AIArticle
Usage time doubled: 114 minutes per user and week
In eight weeks, the time our customers spend working with demi has doubled. And that while demi now answers twice as fast.
· 2 min read
Read more →IndustryArticle
2027: the year machines have to explain themselves
Between 2025 and 2028, several EU rules for industrial products take effect. At their core they ask the same thing: know what is in your product, report fast, keep records for ten years. Documentation becomes part of the product.
· 6 min read
Read more →AIArticle
You hire an agent, you don't install it
Software is bought per user because a person does the work. Once an AI agent finishes a step of work on its own, that logic stops fitting. Why industrial companies will soon staff agents like a role.
· 5 min read
Read more →TechnologyArticle
Your old ERP is about to become more valuable
The usual story says AI replaces legacy systems. For industrial companies the opposite is closer to the truth. An agent is only as good as the record it can check itself against, and that record usually lives in the ERP.
· 5 min read
Read more →IndustryArticle
The new godemi.com: how it works
A tour of the new website: where to find use cases, industries, the platform and all the evidence, and how to find your way around fastest.
· 2 min read
Read more →TechnologyArticle
Kilowatts, not gigawatts: why industrial AI will be small
Headlines measure AI in gigawatts and data centers. A plant manager measures it in cost per answered question. Why small, specialized models are the better choice for factory work, especially in Europe.
· 6 min read
Read more →TechnologyArticle
Starting with AI, without a data project
Many AI projects start with a long data project. demi links your data as it is today. What that means for getting started.
· 1 min read
Read more →IndustryArticle
Projects with TUM chairs
Two chairs of the Technical University of Munich have posted project studies and interdisciplinary projects (IDP) with demi. Here are their public postings.
· 1 min read
Read more →AIArticle
Why your first year with AI looks like a loss
After six months, the controller's spreadsheet shows higher costs and no savings. That is not a sign that AI doesn't work. It is the pattern of every general purpose technology, and if you stop too early you drop out exactly when it starts to pay.
· 5 min read
Read more →AIArticle
Why AI projects in industry fail
Three figures from recent studies show where AI projects get stuck: the results, the data and the permissions. And what AI needs.
· 1 min read
Read more →IndustryArticle
When intelligence is free, what is your price built on?
Open AI models give every competitor close to the best available reasoning, at a fraction of yesterday's cost. For industrial companies, "we are smarter" loses pricing power. What remains is what no model can download.
· 5 min read
Read more →TechnologyArticle
How engineers think
Why every role at demi is an engineering role, and what we mean by that.
· 7 min read
Read more →AIPodcast
AI: efficiency or self-deception?
Philipp Munsch's second visit to “Kevin allein im Marketing”: on the difference between faster and more productive.
· Philipp Munsch
Read more →IndustryPost
When the knowledge retires
One sentence keeps coming back in sales conversations: he retires next year. What gets lost is not the procedure, it is the judgment.
· Niklas Löckel
Read more →TechnologyPost
Who may see what when the AI knows everything?
The first question in an industrial AI project is never about the model. It is: who gets to see, do and authorize what?
· Tobias Wolff
Read more →AIPost
The 30-second test for your AI
Ask your company's AI assistant one question about your own business. Not what, but why.
· Philipp Munsch
Read more →IndustryPost
The document holds the result, not the reason
You can connect every document store to an AI and it still will not know why the best machine in the building is built the way it is.
· Niklas Löckel
Read more →AIPost
Similar text is not the right answer
Many AI tools look for the most similar paragraphs and hope the answer is in there. For an industrial product catalog, that is not enough.
· Tobias Wolff
Read more →TechnologyPost
Where does our data go?
Every industrial company wants AI. Then the legal team asks where the data goes. Philipp Munsch on the answer demi is built on.
· Philipp Munsch
Read more →TechnologyPost
Answers from a model of the company, not from a prompt
Most AI tools generate answers from a prompt. demi answers from a structured model of the company.
· Tobias Wolff
Read more →IndustryPodcast
Industrial marketing in transition: from brochures to AI
Philipp Munsch on the podcast “Kevin allein im Marketing”: where industrial marketing stands between trade-fair stand, brochure and AI.
· Philipp Munsch
Read more →IndustryArticle
From gut feeling to strategy: AI marketing in industrial midsize companies
Philipp Munsch was on the program of the conference “KI für den Mittelstand 2025” in Munich with this pitch.
· 1 min read
Read more →IndustryPost
We have founded demi
Tobias Wolff announces the founding of demi, together with Niklas Löckel and Philipp Munsch: AI that truly helps Europe's industrial SMEs compete.
· Tobias Wolff
Read more →
