The platform
Data first.Then rights.Then AI.For each use case.
Your existing systems stay, the Data Layer is built for them, your org chart becomes rights and only then does AI work. Nothing to structure up front, no data project.
Your systems
demi connects to every system
From SAP to the specialist software only your industry knows. If yours is missing, we build the adapter with you.
- Your systems stay
- demi writes only after approval
Microsoft 365
Google
Dropbox
SAP
Salesforce
Dynamics 365
Infor
HubSpot
Power BI
SharePoint
Outlook
Teams
Workday
- And every other systemSearch all systems
Choose your systems: ERP, machines, databases, legacy systems
Click the systems you work with and request a demo with your selection. That way, we know what matters to you before the call. Your systems stay as they are. We set up the connections with you.
ERP
39 systemsPLM, CAD and PDM
28 systemsECAD and plant engineering
17 systemsMES and shop floor
21 systemsCAQ and quality
15 systemsMachines, PLC and SCADA
30 systemsHistorians
11 systemsProtocols and interfaces
34 systemsDatabases (SQL and NoSQL)
23 systemsLegacy and host systems
15 systemsData standards and formats
31 systemsProgramming languages
14 systemsDMS and archives
27 systemsCRM and CPQ
22 systemsPIM
14 systemsEDI and procurement
15 systemsService
13 systemsBI
18 systemsHR
15 systemsMail and office
17 systemsDon’t see your system? We’ll connect it.
Legacy system, in-house build or machine: we build the connection for you. Your system stays as it is. Your team works on it with AI straight away.
The data standard
DemiDocs: the open data standard
DemiDocs is the format demi keeps everything in: documents and business data as blocks with their source. Open, because it can be read as Markdown and JSON, even without demi. Everything that comes in goes through four steps.
Standardized
Every source ends up in the same model: whether PDF, spreadsheet, mail or record, it becomes typed blocks.
Data should be findable, accessible, interoperable and reusable, and that goes for machines too: the FAIR principles. Wilkinson et al., 2016
Normalized
Numbers, dates, units and identifiers take one uniform form. The original spelling is kept.
The need for data cleaning rises sharply once several sources come together. As early as 2000 it counted among the biggest problems of data warehousing. Rahm & Do, 2000
Harmonized
The same customer, the same part in two systems becomes one record. Every value keeps its source, and contradictions stay visible.
When two records mean the same thing was put into mathematics in 1969: match, non-match or possible match, at stated error levels. Fellegi & Sunter, 1969
De-noised
Headers and footers, page numbers, repeats and hidden text are recognized and removed. What remains is what the document really says.
92% of 53 AI practitioners interviewed in high-stakes fields in India, East and West Africa and the USA reported data problems that compound unnoticed. Sambasivan et al., 2021
Everything goes into any format
What is in DemiDocs can be converted into any format.
- JSON
- Markdown
- DOCX
- XLSX
- PPTX
- CSV
- HTML
- SVG
- PNG
- EML
- and any other
Your data stays yours
Your data belongs to you: export at any time as JSON or through the API, and for at least 30 days after the contract ends. What applies to model training on your data is governed by the DPA (§ 3 para. 3).
Open standards shape industry at every level. DemiDocs brings that idea to your company's data.
How demi works
Three layers on the systems you already use
Connect first, then set rights, then let AI work.
- Data Layer
One picture of your whole company
demi links customers, products, orders, machines and documents: one record per product, customer and order. Every value keeps its source, so you can check it.
The Data Layer in detailExample with fictional data
Centrifugal pump KP-500
- Stadtwerke Nord · CRM
- Order A-2024-118 · ERP
- Datasheet, page 4 · PDF
- Permission Layer
Your org chart becomes rights
An AI agent never has more rights than the person it works for. If an agent has read content that may carry instructions from third parties, every send, publish or share to the outside needs a person's click. A log shows who acted for whom.
The Permission Layer in detailExample with fictional data
Anna's agent, Sales North
- sees: customers and quotes North
- does not see: personnel files
- writes: after Anna approves
- Intelligence Layer
AI that calculates and shows its sources
People and their AI agents do routine work together. Answers name their sources, and figures are calculated, not guessed. The right AI model for each task; agents work with models in EU data centers.
The Intelligence Layer in detailExample with fictional data
Which customers have a KP-500 older than eight years?
- 14 customers, 5 with a service contract
- Sources: ERP · orders, CRM · service
Apps
Apps on the Intelligence Layer
Every app works on the same Data Layer, with the same rights. No new silo.
- demi SpecReads tender packages and returns one reviewable list, every value with its source.
- demi AskAnswers technical inquiries from your own product knowledge.
- demi ChatAsk your company's knowledge, within your rights.
- demi AgentsRun recurring work on a schedule, in your role.
- demi ProjectsTurns a goal into a plan and hands out the tasks.
- demi TalkSpeak instead of typing; speech recognition runs on your own computer.
Your first day
The first case in about a day
Read-only on your systems. Nothing is replaced.
1. Connect
Your systems stay
demi works with read-only rights and connects to every system. If an adapter is missing, we build it with you.
2. Set rights
Everyone sees only what they’re allowed to
If someone can't open a file today, they won't see it in demi either. That holds for people and their agents.
3. Work
The first case runs on real data
A spare-part inquiry, for example: demi answers it from your data, with sources, and writes back only once you approve.
Create task: replacement quote for Stadtwerke Nord
Based on: ERP · orders, datasheet p. 4
See demi work on your case
In 20 minutes we show you demi on a case from your industry: bring the data together, set the rights, let AI work.
