Hey everyone, first post here! But I thought that this would be a good spot to share what I have been working on here in Airtable. I am a social entrepreneur within the education social impact space that has been using AI to reimagine how to better allocate education funds to reach benchmarks such as adequacy, equity and joy via our education investments.
This is my third project and this time i tried to make it a little more simple meaning not aiming for a “full stack” product. So, recently I built a redistribution simulator for state education finance using Airtable as the entire backend and interface layer.
The aim is for the platform to enable policymakers describe a redistribution goal in plain English and receive a structured policy analysis — what the move accomplishes, what it costs, who supports it, who opposes it, and whether to proceed.
It is using real data from all 500 Pennsylvania school districts, representing a $4.5 billion statewide adequacy gap identified after a 2023 court ruling that declared the state's funding system unconstitutionally inequitable.
The architecture:
Three tables, two AI agents, formula-driven impact computation.
Table 1 — Districts: 500 rows, one per PA school district. Adequacy gap data, county, demographics.
Table 2 — Scenarios: The user workspace. Each row is a proposed redistribution move. The user writes a Policy Goal in plain English, links affected districts, and sets a mechanism and magnitude.
Table 3 — Impacts: One row per affected district per scenario. Formulas compute post-move revenue, post-move gap, and an impact classification (Recipient Reaches Target, Donor Stable, Donor Weakened, etc.). Rollups on the Scenarios table aggregate the results.
The two AI agents are native Airtable AI fields:
- The Architect reads the Policy Goal and linked districts, then suggests a redistribution mechanism, proposes a dollar magnitude, and writes scaffolding notes explaining its reasoning.
- The Arbiter reads the computed impacts and produces a four-paragraph policy verdict: accomplishment, donor cost, political surface, and recommendation.
Both are set to automatic generation — fill in the fields and the agents respond. No automations needed.
The interface:
Three published dashboards built in Airtable Interfaces:
- Scenario Control Center — summary stats, scenario list with mechanism badges, click-to-expand detail view showing full Architect and Arbiter output side by side
- Impact Financial Flow Intelligence — donor-recipient flow analysis, charts showing distribution by role and impact class
- Adequacy Progress Tracker — the $4.5B statewide gap as a headline number, how much the platform has modeled, and which districts have never been touched by a scenario
12 scenarios are currently loaded across six different mechanisms: targeted gap closure, flat per-student supplement, donor cap reallocation, property tax equalization, formula weight adjustment, and phased multi-year plans.
What I learned building this:
The biggest unlock was switching from Airtable Automations to native AI fields. I originally wired the agents as automations triggered by status changes, which forced a clunky multi-step workflow (change status to "Scaffolded," wait for automation, change status to "Computed," wait again). Converting to AI fields with automatic generation made the platform feel like a product instead of a process — fill in the inputs and the analysis appears.
The formula chain in the Impacts table is where Airtable earns its keep as a simulator. Lookups pull baseline data from Districts, formulas compute post-move state, and a nested IF formula classifies each impact into one of five categories. Rollups on the Scenarios table aggregate the counts. No scripting, no external compute — just linked records, lookups, formulas, and rollups doing what they were designed to do.
Links:
- Live simulator (published Airtable interface): https://airtable.com/apprRQqfZHALhfvY5/shrRrpC5jgagFuXQ7
- Website: https://edfundingsimulator.carrd.co/
- Video walkthrough: https://youtu.be/HdxHsaGXWCk
This is a sibling platform to GrantRail, a portfolio intelligence platform for K-12 philanthropy that I also built on Airtable. One observes philanthropic flows, the other models state redistribution. Both are solo builds.
Happy to answer questions about the architecture, the AI field setup, or the formula chain. Or just general interest, I used to teach so always open to any general interest in the topic.