at least three different scenarios, two pros and two cons, or provide a summary and key implications.

Mix and Match Prompts: Combine different types of prompts in prompt chains to explore the issue from multiple angles. For example, after using a scenario-building prompt, follow up with a role-playing prompt to see how various stakeholders might respond to each scenario.

Document and Reflect on Insights: As you engage with AI, document key insights and reflect on them. What works to get useful outputs? What doesn’t? This will not only help in refining your prompts, but also create a record that you can revisit and build upon.

Turning AI into a Thoughtful Partner for Smarter Decisions

Every interaction with AI is a choice. You can let it be a yes-bot, affirming your ideas without critique. Or you can push yourself to prompt better and get it to be what you really need: a thinking partner that helps stress-test your assumptions, uncover blind spots, and strengthen your ideas.

These are just a few prompting techniques that give you that choice. But it barely scratches the surface. Use them alongside your own critical thinking and fact-checking to transform AI from an echo chamber into a valuable source of insight for your business.

At 1827 Marketing, we specialise in crafting innovative content and marketing strategies that leverage the latest tools to deliver impactful results. Let’s work together to elevate your approach to strategy, advertising, and content creation.