Current State Opportunity The Harness Roadmap Business Case The Ask →
Phase 1 · Strategy & Planning · June 2026

Turn scattered AI
experiments into one
governed capability.

REALTOR.ca already runs on AI — just unevenly, invisibly, and without a way to prove or scale what works. This plan recovers the value, controls the risk, and pays for itself in the first year.

$1.5–2.9M
Annual productivity value, O*NET-grounded range across 72 people
~898 hrs/wk
Recoverable capacity at org scale (v2 projection)
1.6–3.2×
Value offsets the cost of a 4–6 FTE central team
60 days
To a measured pilot readout and a scale decision
01 — The problem

Adoption is real. Coordination is missing.

Teams build local automations and find genuine gains. The momentum proves the value — but fragmentation caps it and raises risk at the same time.

Duplicated

Same work, solved twice

No shared prompts, skills, or templates. Every team re-solves problems others already cracked, and quality swings by person.

Invisible

No telemetry, no ROI

Spend is uncontrolled and value is anecdotal. Nothing is baselined, so nothing can be proven, attributed, or scaled.

Unmanaged

Risk without guardrails

No acceptable-use standard, no review gates, no data-exposure controls. Shadow AI grows faster than governance.

02 — Where the value is

Two value streams, sequenced by risk.

Start with low-risk capacity recovery that funds the program. Layer in revenue bets only once the capability is proven.

Near-term · Primary

Productivity & capacity

Recover time and lift quality in existing workflows — documentation-heavy, repeatable knowledge work. Fast, low-risk, self-funding.

  • PRD, roadmap and business-case first drafts
  • User stories, acceptance criteria, process mapping
  • Portfolio reporting and exec-deck assembly
  • Code scaffolding, unit-test and regression generation
Tier 3 · Deliberate

Revenue & strategic bets

AI-enabled products and data services. Validated bets, sequenced after the capability is in place — not day-one work.

  • Enhanced DDF — depth and breadth of MLS data
  • Paid property valuation (AVM) product
  • Raw data & insights feeds for banks and consultancies
  • REALTOR-tailored CRM and richer user profiles
03 — The durable asset

One governed harness. Every entry point.

The lasting value isn't any single tool or model. It's the harness — shared context, skills, policies, routing, telemetry and cost controls that every team reaches through the interface that fits its work.

Front end Harness API policy / model / tool routing approved systems audit + telemetry executive dashboard
Entry channels

The interface that fits

M365 Copilot, Copilot Studio, Power Automate, Claude Code, GitHub Copilot, custom HTML — all governed entry points, not disconnected experiments.

Source of truth

The harness as code

A Git repo versions context, prompts, skills, policies and evals under PR review. It holds no secrets and grants no privileged access.

Control plane

Measured and bounded

Delegated identity, least privilege, sensitivity labels, DLP, and per-task telemetry feeding one executive view of value, cost and risk.

04 — Go deeper

The Phase 1 deliverables.

Each builds on the one before it: assess, map, design the asset, sequence the work, then commit the investment.

05 — The decision

Three approvals to start.

Each is reversible and measured. The risk of waiting is continued fragmentation, duplicated spend, and unmanaged data exposure — with no view of cost or value.

1

The operating model

A central AI Adoption Team starting at 5 FTE, plus a ~15-person champions network.

2

A 60-day measured pilot

Six to eight prioritized use cases, baselined before/after, with an executive readout.

3

Initial prioritization

Tier-1 productivity workflows plus engineering enablement — the fastest, safest wins.

The recommendation

Approve the model and
the pilot now.

The recurring productivity value covers the central-team cost 1.6–3.2× before counting delivery speed, quality, or revenue upside. The capability is self-funding from the floor.