Turn AI From Yes-Man to Consigliere

It's fascinating to watch managers interact with ChatGPT these days. They usually fall into two camps:

Camp 1: When asked to open ChatGPT, they nervously fumble around: "Oh, must have logged out..." or "Not working right now..." (Translation: they've never actually used it).

Camp 2: Those who use it so confidently you'd think it was their most trusted advisor, rubber-stamping every business decision.

As we'll explore today, both groups are missing out on what AI can really do – act as your Consigliere. (You’ve probably heard of them – it's the guy on the left).

But before we dive into how to turn AI into your personal Consigliere, let's look at some research that shows why this matters so much...

Let’s dive in!

The Yes-Man Problem of Generative AI

When it comes to using Generative AI for business decisions, recent research paints a somewhat concerning picture.

An MIT Sloan study found that managers who used ChatGPT for problem-solving were twice as likely to propose controlling, surveillance-heavy solutions compared to those who relied on their judgment alone. Fascinatingly, this happened regardless of what ChatGPT actually suggested. Just using the tool seemed to prime managers for a more mechanistic style of thinking.

BCG found that the performance impact of ChatGPT varied dramatically by task type. On some tasks, ChatGPT users outperformed controls by 40%. But on others, they actually performed 23% worse than those who didn't use the tool – especially on tasks like business problem solving.

On the other hand, a recent Harvard Medical School study had an expert panel review 195 medical Q&As from both doctors and ChatGPT. In about 80% of answers, ChatGPT outperformed the physicians in both quality and empathy.

What's going on here?

The key insight is that GenAI isn't actually "thinking" or "reasoning" in the way humans do. It's performing sophisticated pattern matching and token prediction. Like a hyper-intelligent autocomplete, it's suggesting what comes next based on patterns in its training data - not working through problems step by step like a human would. (No, even GPT-o1 isn't doing that).

This is why treating AI as a yes-man leads to poor decisions. Instead, we need to use it like a Consigliere. Here's what a Consigliere does:

By turning ChatGPT from a Yes-Man to a Consigliere, you'll see blind spots before they become problems, build stronger business cases, and make decisions you won't regret. The goal is to take advantage of the areas where (generative) AI really shines and let it augment your thought process - enhancing rather than replacing your own judgment.

But do we actually do that?

For a hands-on walkthrough, check out my upcoming workshop. If you just want a high-level overview, read on. The gist is that we don't want to rely on the AI's knowledge, we want the AI to extract knowledge from us.

Turning AI into a Consigliere

One secret lies in having AI fill the gaps. Let's run through a simple example.

You're doing a SWOT analysis for a new product launch. Most people would start with Strengths - because that's what excites us. We love talking about what makes our product great!

But by following this typical order, you're actually limiting your analysis. Why? Because we're all naturally biased towards confirming what we already believe. When we start with strengths, we subconsciously look for information that supports our positive view.

This is where AI can be incredibly powerful - not as a yes-man, but as a devil's advocate. Instead of asking ChatGPT to "help me with a SWOT analysis," try this approach:

1. Start with Threats: "What are potential threats we might be overlooking for [product]?"

2. Then Weaknesses: "Given these threats, what weaknesses might make us particularly vulnerable?"

3. Only then move to Opportunities: "Based on these challenges, what unique opportunities might we have?"

4. Finally, Strengths: "Which of our strengths would best help us capitalize on these opportunities?"

See what we did there? Instead of using AI to confirm our biases, we're using it to challenge our thinking and fill gaps in our analysis.

Example: SWOT Analysis Deep Dive

Let's say you're launching a new AI-powered email marketing tool. Here's how the conversation with ChatGPT might go:

First, push for threats (notice how specific we get):

"You're a seasoned marketing tech analyst. What are the top 3 threats that a new AI-powered email marketing tool might face in 2025 that most founders overlook? Think beyond obvious competition - consider regulatory, technological, and market dynamics that could blindside us. For each threat, explain why it might be particularly damaging."

Then drill deeper into weaknesses:

"Given these threats - especially [reference most surprising threat from above] - what specific weaknesses should we audit in our product and team? Focus on non-obvious vulnerabilities that could amplify these threats."

By now, you've probably uncovered several blind spots you wouldn't have thought of otherwise. But here's where it gets interesting - let's flip these insights into opportunities:

"Looking at these threats and weaknesses together, what unique opportunities might they actually create? For example, which customer segments might be particularly motivated to solve these problems?"

Finally, strengths:

"Based on the opportunities we've identified, which capabilities should we prioritize building? Specifically, what would make us uniquely positioned to tackle these challenges compared to both established players and other startups?"

The difference in insights you get from this approach versus a standard "Help me do a SWOT analysis" is night and day. Instead of getting generic responses that just confirm what you already know, you're pushing the AI to help you think through angles you might have missed.

5 More Ways to Make Better Decisions with AI

The SWOT exercise demonstrates one key principle: giving AI a framework to explore ideas systematically. But this approach works far beyond SWOT analysis.

Here are 5 other powerful decision frameworks that work exceptionally well with AI:

1. Issue Trees: Break complex problems down by up to 80% faster and identify 2x more solution paths than traditional brainstorming. Often spots critical dependencies you'd otherwise miss.

2. Phineo Model: Map your initiative's full impact chain - from immediate outputs (like "50% faster processing") through outcomes ("30% cost reduction") to long-term impact ("culture of innovation"). Perfect for building business cases that actually get approved.

3. Situation-Complication-Resolution: Structure your thinking like top consultants to get stakeholder buy-in. Studies show structured communication like this increases project approval rates by up to 50%.

4. Pre-Mortem Analysis: Identify potential failure modes before they happen. Teams using this method spot up to 30% more critical risks than traditional risk assessment.

5. Decision Journals: Track and analyze your decisions systematically. According to Nobel Prize winner Daniel Kahneman, this is the best way to test the process by which you make decisions and ultimately improve this process for better quality decisions.

Want to master these frameworks and build your own AI Consigliere? Join my upcoming workshop where we’ll do exactly that.

Conclusion

Better decisions need better information and better ways to challenge that information. While AI won't make decisions for you, it can help you think more systematically and spot blind spots you'd naturally miss.

The key is to move beyond using AI as a yes-man. Don't ask, "Is this a good idea?" - AI will only confirm your biases. Instead, use frameworks to let AI fill in the gaps in your thinking. The magic happens when AI helps you explore the angles you naturally avoid. Just like a good consigliere would do.

Remember: You're still doing the actual reasoning. But used right, AI becomes an invaluable partner for making more profit by making more thorough, well-considered decisions.

See you next Friday!
Tobias