Case Study
Build and debug DocuSign CLM workflows visually
At Bitwise Industries my team maintained large-scale DocuSign CLM automations for enterprise clients. When a workflow misbehaved, the only way to inspect it was to export the definition and read the raw JSON. Hundreds of steps, variables, and routing rules were flattened into a format meant for machines.
In day-to-day use by my team, DSDebug turned an afternoon-scale raw-JSON inspection task into minutes of tracing on an interactive graph. This is an observed workflow improvement, not a formal benchmark. The 2022 original was built for exactly that one job: tracing variables through a workflow. The 2023 rebuild kept the tracer and shifted the focus to editing, adding drag-and-drop authoring, a template library, and a console that turned a debugging aid into a full workflow workbench. Both versions are linked above.
The workflows behind this tool are classic document automation: a document arrives, gets classified, has data extracted, and is routed through review and signature. At Bitwise I built those pipelines end-to-end, including HR onboarding that ran through Laserfiche processes wired into DocuSign eSignature. DSDebug exists because pipelines like that fail in production and someone has to trace why. The same operational question applies whenever automated workflows or agents act on business processes: when the automation misbehaves, can you see exactly what it did? My Ticket System takes that question on directly for agent-run work.
Walkthrough
The workbench opens with a library of real workflow patterns: indexing, routing, contract management, internal approvals, and signature flows. There is also an upload button that accepts any exported workflow definition. The palette on the left holds every DocuSign CLM step type, from Salesforce tasks to signature checkpoints, ready to drag onto the canvas.
Here is a 30-step contract-management workflow, one drag later. Phases render as labeled groups, including document generation, departmental review, and exception handling, with decision diamonds, success and failure edges, and finish states laid out automatically. What was a wall of JSON is now something you can pan, zoom, and reason about in seconds.
Loading the template merged 25 variables into the Variables panel, typed, grouped, and expandable: params, vendor names, contract folders, e-form configs, signature dates. Variable tracing is the job the 2022 original existed for, and it survived the rebuild intact: when a workflow of this size breaks, the question is almost always "where does this variable get set, and who reads it?"
A built-in command line logs every action (template added, variables merged) and accepts commands like alias nodes with friendly names, list them, or rebuild the start activity.
How it's built