Today
Tens of thousands of photos and hundreds of hours of video sit on the NAS, on old hard drives and in project folders, named IMG_4711.jpg or by date. When marketing needs a picture of a plant in action for a brochure, a trade fair or a quote, someone searches for hours or books a photographer again. The best shots were taken long ago; nobody finds them when the deadline is close.
An example
At Montanis Hartsteinwerke, around 180,000 photos and 600 hours of video are scattered across the NAS, three old hard drives and one folder per construction site: drone flights over the quarry, blasts, the crushing plant, deliveries of track ballast and 0/32 base course. For a quote to a rail construction customer, sales searches for “track ballast being laid, rail site, summer”. Within seconds the agent finds 23 photos and four video scenes, each with date, delivery note and customer, and marks five images in which someone stands on the track without a high-visibility vest. For the brochure on quarry restoration it also finds the eagle owl nest in the old quarry face, photographed in 2019 by a shotfirer, and marketing approves 14 images.
Your data, wherever it lives
Your photos and videos live today on the NAS, on old external hard drives, in project and site folders, in SharePoint and in the media library, mostly as IMG_4711.jpg without any description; the orders, deliveries and customers behind them are in the ERP and CRM. demi reads, understands and links them, with your rights. So the AI finds the right picture or video scene in seconds, and marketing decides what is published.
Where the data lives today
- Photos, video & audio
- NAS & hard disks
- Folders & drives
- SharePoint
- ERP
- CRM
- PLM
How demi solves it
Three layers in this order: connect the data first, then the rights, then the AI.
- 1 · ConnectData Layer
The case needs data from the media library, NAS, old hard drives and project folders, plus the ERP, CRM and PLM. demi replaces nothing and indexes every photo and video scene: what it shows, when and where it was taken, and which order, customer, site and product it belongs to. demi groups duplicates and near-identical burst shots; the originals stay where they are.
- 2 · PermitPermission Layer
Everyone finds only shots from folders and records they may see; photos from damage cases or confidential projects stay with the departments responsible. The agent proposes images with recognisable people for publication only with recorded consent, and shots at customer sites only with the customer's release in the CRM. Marketing decides what is published, and every approval is recorded.
- 3 · ActIntelligence Layer
The agent describes every image and scene in words: machine, product, place, season, people and protective equipment. Using capture date and GPS position, it links each shot to a delivery, order and customer, and it transcribes what is said in videos. You search with one sentence in any language, and every hit shows file, timestamp and usage rights.
How it runs
- 1
demi reads photos and videos where they are: on the NAS, on old hard drives, in project folders and in the media library.
- 2
The agent describes every image and video scene and reads capture date and GPS position.
- 3
It links the shots to site, order, customer and product from the ERP, CRM and PLM, and groups duplicates.
- 4
You search with one sentence, such as “track ballast being laid, summer”, and get matching images and scenes with timestamps.
- 5
Before use, the agent shows consents, customer releases and usage rights, and marketing decides what goes out.
What changes
The right picture in seconds instead of hours of searching
Every shot with place, order, customer and rights
Fewer new photo shoots, because existing material is found
What does it bring you?
Check it with your own numbers
Every starting value is an assumption you can change. The calculation sits right below, step by step.
Your numbers
Result
+128 hfreed per year
Assumptions and calculation
demi finds matching photos and video scenes with place, order and rights; you choose and approve.
40 min → 24 min per case
How we calculate
40 image and video searches × 40 min × 40% × 1.0 (manual) × 12 months = 7,680 min = 128 h
Value per year, before demi costs
Freed per year+128 h
Assumption: labour cost of €49.50 per hour in German manufacturing in 2025 (Destatis), plus workplace, IT and overheads.
= Value per year, before demi costs
+€8,320
gained per year
128 h ÷ 1,760 h per role = 0.1 full-time roles
An estimate, not a promise. The starting values are assumptions for an industrial company with 150 to 500 people.
Combine with more cases and get the packageQuestions about this use case
Do we have to move or rename photos and videos for this?
No. demi reads them where they are: on the NAS, on drives, in project folders or in the media library. Descriptions and links live in demi; the agent only proposes new file names or a folder structure and applies them after your approval.
How does demi know where and for which order a photo was taken?
From image content, capture date, GPS position and file path. The agent matches them with deliveries, orders and sites in the ERP and CRM, for example the construction site that was supplied on that day. Where the match is uncertain, it says so instead of guessing.
What applies to people, customer sites and agency photos?
The agent marks images with recognisable people and proposes them for publication only with recorded consent. It proposes shots at customer sites only with the customer's release in the CRM. The usage rights of photographers and agencies are shown on every image, for example print only or only until a certain date.
Can service and engineering search the image archive too?
Yes, with their own rights. Service finds, for example, every photo of wear on crusher jaws or corrosion on a flange, and engineering finds older shots of a plant before its conversion. Photos from damage cases are visible only to those who may see the case file.
