Data entry ops · Production

Miller Tanner saves 800 hours yearly by automating travel PDF data extraction with Box AI API

The problem

Miller Tanner agents manually transcribed data from incoming travel PDFs into their operational database, consuming roughly 800 hours per year, because every travel agency formats their booking PDFs differently, making fixed-format automated extraction impractical.

Workflow diagram · grounded in source
1
PDF uploaded to Box folder
Trigger
A staff member uploads incoming travel PDFs to a folder on Box, triggering the automated extraction process.
▾ source quote
“As PDFs come in, a staff member uploads them to a folder on Box, and the automated extraction process kicks off”
2
Ask API classifies document
Ai action
The Ask API identifies the document type, language, and airports listed, automatically converting airport names to codes.
▾ source quote
“The Ask API provides some important info about the PDF — for instance, what type of document it is, which language it's in, and which airports are listed (automatically converting names to airport codes)”
3
Extract API pulls traveler data
Ai action
The Extract API retrieves traveler name and details, agency information, flight information, train information, and other critical details.
▾ source quote
“The Extract API prompts for traveler name and details, agency information, flight information, train information, and other critical details”
4
Data saved to staging tables
Output
The workflow automatically saves the extracted data to staging tables.
▾ source quote
“The workflow then automatically saves the right data to staging tables, which can then be reviewed for accuracy”
5
Reviewer checks accuracy
Human review
A reviewer checks the staged data for accuracy before approving it.
▾ source quote
“which can then be reviewed for accuracy”
6
Data moved to operational database
Integration
Once the reviewer hits Save, the data automatically moves to the operational database.
▾ source quote
“Once a reviewer hits "Save," the data automatically moves to the operational database”
Reported outcome

A custom workflow built on the Box AI API automatically extracts unstructured data from travel PDFs and routes it to the operational database, saving 800 hours of manual entry per year and enabling Miller Tanner to serve customers faster and more accurately.

Reported metrics
Manual entry hours saved yearly800 hours
content stored in Box95%
Links shared82K
content migrated to Boxtwo-and-a-half terabytes
Show all 5 reported metrics
manual entry hours saved yearly800 hours
content stored in Box95%
links shared82K
content migrated to Boxtwo-and-a-half terabytes
content volume growth since migrationseven- or eightfold
Reported stack
Box AI APIAsk APIExtract APIBox ShuttleBox GovernanceBox NotesBox Sign
◆ Does this fit your context?

Compare to your context

Tell us your scale, team, and constraints. We'll show what changes at your size, what fails at your scale, and whether this case is a fit, needs adaptation, or won't scale to you. Free demo, no signup.

Compare to your context →
~30 seconds · free
Source
https://www.box.com/customers/miller-tanner
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

A custom workflow built on the Box AI API automatically extracts unstructured data from travel PDFs and routes it to the operational database, saving 800 hours of manual entry per year and enabling Miller Tanner to se…

What tools did this team use?

Box AI API, Ask API, Extract API, Box Shuttle, Box Governance, Box Notes, Box Sign.

What results were reported?

Manual entry hours saved yearly: 800 hours; content stored in Box: 95%; Links shared: 82K; content migrated to Box: two-and-a-half terabytes (source-reported, not independently verified).

How is this data entry ops AI workflow structured?

PDF uploaded to Box folder → Ask API classifies document → Extract API pulls traveler data → Data saved to staging tables → Reviewer checks accuracy → Data moved to operational database.

Related data entry ops cases

WHAT TO DO WITH THIS

Now compare it to your context

This case is one data point. Whether its pattern fits you depends on your volumes, your stack, and your exception load — that comparison is the step no case study can do for you.

Browse the Data entry ops cases →