Cut a 2-day document workflow to under 2 minutes
“I just renewed today for another 3 months. I need to keep myself in these classes, keep going, see what everybody else is doing.”
Chris runs a small civil engineering business focused on program, project and construction management out of the San Francisco Bay Area. He has spent years inside project controls, working with data across BigQuery, SQL and Tableau, with Adaptive Work by PlanView as the project management layer underneath everything.
The problem he kept hitting was simple to describe and brutal to live with. Government agencies sent 25 to 30 page PDF reports that took days to manually parse. Word files were worse. Clients had archives of 30 or 40 documents, each seven or eight pages long, with no clean way to track changes across them. The data existed. Nobody could actually use it.
He had been tracking n8n on YouTube for about a month before joining. Tutorial after tutorial, but nothing practical was getting built. The CRMs he had worked with for a decade all carried the same trap: every renewal got more expensive, all the data sat inside someone else's database, and switching tools meant abandoning years of records.
"That's always really scary for me, because if I want to switch, how do you make the switch? You're basically stuck in one system forever." That data-ownership anxiety was the thing that finally pushed him to try n8n properly, instead of watching another tutorial.
The first build was a project management sync, version 26 by the time it shipped. A webhook trigger fires the workflow, which clears the relevant BigQuery table, then pulls fresh data from Adaptive Work across project names, tasks, resource links, non-labour costs, mileage, timesheets, expenses and time-phased projections. A series of code nodes parses each object into the right shape and writes it into BigQuery, where Tableau picks it up so clients can filter and explore the data themselves.
End to end, the sync runs in roughly 80 seconds. Whatever Chris updates in Adaptive Work flows through to the client-facing Tableau view in the same minute. The Tableau view is the part that wins clients over, because they can drill into their own data without him sitting on the other end of an email thread.
The second build is a Gmail-triggered document intake workflow, which is the one he calls his most fun build, the genesis of why he wanted n8n in the first place. The workflow checks a dedicated processing inbox every five to ten minutes, handles up to 20 attachments per email, and routes each one by file type. DOCX files copy across to Drive. PDFs get parsed, with Claude Sonnet 4.0 doing the extraction. The structured output lands in Google Sheets, and the email gets marked as read so it does not get reprocessed.
Chris layered quality checks on top. Totals get cross-referenced in two different places inside each parsed document. If they do not match, the workflow flags it as an error and waits for him to investigate, rather than silently writing bad data into the sheet. The result is a system that gets better as it runs, instead of one that quietly poisons his database.
Mid-build, the n8n 2.0 release dropped. Workflows he had been running on the older version started running noticeably faster on the new one. He stopped fighting performance and started shipping.
A two-day manual document process now runs in under two minutes, with fewer errors and a built-in path to data quality checks. "You've got 29 different documents that are 7 pages long with all this different data. The manual entry would be enormous, and meanwhile, I don't have to do anything, it just runs in the background."
I just renewed today for another 3 months. I need to keep myself in these classes, keep going, see what everybody else is doing.
The deeper shift is data ownership. Chris is no longer locked into any one CRM or project tool. If he wants to swap a database, he writes a workflow that moves the data. If a client throws a new file format at him, he extends the intake workflow. The freedom from vendor lock-in is, in his words, the part that surprised him most.
There is a quieter shift underneath all of that. Chris stopped being the consultant who manually pushes data around and became the operator who can sit between every system in his client's stack. "You just need one operator, which is what you're training us to do, who can connect those dots, and no one else has to get involved."
Chris sees the mentorship as a forcing function for ideas. Every time he ships a workflow it spawns the next one. The group calls are the place those ideas surface, because somebody else's problem is usually his next build. He renewed the same day his case study was recorded, and the reason was not the technical content. It was the queue of new problems waiting for him in the next cohort.
Anyone looking to save some time, especially as a business owner, you have so many extra things you have to do after hours and on weekends. Why wouldn't you automate those tasks to the best of your ability?
Built agentic work-order dispatch, became company's head of AI
Read →Rebuilt a 7-year content workflow into a system that runs in minutes
Read →Launched solo AI agency, 7 demos and first client in 30 days
Read →