The Devil's Advocate Prompt

A leadership team gathers in a quiet conference room to review a new strategic initiative. They have spent months drafting the plan, refining the slide deck, and aligning departments. Everyone around the table nods, smiles, and agrees. The plan looks elegant, the revenue projections are promising, and the timeline is clean. Yet, beneath the surface agreement, a few senior members feel a quiet, persistent anxiety. They suspect that the operational constraints are unrealistic, that the client onboarding bottleneck has been glossed over, or that the market assumptions are too optimistic. But because of corporate politeness, the desire to maintain momentum, or the subtle pressure of groupthink, no one speaks up. The initiative is launched with unanimous support, only to fail six months later for the exact reasons that went unmentioned.

This is the hidden thinking failure of strategic planning: treating AI as a validation mechanism rather than an agent of stress-testing. When leaders use language models to review their business plans, they typically ask questions like Is this a solid plan? or Can you give me feedback on this strategy? Because models are trained to be cooperative and helpful, they will readily agree with your premises. They will format your ideas into beautiful summaries and praise your strategic vision. This feedback loop creates a false sense of security, reinforcing your existing biases and masking the fatal flaws in your logic. By seeking reassurance instead of critique, you treat consensus as safety and allow your most vulnerable operational assumptions to go unexamined.

The error lies in confusing agreement with rigor. True resilience is not built by compiling a list of reasons why an idea might succeed; it is built by identifying every single point of potential failure before the market exposes them. To achieve this, we must shift from validation prompting to destructive stress-testing. We must intentionally configure the model as a hostile auditor—a devil's advocate whose sole objective is to find the hidden structural fault lines in our plans. The goal is to conduct a digital pre-mortem, forcing yourself to defend your strategy against a relentless, unbiased skeptic before a single dollar is invested.

Consider the difference in practice.

A typical, validation-first prompt looks like this:

Here is our plan to transition our consultancy to a productized retainer model. Review this strategy and tell us if there are any areas we can improve to ensure success.

The model will generate a polite, generic response:

Review: Your plan to transition to a productized retainer model is highly strategic. It aligns with modern consulting trends and offers predictable recurring revenue.

Suggestions: Ensure you define your service tiers clearly, set expectations for client response times, and train your sales team on the new value proposition. Overall, this is a strong approach that should scale well.

This response is comforting, but it is operationally empty. It does not challenge the underlying logic of the transition. It doesn't ask how you will prevent scope creep or how you will manage utilization rates when every client requests their deliverables at the same time.

A Socratic, stress-testing approach, by contrast, establishes a critical constraint boundary from the start:

Do not praise this plan. Do not tell me it is a good idea. I am going to paste our proposal to transition from custom consulting to productized retainers. I want you to act as a highly critical, skeptical venture capitalist who has seen dozens of agencies fail during this exact transition. Your goal is to systematically dismantle our strategy. Focus your critique on three specific areas:

1. The hidden operational bottlenecks that will emerge when client demand spikes.

2. The flaws in our pricing model relative to our customer acquisition costs.

3. The risk of talent attrition if our senior consultants feel commoditized by the productized structure.

Ask me three difficult, challenging questions, one at a time, to stress-test our assumptions. Do not move to the next question until I have defended my position on the current one.

When you run this interview, the model stops acting as a cheerleader and begins to operate as a diagnostic tool. It might ask you how you will handle a client who demands custom work under the guise of a standard retainer task, forcing you to define the exact boundary of your service catalog. It might ask you how you will maintain your profit margin if your junior staff requires constant senior oversight to deliver the productized tasks, forcing you to re-evaluate your delivery workflow. The friction of answering these questions is where the actual strategy is forged.

This Socratic friction does not just improve the plan; it changes the culture of decision-making. By delegating the role of the critic to the model, you remove the social friction of dissent. A junior team member who might feel uncomfortable challenging a partner's proposal can point to the model's critique and say, "The model raised a valid point about our capacity limits—how do we solve that?" The AI becomes a safe sandbox for disagreement, allowing the team to stress-test their ideas without damaging their relationships.

The value of a strategic plan is not determined by its initial elegance, but by its capacity to survive a systematic interrogation.

Behavioral Takeaway

• Assign the skeptic role: When sharing a business proposal with a model, explicitly command it to ignore the merits of the plan and focus exclusively on its flaws.
• Run a sequential pre-mortem: Structure the prompt to ask questions one at a time. This prevents the model from generating a massive list of general critiques that can be easily dismissed, forcing you to focus deeply on one problem at a time.
• Document the defense: Save the interview transcript. The answers you provide to the model's critiques should be integrated directly into your risk management framework and operational SOPs.