Automation & AI work companion
Intelligent digitisation for the trades
On the office shelf: twenty years of project folders. On the server: "Photos_new_final", 40,000 images, none of them filed. In the desk drawer: measurement pads with pencil notes only one person can read. All of it valuable – none of it findable.
You'll get …
- 1
Records you can search
Delivery notes, measurements, project files and site photos are structured in your trade software, a spreadsheet or your project files.
- 2
Automatic recognition, spot-checked
The batch goes through automatically and someone in the office checks samples – handwriting included.
- 3
Your project history, put to use
Old files and job costings become easy to find, and new paperwork can be processed automatically if you want.
01Intro
AI can now read even poor-quality scans and handwriting: site measurements become structured dimensions, delivery notes become line items ready to book, and site photos sort themselves by project and trade. The records go through as a batch, someone in the office checks samples, and at the end the data is where you work – in your trade software, a spreadsheet or a knowledge base. Then the question "what did we fit at the Bergers' in 2019?" takes one search, not an afternoon at the filing cabinet.
There's no such thing as 100% recognition – especially with handwriting, someone always checks samples. What can go to the cloud and what stays local is agreed BEFORE processing starts. Audit-proof archiving (GoBD) remains the job of your specialist systems – we make records usable, we don't replace an accounting archive. And if a set of records isn't worth the effort, I'll tell you.
02How it works
- 01
Take stock
Project folders, the photo server, measurement pads, filed delivery notes – what's where, how much of it, and what should the data be good for afterwards? An afternoon spent taking stock saves weeks heading in the wrong direction.
- 02
Decide on the target structure
Where should the data go – trade software, spreadsheet, project files? And which fields do you need: project, client, trade, dimensions, amounts, date? The target structure determines what the recognition has to look for.
- 03
Run a pilot on a real batch
We take a representative sample – one year of projects, one type of folder, one month of photos – and set up recognition and sorting for it. The pilot shows what the records can yield.
- 04
Check quality and refine
Someone in the office checks samples: are the dimensions, line items and project assignments correct? Where the recognition gets it wrong, we refine the rules – until the accuracy rate is good enough.
- 05
Process all the records
Then everything goes through and ends up in the target system. If you like, the pipeline stays in place: from then on, new delivery notes, photos and scans are processed automatically too.
03Stages
The entry level is deliberately small – a single set of records makes a good, measurable pilot.
Starter: one set of records, opened up once
A clearly defined set of records – a year's delivery notes, a photo archive, a shelf of project files – is processed in a single run. The result: structured, searchable files.
Next step: the ongoing pipeline
The one-off job becomes a repeatable pipeline: new delivery notes, site photos and scans are structured automatically and filed in the right place.
Full build: records that do some of the work
The project data flows into a knowledge base: job costings, material lists and project history can be queried – the business's experience, available to everyone.
04Try it yourself
Three documents of the kind lying around in any business. Start the recognition – on the right you see what it reads out. For the flagged field, you decide: that's exactly what the spot-check is for.
The route behind it is a pipeline: recognition and rules work through the stack – and anything uncertain branches off to a human instead of quietly landing in the table. Only once the spot-check rate holds does the whole batch run through.
The dashed loop is the core: what's uncertain goes to the check and back – so the recognition gets better with each stack.
05Technical implementation
- Recognition: OCR plus AI extraction – modern vision models can read even crooked scans, stamps, carbon copies and handwritten notes on measurement sheets.
- Classification: document types (delivery note, invoice, measurement, photo, memo) are recognised automatically and sorted by project and trade; duplicates are weeded out.
- Photo archives: image recognition sorts by site, site-diary period and subject – "Photos_new_final" becomes a project gallery.
- Processing as a pipeline: scripts or a small processing setup – run locally or in the cloud, depending on how sensitive your material is.
- Review interface: a simple view for checking samples and making corrections. A person stays in the loop.
- Output: structured data sent to wherever you work – trade software, CSV/spreadsheet, project files or knowledge base.
for context · This shows how something like this usually gets built – not how it has to be built for you. What actually fits is worked out in the project: around your existing setup, your team, and what stays easy to maintain.