Today
For trade fairs, the website, LinkedIn and press releases, people take photos in the workshop, reuse slides from customer projects and write about current orders. One photo shows a customer part under NDA, one slide a cycle time from a customer project, and one post makes the customer recognizable even without a name. Those who publish do not know the NDAs, and those who know the NDAs see the posts only when the customer calls.
An example
Norvelta is preparing its trade fair presence: 46 photos, a 30-slide presentation, three LinkedIn posts and a press release. The agent finds the 2K mold SW-2231 in four photos; it was built under NDA for an automotive supplier, with confidentiality until the end of 2028, and in one photo the customer's part number is legible on the insert. Slide 17 gives a cycle time from exactly this project, and one post describes customer and vehicle program so precisely that both are recognizable. The agent proposes photos of three molds whose customers have given reference approval according to the CRM. The sales manager swaps the images, removes the cycle time from slide 17 and deletes the sentence, then management approves.
Your data, wherever it lives
Photos and videos sit today in the media library and in project folders, presentations and press texts as files on drives, NDAs as scans in the DMS or a mailbox, projects in PLM and CAD, and reference approvals, if anywhere, in a note in the CRM. demi reads, understands and links them, with your rights. So the AI checks every publication for confidential content in minutes, and a person approves what goes out.
Where the data lives today
- Photos, video & audio
- Folders & drives
- Office files
- DMS & archive
- CRM
- PLM
- CAD & drawings
- E-mails
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, project folders, DMS, CRM, PLM, CAD and e-mail. demi replaces nothing and links every photo, slide and text to the molds, parts and projects it shows, and these to customer, NDA, confidentiality period and reference approval from the CRM. So for every image demi knows which order the part belongs to and whether the customer allows it to be shown.
- 2 · PermitPermission Layer
The NDAs and customer projects themselves are visible only to management, sales and project leads. Whoever prepares a post sees only what is blocked and why, without contract details. Nothing is published until an authorized person approves it, and every exception is recorded with its reason.
- 3 · ActIntelligence Layer
The agent recognizes molds, parts, drawings and engravings in photos and videos and matches them with 3D models and part numbers from PLM and CAD. In texts and slides it finds customer names, part numbers, quantities, cycle times and descriptions that make a customer recognizable. Every mark names the NDA or restriction it rests on and proposes a way out: crop, blur, replacement photo or different wording.
How it runs
- 1
Photos, slides, posts or a press release go into approval.
- 2
The agent recognizes parts, molds, drawings, engravings and customer names in images and text.
- 3
It assigns every find to project and customer and checks NDA, confidentiality period and reference approval.
- 4
It marks blocked content with the reason and proposes a replacement photo, a crop or new wording.
- 5
The responsible person swaps or deletes and approves; only then is anything published.
What changes
Confidential content is caught before publication, not when the customer calls
References only with customers who have agreed
Faster approvals, because nobody has to check every image by hand
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
+45 hfreed per year
Assumptions and calculation
demi checks images, slides and texts against NDAs, customer projects and reference approvals; you approve.
25 min → 17.5 min per case
How we calculate
30 publications × 25 min × 30% × 1.0 (manual) × 12 months = 2,700 min = 45 h
Value per year, before demi costs
Freed per year+45 h
Assumption: labor cost of $48.62 per hour in US manufacturing in June 2026 (BLS), plus workplace, IT and overhead.
= Value per year, before demi costs
+$2,925
gained per year
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
How does demi know which parts and projects are under NDA?
From the signed NDAs, customer contracts and project files, scans included, linked to customer and project in CRM and PLM. For every project demi records what is confidential, for how long and whether the customer has agreed to be named as a reference. If information is missing, the project counts as blocked until someone clarifies it.
Does demi recognize a customer part in a photo?
Yes. The agent compares what the photo shows with the 3D models and drawings from PLM and CAD and reads engravings, labels and part numbers in the image. If it is unsure, it marks the photo for review, and someone from project management or design decides.
What about texts that do not name the customer at all?
The agent checks those too. Industry, region, quantity and vehicle or product program together can make a customer recognizable, even without a name. It marks such combinations with a reason and proposes more general wording.
Can a project be shown once confidentiality has ended?
Yes. demi knows confidentiality periods and reference approvals and reports when a project's confidentiality ends or a customer has allowed a mention. Marketing then gets a list of the projects it may use for case studies, trade fairs and the website.
