RESULTS
Case Study

Scaling AP operations with AI copilots

How Colliers built an end-to-end invoice automation workflow that reduces manual effort, improves coding consistency, and maintains use of legacy systems.

Colliers International
Commercial Real Estate
Planning Cycle Time
90% faster
Down from 4 weeks to 10 days
Forecast Accuracy
+22%
Better alignment to actuals
Analyst Hours Saved
725 hrs
Per quarter
01

The Challenge

Colliers' Accounts Payable workflow spans many properties and owners, with coding decisions that can vary by property even for the same vendor. The team needed faster processing without weakening controls, and without forcing property managers to adopt a new tool for approvals.

GL coding is nuanced (vendor + property + description), with some owner-specific GL charts.

Portfolio assignments change frequently, making access/scoping maintenance a recurring pain.

Approvals must remain in the legacy accounting system to avoid workflow disruption, while still increasing automation upstream.

02

Our Approach

The nuance and risk profile of an accounting flow often makes for a dense project that requires care and caution. We treated automation as a full workflow system, not a point solution. We prioritized putting a human in the loop to review the digital co-workers' output and provide feedback:

Map the end-to-end process and identify where automation can safely "insert" without breaking approvals.

Use structured data (historical coding exports, GL catalogs, recurring mappings) to build a context graph and bootstrap accuracy and consistency quickly.

Prioritize explainability, audit logs, and exception handling so automation can scale with governance intact.

03

The Solution

We delivered an end-to-end workbench that combines a unified data layer with AI copilots embedded in daily accounting workflows. Teams can now oversee and approve their digital co-workers' work product, and complete this tedious work in hours instead of weeks. The system provides automated variance explanations, highlights risk factors, and keeps leadership aligned with real-time insights.

Invoices are indexed, ingested, and prepared by an AI Agent.

AI Agents propose GL coding and standardized descriptions, with confidence scoring and reuse of prior vendor/property patterns.

Low-confidence cases route to an exceptions queue for coordinator resolution.

Approved coding is pushed into the legacy accounting system so employees continue approvals in the tool they already use.

04

Results & Impact

Colliers shortened their account payable cycles by 90% and improved forecast accuracy by 22% within the first quarter. The team now finalizes updates in hours, not weeks, and executive reporting happens on a consistent cadence. The AI copilots freed analysts to focus on strategic modeling while leadership gained confidence in the underlying data and the reduction in time spent on this tedious workflow.

Planning Cycle Time: 90% faster — Down from 4 weeks to 10 days

Forecast Accuracy: +22% — Better alignment to actuals

Analyst Hours Saved: 725 hrs per quarter

Company

Colliers International

Industry

Commercial Real Estate

Location

Toronto, ON

Company Size

18,000+ employees

Project Overview

Duration

2 months

Team Size

2 specialists

Role

Product Strategy, AI Enablement, Forward Deployed Engineer, AI Engineer

Technologies

ReactNext.jsTypeScriptLangGraphPostgreSQLSlack

Explore what an AI planning stack could unlock

Let's map your planning workflow and identify the fastest wins.

Book a Strategy Call