Vertex SaaS

Vertex inherited a graveyard. A previous agency had built dozens of automations, documented none of them, and left. Flows double-fired, some had silently stopped months earlier, and nobody on the team was willing to open them. On top of that, support was drowning: 48-hour resolution times, rising churn, and hiring more agents was not sustainable.

Automations nobody understood, breaking silently

We found 31 live automations. Eleven were firing duplicates, seven had failed silently months earlier and nobody noticed, and four were doing nothing at all while still consuming plan limits. None had documentation. The team had stopped trusting the system entirely and was manually double-checking work the automations were supposedly handling, which is the worst of both worlds.

Rebuild what earns its place, delete the rest

We rebuilt the flows worth keeping with proper error handling and alerting, deleted the ones doing nothing, and documented every remaining piece in language their team can read. Then we added the layer they actually needed: an AI support agent trained on their own documentation and ticket history. It resolves routine tickets directly and escalates the rest to a human with full context and a suggested resolution already attached.

No lock-in, because that is how they got here

Everything runs in their own accounts with credentials they hold. Every flow has a documentation page and a named owner. If they never speak to us again, the system keeps working and their team can extend it.

What changed