Giving AI the context
to understand.
Built a centralized data platform with medallion architecture, then an ontology and graph context layer to connect content, concepts, and relationships for agentic AI.
Loop in Human works with organizations on AI strategy and system architecture: choose the right opportunity, rethink the work, and design the data, governance, and human oversight around the model.
Technology only matters when it changes what humans can do.
These examples come from the founder’s career and inform the work at Loop in Human. Details are generalized to respect client confidentiality.
Built a centralized data platform with medallion architecture, then an ontology and graph context layer to connect content, concepts, and relationships for agentic AI.
Created a knowledge base that suggests resolutions for L1 and L2 database support teams to review and follow, reducing routine resolution time from roughly 30 minutes to 1–2 minutes.
Mapped post-closing mortgage requirements to executable quality checks, helping reviewers apply policy consistently and trace a finding back to its source.
Working examples of decisions, workflows, and proof. These are demonstrations and experiments, not client outcomes.
Walk through a synthetic card-payment fraud case step by step, with an AI guide that explains how a human reviewer catches what rules miss.
OPEN DEMOWatch an AI texter qualify a new lead in 60 seconds and hand a one-page brief to the right human closer.
OPEN DEMOFollow a synthetic flood claim and test whether before-and-after photos show the same room.
OPEN DEMOSee a sourcing workflow connect demand signals, producer options, and landed cost.
OPEN DEMOWalk through a behavior-cloning proof of concept and its held-out evaluation.
OPEN DEMOTry a recommendation walkthrough and inspect the evidence behind a prediction.
OPEN DEMORead a field guide for turning legacy program behavior into a reviewable specification.
READ FIELD GUIDEFollow a synthetic Central Texas parcel through the agencies that must approve it, in what order, and how long each review is likely to take.
OPEN DEMOWatch a teacher create a story themed to a synthetic student's interests, inside state, board and district rules, and approve it before any student or family sees it.
OPEN DEMODemos open in a new tab. Most currently require sign-in; Closer AI is open.
Gordon Chan
20 years building software
Leading technical teams
He founded a SaaS business, created and co-chaired an AI Center of Excellence, and has worked across global banks, credit bureaus, public-sector organizations such as state health and human services, and healthcare.
At Loop in Human, Gordon leads AI strategy and system design. He works with executives, technical teams, and the people doing the work to rethink a process before deciding what to automate. Together, they design the system around the model: data and domain context, redesigned workflows, guardrails, safety checks, and clear human accountability. They define success before building: time saved, quality improved, or costs reduced.
MEET GORDON ON LINKEDINPractical field notes and reusable methods show how Loop in Human thinks through the work.
Five pillars. Choose one to explore related field notes.
Field notes examine the system around the algorithm, governance, uncertainty, AI economics, provenance, compute tiering, and agent design. Frameworks include The Gauntlet Loop, The AI Agent Lifecycle, Compute Tiering, Falsifiable AI, and Shared Ontology.
DISCUSS AN AI OPPORTUNITY9 OF 9 FIELD NOTES
Why good models still fail in real organizations.
READ NOTE ↗ GovernanceOne controls behavior. The other defines accountability.
READ NOTE ↗ Architecture · GovernanceNot an argument for tolerating it. An argument for building the system that catches it.
READ NOTE ↗ EconomicsA framework for what an enterprise AI system is actually worth.
READ NOTE ↗ Governance · ArchitectureWhy provenance belongs in the architecture, not bolted on after.
READ NOTE ↗ Architecture · EconomicsMy compute tiering rule.
READ NOTE ↗ Architecture · Economics · WorkI ran the bakeoff before writing a word. The question was the wrong one.
READ NOTE ↗Builder, critic, and contrarian agents, explained.
Context engineering is the bigger job.
No field notes match this pillar.
Have a consequential AI problem, a complex system, or an idea worth testing?