CASE 01 · BROKERAGE NETWORK · ONTARIO
Listing content system across 38,000 active records
The problem: A Canadian brokerage network inherited three MLS feeds and two generations of copy templates. Agents spent evenings rewriting descriptions that still sounded nothing alike — and compliance could not audit fair-housing language at scale.
What we built: A listing content generation pipeline with brand voice rules, neighbourhood lexicons and automated fair-housing checks. Virtual staging passes and image enhancement flagged low-quality media before publish. Editors reviewed batches through a human expert loop queue — approving, rejecting or rewriting with full audit trail.
Outcome: Median draft-to-publish time dropped from 47 minutes to 11 minutes for standard residential listings. Compliance review coverage reached 100% of machine-assisted copy. The brokerage retained full editorial control; we did not guarantee leads, sale prices or returns.
CASE 02 · NATIONAL DEVELOPER · CANADA
AVM support for pre-construction pricing decisions
The problem: A national developer priced pre-construction inventory from comparables that were three to six weeks stale. Absorption forecasts lived in a spreadsheet the sales desk did not trust.
What we built: Automated valuation model support with geospatial analysis, market analytics layers and explicit confidence bands. Analysts challenged every estimate in a review workflow before numbers reached pricing committees. Model QA guardrails flagged when comps drifted outside training distribution.
Outcome: Pricing refresh cadence moved from monthly to weekly. Forecast error on pilot towers improved within the band the team set — still estimates, still reviewed by humans, not certified appraisals and not investment advice.
CASE 03 · PROPERTY MANAGEMENT GROUP · GREATER TORONTO
Lease abstraction for a 12,000-unit portfolio
The problem: A property management group handling mixed-use assets across the Greater Toronto Area stored leases as unstructured PDFs. Rent escalations, option dates and recovery clauses were invisible until someone opened the wrong file folder.
What we built: Document automation flows for lease abstraction with exception queues for ambiguous clauses. Extracted fields synced to occupancy analytics dashboards and CRM automation triggers for renewal outreach. PIPEDA-aligned data governance documented who accessed tenant personal data and why.
Outcome: Structured coverage rose from 34% to 91% of active leases within two quarters. Analysts spent less time hunting PDFs and more time on exceptions the model correctly escalated. No guaranteed rental yields or ROI — operational efficiency gains only.
CASE 04 · REGIONAL BROKERAGE · WESTERN CANADA
Lead scoring without discriminatory targeting
The problem: A regional brokerage wanted demand forecasting and lead scoring to prioritize agent follow-up, but prior vendor models could not explain rankings — and legal was nervous about fair-housing risk.
What we built: A lead scoring model on inquiry behaviour, showing history and price-band signals — with property search and recommendation transparency reports. Fair-housing compliance tests ran before production delivery. CRM automation routed leads with documented reason codes.
Outcome: Agent follow-up within 24 hours improved on scored tiers. Compliance signed off on explainability documentation. The brokerage understood this was decision-support, not a guarantee of conversions or sale prices.
CASE 05 · REIT OPERATIONS · NORTH AMERICA
Portfolio analytics and model governance retainer
The problem: A REIT operations team ran occupancy analytics across assets in multiple provinces but lacked consistent model QA after an internal data science hire left. Valuation and forecasting tools drifted quietly for two quarters.
What we built: A governance retainer covering model QA, guardrails, portfolio analytics standardization and quarterly fair-housing reviews. Senior data scientists retrained demand forecasting models on refreshed property data and documented retraining triggers for enterprise audit.
Outcome: Drift incidents surfaced in monitoring instead of board meetings. The REIT kept ownership of all models and data; we remained the proptech studio on retainer — not an investment adviser and not a fund manager.
