How to Think with AI: A Guide to Chain-of-Thought Prompting

At Hapax, we work with banks and credit unions every day to help them get more out of generative AI. One of the most valuable techniques we share with our customers is something called chain-of-thought prompting. It’s a simple but powerful way to encourage AI to show its reasoning step by step—making outputs more transparent, accurate, and useful.

We’ve seen firsthand how this method helps compliance teams think more critically, operations teams spot issues faster, and analysts explore more complex scenarios without getting stuck. Now, we’re sharing this same approach more broadly—so whether you’re a Hapax customer or just starting to explore what AI can do for your institution, you can begin designing smarter prompts that mirror how your organization works and reasons.

What Is Chain-of-Thought Prompting?

Chain-of-thought prompting is exactly what it sounds like: asking your AI assistant to walk through its reasoning before giving you a final answer.

Instead of saying:

“Summarize our recent customer satisfaction survey results.”

You might try:

“Walk through the key trends from our recent customer satisfaction survey results step by step. First, identify patterns in customer complaints. Then, summarize any major improvements compared to last quarter.”

This kind of structured, process-oriented prompting helps the AI slow down, focus on intermediate steps, and reduce the risk of rushing to a generic or incorrect answer. It also gives you more insight into why the model is making a recommendation—something that’s critical when accuracy, context, and regulatory alignment matter.

Why We Recommend It to Hapax Customers

Our Hapax customers use chain-of-thought prompting to:

• Improve decision quality: By seeing how the AI gets to its answers, teams can catch incorrect assumptions or clarify the scope of a problem before acting.
  
  
• Build better audit trails: When AI shows its thinking, it’s easier to document how a recommendation or summary was formed—useful for both internal reviews and external audits.
  
  
• Promote learning and understanding: Especially in technical or compliance-heavy areas, the model’s explanation often helps users better understand the issue themselves.
  
  
• Iterate more effectively: If an answer isn’t quite right, you know exactly which part of the reasoning to revise or redirect.

We’ve even seen this approach used effectively in training, where new team members can use Hapax to learn not just what to do, but why.

How to Start Using It

Here are a few ways we’ve advised Hapax customers to integrate chain-of-thought prompting into their daily workflows:

• Use language like “Walk me through…” or “Explain this step by step…” to encourage structured reasoning.
  
  
• Break down the request into stages: Ask the AI to identify a problem, evaluate options, and explain trade-offs before recommending a path forward.
  
  
• Ask follow-up questions like “Why do you think that’s the best choice?” or “What assumption are you making here?” to dig deeper.
  
  
• Combine with other techniques: For even stronger results, use chain-of-thought prompts with contextual information, role-playing, or examples.

A Prompting Habit Worth Building

Whether you’re a current Hapax customer or an institution exploring how to adopt generative AI more strategically, chain-of-thought prompting is a habit worth building. It’s one of the most reliable ways to improve accuracy, strengthen trust in AI-generated content, and make the model a more valuable thinking partner for your team.

When your prompts encourage AI to “think” more like your institution does—carefully, logically, and in context—you get results that are not just technically correct, but operationally useful.

Looking to learn more about how to structure your prompts effectively? Our AI Prompt Guide for Banks and Credit Unions includes additional techniques to take your interactions from good to exceptional.