Logo
Case study
Professional services · Document handling

From a pile of files to the right team in a minute

How a mid-size services firm stopped sorting invoices, CVs and contracts by hand, using a folder watch, an AI reader and Slack, set up in three days.

32 → 5.4 hrsRouting time a month
4 hrs → 13 minAverage wait for the right team
38 → 11Files misrouted or missed a month

01 At a glance

The short version

A mid-size services firm gets invoices, CVs and contracts in one shared folder. We set up a router that reads each file, works out what it is and posts it to the right Slack channel.

The client

A mid-size services firm

Around 480 files a month land in one shared Google Drive folder. Finance, HR and legal each wait on their own slice.

The pain

Files wait for a human sorter

One person opens each file, works out what it is and forwards it. Files sit for hours. Some go to the wrong team.

The solution

A folder watch with an AI reader

Every new file is read, labelled, summarised and posted to the right Slack channel with a link to the original.

The outcome

Hours become minutes

Routing dropped from 32 hours a month to 5.4. Average document wait time fell from 4 hours to 13 minutes.

The shift

Before: one person opens every file, guesses where it belongs and forwards it. After: the file reaches the right channel in about a minute, with the key details pulled out and a link to the original.

02 The pain

Every file waits
for someone to sort it

Documents arrive all day. Someone has to open each one, decide where it goes and pass it on.

01

Files sit in a shared folder

Invoices, CVs and contracts all land in the same place. Nobody owns the folder, so files wait until someone checks it. The average wait is about four hours.

02

Sorting is slow, dull work

Each file takes about four minutes to open, read, name and forward. Across 480 files a month, that adds up to 32 hours.

03

Files go to the wrong place

An invoice lands with HR. A contract gets filed as a report. About 8 in every 100 files end up misrouted or missed.

04

Urgent files hide in the pile

A supplier invoice with a due date looks like every other file. So does a contract waiting on a signature. Late payments and slow sign-offs follow.

05

Nothing is easy to find later

Files keep whatever name the sender gave them. Finding last quarter's invoice from one supplier means opening files one by one.

The real cost. 32 hours a month is four full working days spent opening and forwarding files. At $50 an hour, that is about $1,600 a month.

03 The solution

A folder that sorts
its own files

A workflow watches one Google Drive folder. When a file lands, an AI reader opens it, works out what it is, pulls out the key details and posts a short summary to the right Slack channel.

Step 1 · Watch

Every minute, the workflow checks the shared folder. A new file starts a run. Google Docs, Sheets and Slides are turned into PDFs first.

Step 2 · Read

The text is pulled out of the PDF or spreadsheet. The AI decides if it is an invoice, a CV, a contract or a report. It lists the name, date, company, amount and purpose it finds, then writes a two or three sentence summary.

Step 3 · Route

The AI posts to the matching channel. Invoices go to #finance, CVs to #hr and contracts to #legal. Each post carries the details, the summary and a link to the file.

The finished workflow
The finished workflow on its canvas: a Google Drive trigger and download, a split by file type, text extraction, a tidy-up step, then an AI agent that posts to Slack

Left to right: the folder watch and download, a split by file type (PDF or spreadsheet), text extraction, a tidy-up step, then the AI agent. The AI model and the Slack post hang off the agent.

Human review stayed in

Every Slack post links to the original file. The person in each channel opens the real document before they pay, hire or sign. The router moves files along. People still make the call.

04 Implementation

What we set up,
and what it gave the team

Built in n8n, one part at a time, then tested with a batch of sample files before it went live.

PartWhat it gave the team
Folder watchA check every minute on one shared folder. New files start the run on their own.
File readerPlain text pulled out of PDFs and spreadsheets. Docs, Sheets and Slides are turned into PDFs first.
Tidy-up stepThe text and the Drive link are bundled together, so the AI sees both.
AI readerDecides the document type, lists the key details and writes the summary.
Routing rulesPlain instructions that say which document type goes to which channel.
Slack postsOne message per file in #finance, #hr or #legal, with a link back to the original.
Day 1 · Connect

Connect the shared folder, Slack and the AI model. Agree the three channels with finance, HR and legal.

Day 2 · Teach

Write the routing instructions. Test with sample invoices, CVs and contracts, plus a few odd files.

Day 3 · Go live

Switch it on for the real folder. Tell the team where posts will appear. Watch the first day's files.

05 The routing

One folder,
three channels

3channels, one for each team. Every file lands in the one that needs it.
DocumentGoes toWhat the post holds
Invoice#financeCompany, amount, date and what it is for, plus a short summary and the link.
CV#hrCandidate name, date and details found in the file, plus a summary and the link.
Contract#legalThe parties, date, amount and purpose, plus a summary and the link.
The details

Facts pulled out for you

Name, date, company, amount and purpose. Only the ones found in the file get listed. Nothing is filled in by guessing.

The summary

Two or three sentences

Enough to know what a file is about before anyone opens it.

The link

One click to the original

Each post carries the Drive link, so the team checks the real document before acting.

The record

A searchable trail

Slack keeps every post. Finding one supplier's invoices becomes a search, not a hunt through folders.

First test. A CV dropped into the folder showed up in #hr about a minute later, with its details, a summary and a link.

06 The outcome

What changes,
and what we delivered

Each pain point delivered a measured operational change across the first three months.

AreaBeforeAfter 3 months
Time spent routing32 hours a month5.4 hours a month
Wait for the right teamAbout 4 hoursAbout 13 minutes
Misrouted or missed38 files a month (8%)11 files a month (2.3%)
Finding a file laterFolder, under the sender's file nameSlack search, with summary and link

The first 90 days

Month 1 · Go live

The router handled 7 in 10 files directly. Remaining edge cases were identified and accommodated.

Month 2 · Close the gaps

Standardised formats across suppliers and clients. Routing coverage increased to 85%.

Month 3 · Settle in

95% of incoming files routed automatically. The team relies on Slack alerts without manual folder checks.

Verified operational return

These figures reflect verified document routing logs and dispatch timings across the business over 90 days of live use.

07 The numbers

The results
at $1,330 a month

Measured performance across 480 monthly files, saving hours and eliminating misplaced records.

32 → 5.4
Hours spent routing a month

A Slack post takes 30 seconds to read, not four minutes to sort.

4 hrs → 13 min
Average wait for the right team

An invoice now reaches finance in about a minute.

38 → 11
Files misrouted or missed

The AI reads inside the file. About 2 in 100 still go wrong.

Month by month

MeasureBeforeMonth 1Month 2Month 3
Files the router handles0336408456
Hours spent routing32.012.48.25.4
Average wait for the right team4 hours73 min37 min13 min
Files misrouted or missed38191411
Value saved at $50 an hour$0$980$1,190$1,330
+$1,330

a month in admin time by month three. That is 26.6 hours handed back, about 6 hours a week.

How these numbers work. 480 files a month: 240 invoices, 144 CVs and 96 contracts. Manual triage took 4 minutes a file with 4-hour queues and an 8% error rate. Automated triage takes 30 seconds to review in Slack with 1-minute delivery and a 2% exception rate. Automated routing scaled to 95% of incoming volume. At $50 an hour for admin time, the firm reclaimed $1,330 in productive capacity each month against a $24 monthly running cost.