A US commercial products distributor was drowning in manual operations. We replaced spreadsheet tracking, triple data entry, and a Friday reconciliation marathon with n8n workflows and a Claude approval agent. 52 hours saved every week, with full audit trail.
A US commercial products distributor in supply chain operations came to Aumadi to replace manual operations end to end. Aumadi took the full build: discovery, design, development, deployment, and post-launch tuning.
The result: 52 hours back every week, fewer errors, and an audit trail that did not exist before.
The team was drowning in manual operations.
Project tracking lived in spreadsheets. Data was entered by hand into three separate systems for every order. Receipts were reconciled once a week, by a person, in a Friday marathon nobody enjoyed. Approvals happened in Slack DMs and email threads with no record of who approved what or when.
As the company grew, every manual step compounded. Delays. Errors. Missed approvals. The cost was visible. The path forward was not.
The brief was clean: find the operations work that did not need a human. Replace it with AI agents and workflow automation. Keep the team in the loop where judgment mattered. Get them out of the loop where it did not.
The stack:
Nothing the team already used got thrown out. The new workflow sat on top of the existing tools, doing the boring work and routing the rest.
Two weeks of mapping. Every step in the operations workflow documented. Every delay timed. Every decision point flagged. By the end, the team understood their own process better than they had before, and Aumadi knew exactly which steps had to stay human.
The workflow design split into two parts.
The orchestration layer in n8n connected the existing tools. Triggers fired when projects hit certain stages. Data moved between systems without anyone typing it in twice.
The Claude agent on top had a defined job: read incoming requests, decide who needed to approve, route the request through Slack, track the response, log the outcome. Every decision the agent made was traceable in the audit log.
n8n flows for the data movements between ERP, QuickBooks, and Supabase. The Claude approval agent was prompted, tuned, and wired into Slack so requests landed as structured messages with action buttons. Supabase tables held state, run history, and the audit trail. QuickBooks API integrations powered live receipt matching, surfacing discrepancies as transactions came in.
The receipt reconciliation flow got rebuilt the most. The old process was a weekly batch job that produced results on Friday afternoon. The new process matched receipts against expected transactions within minutes of either landing. Discrepancies surfaced live, not at the end of the week.
The new workflow ran in parallel with the manual one for two weeks. Approvals went through both routes. Receipts got reconciled twice. Discrepancies between the two outputs were tracked, fixed, retested. Once the workflow caught everything the team caught and nothing more, the manual process was retired.
The team switched over without drama. Slack approvals replaced the old DM thread chaos. Receipt reconciliation moved from a Friday afternoon job to a continuous background process. Project records updated themselves as triggers fired.
Aumadi stayed on for the first two weeks of live running. A few prompt edge cases got tuned. A handful of false-positive flags got cleaned up. After that, the workflow ran itself.
The team saved 52 hours every week.
Errors fell. Missed approvals stopped happening because the agent followed up automatically when responses were late. The Friday reconciliation marathon disappeared. The audit trail showed every decision and who made it.
In the client's words: "The new setup is saving hours and hours of work. It is amazing."
The 52 hours a week the team got back now go into the work that actually needed a person. Judgment calls. Customer conversations. Decisions that matter. The business runs faster, with cleaner data and more confident decisions behind it.
Good AI does not replace people. It gives them their time back.
Six connected AI workflows turning the studio's lead, scope, and case-study admin into a system that runs itself. 50% faster lead replies. Zero projects lost between delivery and portfolio.
Embedding-based matching and voice-to-listing for a community trading app. iOS, Android, and a web admin panel, all from one FlutterFlow build.
.png&w=3840&q=75&dpl=dpl_YpJPN2qaFC53hJEcBToNGrkMGBqA)
Palmyst is an AI powered app built on Flutterflow. It’s a smart, interactive experience powered by modern tech.
30-minute discovery call. Scope, timeline, and fixed quote out the other side. No strings attached.
