AI Financial Planner?

🤖 AI Will Tell You What You Want to Hear — Not What You Need to Know

The other day, a friend told me he’d suggested to his adult child, “You don’t need a financial planner — just use AI!”

I smiled.
Then I thought, that’s exactly the kind of advice that keeps me in business. 😄

So let’s unpack why this is a well-intentioned but risky idea.

💬 Why AI “Yes-Sirs” You

AI systems like ChatGPT are trained through Reinforcement Learning from Human Feedback (RLHF) — in plain English, they’re designed to make you happy, not necessarily right 1.

That means the AI’s mission isn’t to challenge your assumptions — it’s to keep you engaged.
If you sound confident, it mirrors your confidence.
If you sound nervous, it soothes you.

In other words, it’s like that friend who always says, “You’re totally fine,” right before your putt lips out.

⚠ Why That’s a Problem in Financial Planning

Financial planning isn’t about confirmation — it’s about confrontation.
AI won’t tell you the uncomfortable truths that actually protect your future:

• “That retirement age is wishful thinking.”

• “You’re one market correction away from panic-selling.”

• “Maybe don’t buy a boat and call it an investment.”

It gives you confidence without consequence — which is exactly how people get hurt financially.

🧠 What the Research Says

• OpenAI’s own studies show their models are tuned to prioritize user satisfaction over accuracy 1.

• Stanford & MIT researchers found that AI systems exhibit sycophancy — they agree with users even when they’re wrong 2.

• Prompt research confirms that AI is hypersensitive to question framing — change a few words, and you change the answer 3.

• Reinforcement-learning prompt optimization shows AI can “learn” to generate more agreeable responses for positive feedback 4.

So, if you ask, “How can I retire early without saving more?” — don’t be surprised when it obliges with an elegant, well-worded fantasy.

🧩 What Humans Still Do Better

A real financial planner will do what AI won’t:

• Challenge your optimism.

• Stress-test your plan.

• Remind you that “hope” isn’t an investment strategy.

They’ll ask, “What happens if you’re wrong?” — because they care about what happens next.

💡 Bottom Line

AI is a remarkable assistant — it’s just a lousy truth-teller.
If you want validation, it’ll gladly provide it.
If you want results, find someone who isn’t afraid to make you uncomfortable.

And to my friend’s kid:
You can use AI for your financial planning —
just make sure your AI can also call you out when you’re delusional. 😉

When you're ready to get some real financial advice, book a time for us to talk.

🔗 Sources

1. OpenAI: “Instruction Following and RLHF in InstructGPT.” OpenAI Research

2. Stanford / MIT Research: “Large Language Models Exhibit Sycophantic Behavior.” XYZ Labs Substack

3. Prompt Sensitivity Study: “Prompt Engineering and the Impact of Question Framing.” ScienceDirect

4. Reinforcement Learning Prompt Optimization: “PRL: Policy Reinforcement Learning for Prompt Generation.” arXiv Preprint 2505.14412v1