DASTech Consulting built a stage-aware project pipeline board in Azure DevOps Boards for a legal-technology consultancy. Every client migration is now one card that travels through 13 stages in a swimlane for its target platform, and each card fills in its own task checklist as it advances. We also reconciled the board against live matter data, cleared out demo and test records, and delivered a handoff package for a non-technical administrator. Jason Da Silva did the work in tandem with AI, hands on inside our customer's live systems, and personally carried out anything that could not be undone.
Illustrative board. Client names, matter numbers and volumes are anonymized. Dragging a card into a new stage rewrites its task checklist using the task set for that stage and that platform.
Our customer is a legal-technology consultancy running many concurrent client projects, each one a law firm moving onto a new practice-management or document platform. The problem was never the work, it was the tracking: the pipeline lived in people's heads and in manual lists.
That set the clock: a board only one person can interpret is not an asset, it is a dependency.
Three things made this work.
One consultant, working with AI. Not an AI agent running unattended with a human signing off afterward, and not advice delivered from a distance. Jason worked side by side with AI across live systems: reading the Clio and Azure DevOps APIs, designing the stage-by-platform task logic, publishing and debugging the automation, wiring the service hook, creating and cleaning up board items, and producing every polished deliverable (a written document, a spreadsheet, an interactive HTML reference and a branded PDF). He directed every decision and carried out anything irreversible himself. What the pairing bought was pace: one person moving through discovery, build, debugging, reconciliation and documentation through work that would normally be split across a team. That is the capability DASTech now sells.
No-code plumbing. Nothing here asks the consultancy to own a codebase. The integration runs as an Azure DevOps service hook into Zapier, using Catch Hook and Code by Zapier. We wrote and configured the code step, and the administrator never has to open it.
Workflow design before any building. The request arrived fuzzy: keep their board in sync with their matter system and hand it off cleanly. We decomposed that into a board data model, an integration architecture, a data-reconciliation pass, a cleanup pass, and a documentation and handoff layer, sequenced so each stage de-risked the next. Automation laid over an unclear model only produces wrong answers faster.
Two failures are worth naming, both silent.
The first was a field-mapping bug: the automation read the wrong part of the message Azure DevOps sends. We had built it against a test payload, the data the hook receives, that arrived form-encoded, while the real event payload arrived as JSON with a different field shape. It ran, and cards were silently skipped. Working through the live payload with AI, we traced the mismatch, remapped to the real field and verified end to end.
The second was a naming mismatch. We found a stage column that had been renamed out of sync with the automation's lookup key, the name it matches on, and brought the two back into line.
We worked inside the same guardrails throughout: reversible soft-deletes only (items sent to a recoverable recycle bin), Jason performing anything irreversible himself, and access tokens and credentials never handled by the AI tooling.
The board was always going to change hands, so the deliverable was never just a working system. It was a system a non-technical person could operate on day one.
The DevOps 101 cheat sheet covers what each lane and stage means, how the auto-checklist works, day-to-day how-tos, watch-outs and a plain-English glossary. It ships with the written handoff document and the reconciliation and import spreadsheet.
This is the change the consultancy felt first. Previously, knowing where anything stood meant asking: a stand-up to surface blockers, chase updates, and work out who was stuck on what. That is a familiar agile habit, and for a distributed set of client projects it is an expensive one, because the meeting exists mainly to rebuild a picture that should already exist somewhere.
Now the board is that picture. A card's column is its status, so blockers are visible by position rather than by discussion, and the team can see at a glance where someone is stuck and where they can pick up slack for each other. Coordination that used to need a meeting now happens by looking. Across the team, that is saving tens of hours a week.
With the pipeline current, the checklists building themselves and the sync running, the interesting question shifts from keeping the board accurate to what the board can now tell them. A pipeline with real stage data underneath it can start answering planning questions: where work reliably stalls, which platform migrations take longest, and where capacity is about to be tight. That is the natural next step, and it is only possible because the underlying data is now trustworthy.
No. The board runs in Azure DevOps Boards and the automation runs on a service hook into Zapier using Catch Hook and Code by Zapier. We wrote and configured the code step, and the board and its documentation are built to be read and operated by a non-technical administrator.
One consultant hands on in your live environment, with AI working alongside him rather than running unattended: reading the APIs together, drafting the stage logic, debugging the live automation, producing the deliverables. Jason Da Silva directs every decision, performs anything irreversible himself, and handles credentials personally. The result is one person moving at the speed of a team.
Yes. On this engagement the guardrails were structural rather than promises. Deletions were reversible soft-deletes into a recoverable recycle bin, never hard deletes. The consultant performed anything irreversible personally. Access tokens and credentials were never handled by the AI tooling.
This one compressed what would normally run into weeks of setup and documentation into a small number of focused working sessions. The variable is not the building, it is the clarity of the workflow underneath. Where the process is already understood, work moves quickly. Where it is not, the design comes first.
If your operations run on undocumented process, manual checklists and systems that do not talk to each other, that is the problem we solve. See AI Consulting and Workflow Automation for how we scope this work, or get in touch to talk through your own pipeline.
We map where AI and automation genuinely save time in your business, put the guardrails in place, and hand you something your team can actually run.
See how AI & Automation works →