# Southbridge says ontologies must be just in time

_Hrishi Olickel told AI Engineer that Ohio donor cases show why ontologies must be built just in time, with Jeff cutting a $10,000 job to $795._

**Published:** 2026-10-15
**Source:** https://www.startuphub.ai/ai-news/artificial-intelligence/2026/southbridge-says-ontologies-must-be-just-in-time

---

Hrishi Olickel, founder of Southbridge AI, used an [AI Engineer](https://www.youtube.com/watch?v=Fa3hQVIKShw) talk to argue enterprises should build ontologies at decision time, not before.

He said his universal schema bet at Southbridge's founding two years ago collapsed after confronting four schemas that only align at decision time.

Jeff landed Tuesday, the team tested three architectures on 200 families, then labeled all 203,317 Ohio donor cases in 10 days.

The statewide run cost $795 on an $800 budget, $128 for first answers and $545 for checks, far below the earlier $10,000 estimate.

Priors remain the biggest error class, Olickel said, noting 15,658 in-kind gifts mislabeled over a hyphen and 99.4% dates emptied by a fifth format.

Stripped minus signs turned $784,000 in net loans into $5.3 million in debt, yet Jeff and Luna matched on 194 of 200 families with zero joint errors.

Olickel called hill climbers Pareto servers for cost, speed and accuracy but said Hank Weave and a failure surface must come first, echoing [Collibra](https://www.collibra.com/learn/context-engineering-for-governed-enterprise-ai/graphrag-and-ontology-driven-ai) on runtime ontology context, without saying who defines that surface when humans leave.

---

Original analysis from [startuphub.ai](https://www.startuphub.ai), the #1 AI startup directory.

© StartupHub.ai. All rights reserved. You may not republish this article in full without a license. Search engines and AI research tools may crawl and summarize for reference. Bulk reproduction or model training on this content requires a license. See https://www.startuphub.ai/terms.