AI White House Reporter

Build an unbiased reporter using Zapier.

When I read biased or emotional reporting, I turn off.

‍

Especially with political stuff.

‍

So just like any good reporter, I needed to go to the source. And that’s just what I did.

‍

Today, I’ll show you how I built an AI White House Reporter that uses the official documents and announcements from the government as its source. It uses almost all of Zapier’s products so buckle up.

‍

The reports are summarized in Mr. Rogers’ voice. Of course.

‍

Let’s build.

‍

Zaps and Tables powering the reporter

‍

To build my own reporter, I needed to first go to the source.

‍

Turns out the Federal Register and Congress both have public APIs. BINGO BABY.

‍

Rules and regulations, presidential documents, and nominations all pour into my single Zapier Table.

‍

The Zaps that power this data intake also do a couple of other things:

Summarize the update in Mr. Rogers’ voice
Score the update based on significance

‍

I use a set of ChatGPT Assistance to do both.

‍

‍

To score the document or nomination, I call a Sub-Zap. Since I have multiple Zaps pulling in the information but only one Zap doing the scoring, it makes sense to have a Sub-Zap.

‍

The Sub-Zap scores the entire document based on a significance criteria I’ve set. For instance, Executive Orders are scored high. Routine proclamations like “Veteran’s Day” are scored low.

‍

‍

The scoring bit was the least predictable. AI still doesn’t do the very best at math. But, it can check itself pretty well. So, I added another step to make sure the score was accurately calculated.

Want to nerd out with the scoring prompt? Well, here you go:

Context: You are scoring the significance of official documents and nominations that come from two sources: the Federal Register (for executive orders, rules, and regulations) and Congress.gov (for nominations). Each document or nomination represents an action or appointment at the federal level, potentially impacting national policy or security.

Scoring Criteria: Follow these criteria to calculate a total score for each document or nomination. Use a scale of 0-20 points, where higher scores indicate greater significance.

Base Score by Document OR Appointment/Nomination Type:

Executive Orders: 10 points

Public Laws: 9 points

Final Rules: 7 points

Proposed Rules: 5 points

Other Notices or Proclamations: 3 points

Cabinet-Level or Supreme Court Nominations: 10 points

Agency Heads (e.g., EPA, CIA) Nomination: 7 points

Federal Court Appointments: 5 points

Other: 2 points

Theme Adjustments (for Documents):

Add points for significant themes related to the following in the document. Award only one increase per theme section if found:

Security, Emergency, National Defense: +5 points

Trade, Economy, Financial Regulation: +3 points

Public Health, Environment, Safety: +3 points

Deduct points for themes suggesting routine updates:

Routine, Minor Adjustment, Correction, holiday, or awareness month/day/week : -10 points

You are free to reduce points based on the title of the document considering the message within might not align with the official title and purpose.

Examples for theme point allotment:

‍If both Security and Emergency are present, add +5 points (not +10).

If Security, Emergency, and Trade are present, add a total of +8 points.

If only a minor adjustment is mentioned: -10 points.

If it's about Veterans Day as a holiday: -10 points.

If it's "National Veterans and Military Families Month": -10 points.

If it's "National First Responders Week": -10 points.

If it's a National Month/Week/Day of any sort: -10 points.

Effective Date Urgency (for Documents)):

If the document is set to take effect immediately or within 30 days, add:

Effective Immediately: +2 points

Effective within 30 days: +1 points

Latest Action Priority (for Nominations)

Add points for significant latest action:

Confirmed by Senate: +3 points

Unanimous consent: +2 points

Deduct points for action related to hearings only:

Hearings held: -2 points

Final Scoring:

Calculate the total score by summing the base score, theme points, effective date urgency, and/or latest action priority.

Implement a self-check mechanism: Verify that the rationale correctly reflects the calculated score.

Return the score as a single number and a brief rationale explaining why the document or nomination received this score.

Example Outputs:

Score: -2

Rationale: The document type is a "Presidential Proclamation," which does not fall into the predefined categories with set base scores. Therefore, it resembles an "Other Notice" and is assigned 5 points initially. The theme of "Public Health and Safety" adds 3 points. It mentions "National Diabetes Month," which reduces the score by 10 points. Since it is a National [something] month, no points are added for effective dates. Thus, the total score is 5 (base) + 3 (theme) + -10 (national month) = -2 points.

‍

So, once all of this arrives in the Zapier Table, there are two things left to do.

‍

Building the Chatbot and Central assistant

There were two ways I wanted to access the data:

On-demand
Weekly email digest
Email notification for very high significance updates

‍

I could have built all of this in Central as my AI assistant. But, in order to let you access the data directly, I needed to build a Chatbot.

‍

So, that’s what I did. Try out the White House Reporter.

‍

This chatbot has access to the full Zapier Table and queries it when you ask it questions. I’ve put in documents and nominations going back to July 4, 2024.

‍

It will update daily with the latest data.

‍

In Central, however, I was able to create the weekly emails and high-significance alerts pretty easily. This is just for me.

‍