Case studies

The situation

The company ran ComputerEase for job cost, purchasing, inventory, AP, AR and the GL, and a separate CRM for sales, scheduling and field work. The two never talked. Reporting meant emailed report books and Excel exports, and automation meant software robots clicking through CE screens, because the ComputerEase API doesn't cover purchase orders, AP or the GL.

What we built

• A nightly extract on the CE server: 65 tables through a dedicated login that only reads, with payroll and every sensitive column left behind.
• A certificate-signed upload into a walled-off landing zone in Microsoft Fabric, then a checked load into two lakehouses: one for operations data and one for owner-level financials.
• Around 90 readable views that decode CE's cost types, voucher statuses and custom fields, and quietly fix its quirks.
• A Power BI job costing report that reads the lakehouse directly, shared to readers through an app with no database access.
• A purchasing agent that answers questions like "what's arriving this week?" from operations data only, and an executive agent with its own separate, audited access to the financials.
• Email and Teams alerts if a night's load fails or the data goes stale.

Where it stands

It runs every night in production with no one touching it. The job costing dashboard is live, and more agents are being built on the same foundation.

Customers, estimates, jobs, notes and photos had to come across without losing anything, and without flooding the new CRM with duplicates.

• Every photo and document was downloaded to private cloud storage and linked from the customer's CRM record.
• Each customer got a factual job history, built from the data, and a short readable summary.
• The other 13% went to a hand-review queue with clear rules instead of being merged on a guess: customers who already existed in the CRM, duplicates inside the old system (about 4%), and records with no contact info.
• A full dry run sized every outcome before a single live record was written.