Analyzes showing patterns for potential steering violations under FHA (24 CFR §100.70). Computes chi-square steering score across inferred demographic proxies and zip clusters. Buyer names processed transiently by AI per inference requirements; never stored or exposed in output. Examples shown only when potential steering detected.

How it works 16 steps

Inspect the data fetches, transforms, gates, and output this recipe runs.

Data fetch

▼ all_deals ?

▼

Process (15 steps)

■ agent_deals filter

Filter by condition

⚠ check_deals guard

Stop if agent deals is empty

■ total_showings count

Count items in agent deals

■ grouped_zips group_by

Group by properties.property_zip

■ zip_clusters_count count

Count items in grouped zips

■ prepare_ai_input map

Extract from each item

■ ai_analysis ?

skip: config.enable_ai equal false

■ is_high_risk compare

Check if ai analysis steering score is greater than setting steering threshold

skip: step.ai_analysis is_null

■ computed_risk_level switch

Map is high risk to one of: true, false

skip: step.ai_analysis is_null

■ final_steering_score coalesce

Use the first available value from: ai analysis steering score,

■ final_risk_level coalesce

Use the first available value from: computed risk level, AI disabled

■ final_analysis_summary coalesce

Use the first available value from: ai analysis analysis summary, AI analysis disabled. Enable AI to assess steering risk.

■ final_example_showings coalesce

Use the first available value from: ai analysis example showings,

■ agent_summary to_summary

Format results as a summary card

■ example_table to_table

Format results as a data table

Settings 3 configurable

Configurable at install. Defaults shown — change them anytime in Recued.

hubspot setting =

enable ai toggle = on

steering threshold number = 2