Enterprise Automation / Systems Design / AI-Assisted Development

Reconstructing Enterprise Financial Automation

Reverse-engineering legacy Workato processes and rebuilding production invoice-generation workflows using Tray, SQL Server, and AI-assisted development.

Project statusThree invoice-generation workstreams delivered to production.

Client: Regional Healthcare Services Organization
My scope: Three major invoice-generation workstreams
Implementation: Tray orchestration & SQL Server / T-SQL
Validation: Operational UAT & production delivery

Overview

More than a platform migration

I worked directly with stakeholders at a regional healthcare services organization to reconstruct and deliver three major invoice-generation workstreams within a broader migration from Workato to Tray.

Some inherited workflows required additional validation, so the existing implementation could not always be treated as a definitive specification of the intended business behavior.

This was not a one-to-one platform migration. I reverse-engineered the recipes to reconstruct business intent, then redesigned my three workstreams with Tray for orchestration and SQL Server T-SQL stored procedures for substantial data processing.

Interactive architecture reconstruction

Three layers. One coordinated invoice lifecycle.

Illustrative reconstruction, not an original client system diagram. Generic stage labels show a simplified lifecycle; no integrations or financial processing run here.

Layer 01

Tray workflow orchestration

Scheduling, API calls, notifications, and process coordination.

Supporting process layer

Layer 02

SQL Server data processing

Stored procedures for substantial transformations and database-side processing.

Responsible for stage 01 / Consolidation

Layer 03

Human operational review

Review and confirmation of invoice information before processing continues.

Supporting process layer

Invoice lifecycle

Select a stage to explore the challenge, decision, and significance.

Stage 01 / SQL Server data processing

Consolidation

Human verification. Deterministic processing.

Operational challenge: Invoice information required substantial transformation before it could be reviewed.
Implementation decision: Move substantial data manipulation into SQL Server stored procedures, with Tray coordinating workflow execution.
Why it mattered: Put database-side processing in an appropriate technical layer instead of implementing every operation inside the orchestration platform.

Stage 1 of 6: Consolidation. Responsibility: SQL Server data processing.

AI assisted analysis and development; it was not a financial decision-maker. This reconstruction describes responsibilities and handoffs without exposing client records or implementation details.

Challenge

Recovering business intent

I analyzed complex Workato recipes and their interconnected subprocesses, using AI to assist with understanding existing logic and accelerating development. I independently reviewed and validated AI-generated interpretations rather than treating them as the source of truth.

Manual workflow inspection, real input data, sample invoices, expected output examples, and stakeholder clarification helped establish what each process needed to do, including where additional validation was needed.

Architecture

Redesigning where the logic lived

I personally designed and implemented the Tray workflows, separating orchestration from substantial data processing. I also designed, wrote, executed, debugged, and tested T-SQL stored procedures on SQL Server using SQL Server Management Studio (SSMS).

AI coding assistants accelerated development; I owned the backend logic, its implementation, and validation against business requirements.

Architectural shift / My three workstreams

Inherited approach

Complex, interconnected Workato recipes and subprocesses.
Extensive data transformations embedded within automation workflows.
Some inherited processes required additional validation.

Redesigned approach

Tray coordinating execution, scheduling, API interactions, and notifications.
SQL Server stored procedures handling substantial database-side transformations.
Expected outputs validated against business requirements and operational UAT.

Implementation

Managing operational complexity

The implementation had to account for concurrency, stale information, database processing-status flags, and dependencies between upstream and downstream workflow stages. These were part of determining whether information was ready for the next stage of processing.

The work combined scheduled execution with human-triggered stages. Coordinating them required attention to both the current data and the state of the wider process.

Human verification

The invoice lifecycle

Automated processing produced consolidated invoice information for operational verification. Notification, review, and human confirmation preceded downstream invoice generation and delivery.

This handoff made human verification explicit within deterministic automation. AI assisted my development work; it did not make financial approval decisions.

Validation

UAT and production

I worked through extensive UAT with operational stakeholders, checking expected outputs and resolving issues with the people who would use the processes.

All three invoice-generation workstreams passed UAT and entered production—my delivery scope within the broader migration.

Reflection

What made the three workstreams successful

Reconstruct business intent rather than copy the previous implementation.
Validate AI assistance against source data, expected outputs, and real business requirements.
Choose the appropriate technical layer for each responsibility: orchestration, database processing, or human review.
Treat operational reliability and human verification as design requirements.