AI Saves Us ~€1,600 per Month: Automating Document Processing
How we turned 4–8 hours of daily manual work into an automatic nightly cron by integrating Odoo, Dropbox and OpenAI. Concrete numbers and stack.
Damián Bucha · 2/17/2025 · 3 min read
The problem: hundreds of documents in task attachments
As a company that does installations, servicing and deployments all across Slovakia, we receive dozens of tasks in our project system (Odoo) every day. Each task usually has a photo of a handover protocol or a scanned PDF attached.
Manual processing meant:
- Open the task
- Download the attachment
- Figure out whether it's a handover protocol or something else
- Rename the file into the correct format
- Save it to a shared Dropbox in the correct folder
With 50 tasks per day that meant 4–8 hours of human work per day. Monthly, up to 160 hours of routine, frustrating work that nobody enjoys and that serves no strategic purpose.
The solution: a simple script, a powerful combination
We built a simple Python script that integrates three systems:
- Odoo, where it pulls tasks and their attachments from
- OpenAI GPT-4 Vision, which analyses the images and decides on the document type
- Dropbox, where it stores the resulting PDF in a structured form
How the script works
1. Loading tasks from Odoo
- Connection via XML-RPC API
- Retrieving tasks scheduled for the previous day
2. Fetching and analysing attachments
- Extracting files from tasks and decoding their content
3. Identifying the document type
- The OpenAI Vision API analyses the image and decides whether it's a handover protocol or another type of document
4. Saving to Dropbox
- Automatic renaming to a standard format:
XYZ_<id>_<name>_<date>.pdf - Saving into the correct folder by type
5. Logging
- Each run is recorded into
cron.log, including any errors - On error, the designated person receives a notification
Concrete savings
With 50 tasks per day, manual sorting takes 4–8 hours daily, let's take the middle, 6 hours.
- 6 hours × 20 working days = 120 hours per month
- At an hourly wage of €10/hr that represents €1,200
- At the upper estimate of 8 hours × 20 days = 160 hours × €10 = €1,600 per month
And that doesn't include the indirect benefits:
- Fewer errors in records
- Faster access to documents when a client asks for them
- People doing strategic work instead of routine
Tech stack used
- Python, orchestration
- Odoo XML-RPC API, input data
- Dropbox API, output storage
- OpenAI GPT-4 Turbo Vision, image analysis
- Linux Cron Jobs, automatic nightly execution
The whole implementation took a few days of work. Return on investment (ROI) was less than a month.
Conclusion: AI doesn't have to be a mega-project
AI is often talked about as a huge transformation requiring consultants, strategies and years of implementation. The reality is often much simpler. All you need is to find one specific painful process that repeats daily, and build a simple script around it with an LLM API.
If you have a similar process in your company (invoice extraction, email categorisation, ticket classification, document sorting), there's a high probability we can automate it so that it pays for itself within a few weeks. More in AI solutions.
Want to find out where AI will give you the fastest ROI in your company? Let's arrange a short consultation.
Damián Bucha
Omnius
