Using AI to Decode the Federal Reserve

By Dave Wang

Jerome Powell delivered a highly watched speech on Friday showing his hand on rate policy.

The Fed's pressers like this are some of the most consequential releases for market participants - but unfortunately it's delivered in an absurdly convoluted manner.

For regular investors, the Fed's way of talking is like deciphering a new language.

For professional investors, you have no time to sit for hours to understand what Powell is saying between the lines.

This is where AI comes in!

I'm going to show you a prompt chain I used to analyze Jerome Powell's Jackson Hole speech in minutes.

Here's the plan for how our prompt chain will work:

  1. Prompt 1 fetches the raw transcript
  2. Prompt 2 decodes hidden messages
  3. Prompt 3 determines asset class impacts
  4. Prompt 4 gives portfolio suggestions

The Prompt

Prompt 1: Fetch Raw Transcript

First prompt in the chain ensures we have data fidelity. We can both scan the transcript ourselves and double check the AI is referencing the correct speech.

Role: Research assistant
Task: Fetch the official Jerome Powell Jackson Hole 2025 speech transcript from federalreserve.gov.
Output: Verbatim text in Markdown, broken into numbered paragraphs [P1], [P2]…
Sources: federalreserve.gov official PDF/HTML only
Rules: No summaries, no interpretations. Just the transcript and a short metadata block (title, date, link).

Prompt 2: Decoding Hidden Messages re: Policy

Second prompt in the chain uses the transcript to uncover any hidden messages that might show the Fed's hand in policy shifts.

Role: Fed-watcher
Task: Analyze the transcript to uncover “between the lines” messages. Reference only the raw transcript above from the Jackson Hole speech from August 22nd 2025.
Output: Table with columns: | Theme | Exact Quote [Px] | Interpretation | Why It Matters |
Rules:
- Separate FACTS (direct quotes) from INTERPRETATIONS.
- Highlight changes in tone vs prior Fed communication.
- Confidence score (0–100) for each interpretation.
Sources: Cite transcript paragraphs only.

Prompt 3: Asset Class Impact

Based on the speech, what will be the impact across major asset classes?

Role: Macro strategist
Task: Translate the hidden messages into likely market impacts over the next 3–6 months.
Output: Bullet list by asset class:
- Treasuries (yields, curve)
- Credit (IG, HY)
- Equities (style tilts)
- USD, Gold, Oil, Crypto
Rules:
- Tie each view back to transcript signals.
- Include Base, Hawkish, Dovish scenarios with probabilities.

Prompt 4: Portfolio Positioning

How should I be positioning as a portfolio manager?

Role: CIO
Task: Summarize into a one-page “What to Do Now” brief.
Output: 1. One-sentence thesis 2. Top 3 portfolio actions (with rationale + invalidation) 3. Key watchpoints (next data/events that change the view)
Rules: Keep under 400 words. Plain English.

The Result:

Full output here: Link

Our AI analysis gives these situation probabilities based on the raw transcript:

  • Base 55 percent: Gradualist easing bias, one small cut or a glide toward less restriction as growth cools, while watching inflation risks. Signals: balanced risks, policy closer to neutral and may warrant adjustment, tariffs as one-time level shift, expectations anchored, growth slowdown.
  • Hawkish 25 percent: No cuts and a willingness to lean tighter if tariff pass-through risks persistence or expectations wobble. Signals: inflation upside risk, vigilance that one-time shocks do not become ongoing inflation, commitment to act forcefully to keep expectations anchored, core PCE at 2.9 percent.
  • Dovish 20 percent: Faster easing if labor deteriorates. Signals: rising downside risks to employment, growth slowing to 1.2 percent, policy is restrictive and could be adjusted, wage-price spiral seen as unlikely.

AI suggests to buy shorter term bonds on the fixed income side. For equities, it suggests to allocate heavier to higher quality large caps particularly financials like banks.

Interestingly, I noticed AI's interpretation of Powell's speech to be more conservative on cuts vs market probabilities (AI thinks we will cut once vs market thinks we will cut twice in 2025).

Source: Polymarket

Why is this the case?

It comes back to nobody actually reading the raw transcripts...

Many of the "Fed analysts" are reporting Powell said he will remove the 2% inflation target (see Bloomberg report below). When in reality if you read in the text he actually mentions this target isn't being removed long run but rather employment seems higher priority at the moment (ie, we will run the inflation hot for the short term and come back to taming inflation with hawkish rate policy).

Our AI analysis picked up this contextual nuance.

Image
Analysts are misinterpreting Powell's speech .... AI picked up the 2% inflation target is NOT removed over the long run. Source: Bloomberg
Our AI noted high confidence that 2% inflation target will remain

As a reminder, excess returns in investing often come from your differentiated views compared to market consensus.

Let's check back in 3 months to see what the Fed actually does!

Hiring:

I'm working with the team at RBF Capital to build out their AI platform.

If you're a talented AI/ML engineer with experience in quant finance, I think this a special opportunity to take ownership in the frontier of building AI tools for investing.

I talk to many hedge funds and few are as forward thinking as the team here at RBF. They have 30+ years of track record and are rewiring everything to supercharge their process with an AI system replicating proven methods.

Hit the reply button if you're interested or know a quant engineer who would be a good fit!

Posts from Me:

If you made it this far, forward this to a friend who might find this helpful

Get one high impact AI prompt every week

Join thousands of smart investors.

2026 — Built by Dave Wang. Not financial advice, only for educational purposes.