Illinois has 851 school districts. About 600 are below adequacy. I built a seven-agent pipeline to model what closing each gap could look like.

In Illinois, the state's Evidence-Based Funding (EBF) formula calculates what each of 851 school districts actually needs versus what it gets. Roughly 600 districts are below adequacy. The aggregate gap is $6 billion.

I built an Airtable base to hold all of it — every district's adequacy percentage, funding gap, enrollment, revenue breakdown, tier status, redistribution projections.

Then I built seven Hyperagent agents on top of it.

Each agent has a role in a pipeline that takes a single district from raw data to a stakeholder-ready investment memo:

1. EBF Analyst — pulls the district's numbers from the Airtable base and places it in statewide context
2. District Intelligence — enriches with web research (leadership, community, demographics, recent board decisions)
3. Funder Match — screens the district against three philanthropic investment archetypes
4. Policy Simulator — models dual scenarios (current funding versus aspirational) with redistribution math
5. Stakeholder Briefer — produces the investment memo with sourced claims and methodology disclosure
6. Pre-Flight Self-Check — runs five editorial checks during drafting (are claims sourced, are assumptions disclosed, are both scenarios shown)
7. Critical Reader — an independent editorial agent that reviews the finished memo and flags what still needs tightening

The Airtable base is the single source of truth. Every number in every memo traces back to it. When an agent says a district is at 72.1% of adequacy with a $47.2M gap, that is a live pull from the base, not a hallucination.

I have run three districts through the full pipeline so far — Bloomington SD 87, Decatur SD 61, and Rockford SD 205. Each one taught the system something. Bloomington was the baseline before the self-check layer existed; Critical Reader caught six issues. Decatur was the first district run with the self-check in place; it caught four issues during drafting. Rockford ran through the mature pipeline with minimal residual findings. The quality layer evolved in production, district by district.

Two interactive artifacts if you want to see the system run:

🔗 Pipeline Dashboard — switch between all three districts and watch the seven-agent pipeline process each one step by step 🔗 https://hyperagent.com/s/k0ffSM14704WN9Su_mED2w — side-by-side metrics, pipeline traces, investment theses, and Critical Reader findings across all three districts

Stack:

• Airtable — data layer (851 districts, EBF formula outputs, tier classifications)
• Hyperagent — agent platform (seven named agents, each with its own system prompt, tools, and role)
• The Tabula Report — the research project this work lives under

The part of the architecture I keep returning to is the quality layer. Pre-Flight Self-Check and Critical Reader are not just reviewing; they are creating a feedback loop where each district run produces a better memo than the last. The Airtable base gives the agents something to be accountable to. When Critical Reader flags a claim for verification, there is an actual cell in the base to check it against.

Open to questions on the data model, the agent architecture, or how the pipeline works.