TYPO3 AI-powered Workplace Hub - Private AI for Intranet

Idea Title

TYPO3 AI-powered Workplace Hub - Private AI for Intranet

What is my idea about? What is the problem we are solving?

With the right capabilities, an AI-powered intranet becomes a true digital workplace hub, helping employees in any organization work smarter, faster, and more collaboratively. Simply by making it easier to find information, including the source references, with broader search terms, queries and research, and take immediate action upon it, as in a ChatGPT.

Such an intranet will add value to an organization, in particular if it is a hundred percent sovereign. A main condition is that the chosen AI solution is not rented and stored in the cloud but implemented locally and less of a concern when it comes to compliance, privacy and security.

There are many ideas about the use of AI in a CMS. Since TYPO3 is recognized as a robust, enterprise-class open-source CMS that is well-suited for building secure, scalable, and highly customizable intranet solutions, it is a logical step to focus on a solid AI assisted intranet solution first, packed as an extension. This will be the basis for a more solid market position and spark development of other AI related ideas, like assistance in creating and maintaining page trees with optimized content and even complete rollout of websites based on what we call Private AI.

Private AI: refers to artificial intelligence systems and models that are developed, deployed, and operated within an organization’s own secure infrastructure—either on-premises or in a private cloud. These AI solutions are trained on the organization’s proprietary data and are accessible only to authorized users within that entity.

What is the potential impact of my idea for the overall goal?

Key functions that are implemented by the extension may be an 80-90% selection of the list below. This order seems the most one logical but priorities can shift.
Note:
For a thorough analysis of what is feasible as a delivery at the end of Q3, we have to be realistic.
We have a list of detailed functions that fit per bullet point, which will be revealed when the budget is assigned and after a bit more research. Feasibility cannot be confirmed for every detail now up front. Budget is the fixed factor in this project. We can bend time in a sense of spending more hours if necessary, but end of Q3 is solid. So we have to calculate with some slack in the delivery itself.
In order to establish a minimal usable subset, we take this list and focus on the most generic functions inside. For instance, it is likely that Voice access is not feasible in the first delivery. To make this list definitive, we may consult members of the TYPO3 Association.
1. Smart Search & Instant Answers: Quickly find information or get direct answers to questions across all company data.
2. Automated Content Creation: Generate and edit documents, emails, news, and training materials with AI assistance.
3. Personalized Dashboards: Receive tailored news, updates, and recommendations based on your role and activity.
4. 24/7 Virtual Support: Get immediate help with HR, IT, and company processes through an AI chatbot.
5. Onboarding & Training: Access guided onboarding, training modules, and progress tracking.
6. Task Automation: Submit requests, manage workflows, and receive automated reminders.
7. Knowledge Discovery: Discover related documents and resources automatically suggested by AI.
8. Security & Compliance: Benefit from secure access and automated compliance checks.
9. Analytics & Insights: View usage stats and feedback to improve content and engagement.
10. Voice & Multichannel Access: Interact with the intranet via chat, voice, or messaging apps.

Examples
A few tangible examples of User Interactions may underline the advantages of such an AI-powered intranet. Employees with access to the organizations intranet can launch questions and researches like for instance:
• “Show me the latest travel policy and summarize the changes from last year.”
• “Draft a training announcement for the new TYPO3 rollout.”
• “Who in my department has completed the cybersecurity training?”
• “Book a meeting room for tomorrow and invite my team.”
• “What are my outstanding HR tasks this week?”
• “Find all project documents related to ‘Project X’ and suggest potential collaborators.”
• “Launch a quick feedback survey for today’s department meeting.”

How does my Idea align with the strategic goals for TYPO3 v14.

The first 9 Strategic Goals for v14 are served with this project if we follow the right steps. We (ViBiS and Outpacr is a combined entity in this) had to make a choice between a GenAI toolbox as many people may have in mind as a short term solution, and the AI-powered Intranet that has a longer breath. The AI-powered intranet functions can drip down into the backend, the other way around is nearly impossible.

The choice of the intranet first is based on the focus on the public sector and small to medium enterprises that need a high level of AI quality and sovereignty to serve the citizens and store their organizations capital.

An AI-powered intranet is not a new concept and it is available in probably many places already. But as a TYPO3 extension it is new territory. And since the fast development in the last year, we can use state-of-the-art technical solutions.

Delivery Q3

In General

The delivery of the extension incudes MCP(Model Context Protocol)-router functionality. The customer can basically work with an AI stack of his choice. This means, the whole infrastructure including one or multiple LLMs and RAG (Retrieval-Augmented Generation). It is up to the organization and the TYPO3 integrator and other personnel to make the best choices how to configure the set up. This will be guided by documentation, examples, and probably the AI solution itself.

Components

The extension delivery will consist of these high-level components, according to the TYPO3 API standards for v13/14:

  1. An MCP Router (built in Python);
  2. A chat interface (UI) that works on LLM/RAG through the MCP Router and that handles all the functionality as described here above;
  3. Content synchronization with RAG through the MCP Router to feed the chat interface.

Variations

It might be a good idea to wrap the MCP Router in a separate extension. So the delivery then would be two extensions: the “AI-Powered Workplace Hub” and the “MCP Router for AI-Powered Workplace Hub”. But this is something we have to find out during the project.

This extension enables an AI set up as Private AI configuration on a local server as well as on a cloud system, and also on less private configurations. It also enables the inclusion of multiple LLMs.

Which budget do we need for this idea?

10000 Euro

My Name

Vincent Mans

AI outgrows the internet. So it has to be employed.

I see many valuable contributions in this round for Q3 and I would love to have most proposed functions available in the backend.

However it surprises me how few views our contribution “TYPO3 AI-powered Workplace Hub - Private AI for Intranet” has. It is good not to get too much taken by an AI hype, you may call me reluctant in that field - but it is not a hype.

As AI needed the internet to grow, it now outgrows internet.

If we want to stick to, and even expand the TYPO3 mission, it is mandatory to act, quickly and firmly, in this territory. There are initiatives already, but they all are kind of a Lorem Ipsum 3.0. Very useful, no mistake about that, but not the real power TYPO3 needs.

The real power is in the integrated use of AI solutions for enterprises - from small to medium to large. Where people are being laid off, it is part of the mission to enforce these people. A company has its DNA. With this DNA it is capable to excell. A way to work with, as a starter, is a good intranet solution.

If an enterprise runs its intranet on TYPO3, it will ask for a solution like this. If an enterprise takes TYPO3 into account in a tender, but misses this AI driven Intranet solution, I bet TYPO3 will fall off the list.

On top of it all, the solution we propose offers the functionality of using a Private or Local AI stack. Which is more secure and compliant, in times that we need it. And those times are now.

We can wait for Q4? With the pace of AI now, that would be not, least to say, the very best idea.

If you have questions, please reach out to me, vincent@vibis.nl, and I’m more than happy to answer them.

The execution of the project is in the hands of a joint venture of ViBiS and Outpacr.

I see a big list of possible features like “Automated Content Creation: Generate and edit documents, emails, news, and training materials with AI assistance.” or “Book a meeting room for tomorrow and invite my team.” but I don’t see how any of your examples can fit into TYPO3 as those look like just generic AI features.

Don’t get me wrong, we really need to think how AI can be brought to TYPO3 and we need the AI experts to tell e.g. what is missing in the core but we need real concepts which are not “Let’s put an AI button next to each input field” to generate some content but also not some buzzword bingo “Launch a quick feedback survey for today’s department meeting”

So my questions are: Do you have a rough idea how this can look like for TYPO3? IMHO the TYPO3 GmbH is also looking into those questions but I know nothing more

Thank you for your additions Georg, and you’re right. These simple examples were added in a very late stage, to make it more tangible for the mainstream audience… we here had trouble to calibrate the document as Roberto can confirm.

Surely these use case examples are not the core of the solution if you dig deeper. However, these are valuable enhancers of corporate workflows, and there is/will be demand for it. It is part of positioning TYPO3 in the market in the current era.

The concept is broader than that. It is about securing company’s DNA. You will find more hooks in the ten points list above the simple examples. It is hard to extract the most perfect examples now. First thing to do, parallel to developing the base infrastructure, is to prioritize requirements, as there may be so many.

Imagine yourself and your use of AI. Now, how would you have your employees work with AI?

This is not child’s play. We have long conversations with customers about functionality, and they first go like “well, just some content, print the invoices…” until they see that a data combination of “customer history + value of orders + new product releases + dropping market prices (or raising in the near future) + [fill in yourself]” and inform the customer ultra personal, may become a core process of the business. Keeping your AI and your customers close to you is essential.

It takes some conviction that an AI-Driven Intranet is the most powerfull tool to outpace competitors, I am aware, and I perhaps make the wrong examples to let it speak to tech people, but again, these are the easy to understand parts of a bigger picture.

By the way, it is not necessary that we do this for or with TYPO3. The main reason is that I happen to work with it for about 23 years now, I hooked up recently with a start-up Private AI company, and that the idea of an open, compliant, local AI system that supports enterprises in a holistic way, should resonate with the TYPO3 filosophy and values. And I’m sure TYPO3 needs it, in this quarter.

Thanks again for your remarks, it gave me the opportunity to - hopefully - elaborate and make it better to understand the purposes.

If you need to know more about the origanization and how the ideas evolve(d), please ask me.

Hi Vincent,

first of all, thank you for your passionate and thoughtful contribution — it’s clear that you’ve put a lot of strategic thinking into this, and I fully agree that AI is not just a trend. The focus on intranet use cases, enterprise readiness, and privacy-conscious AI solutions is indeed highly relevant. Especially for TYPO3’s strong positioning in professional environments and public institutions.

That said, I’d like to offer a few reflections from the perspective of ongoing core development and roadmap planning:

As you mentioned, AI-related initiatives are already underway within TYPO3 — most notably the GenAI Toolbox roadmap goal, which is currently being developed by a dedicated team in the core. To make real progress and avoid fragmentation, I believe it’s crucial to align closely with that effort rather than starting an isolated initiative.

While your proposal outlines a list of features, some of the terms used — such as “TYPO3 API standard” — are not fully understood especially their technical scope. It’s also not quite clear how the mentioned MCP component in Python would fit into TYPO3’s PHP-based architecture. If this component is essential to the project, more information would be needed on how it should be integrated, deployed, and maintained within TYPO3 environments.

Furthermore, the proposal refers to both TYPO3 v13 and v14, and appears to be based on implementing the solution as an extension. From experience, this approach can lead to problems with long-term adoption and sustainability — especially when the functionality is so closely tied to a roadmap goal. Developing this outside the core increases the risk of inconsistency, lack of maintenance, and limited community uptake.

In summary: this budget request clearly addresses important needs, and your motivation is highly appreciated. However, to make this work within the TYPO3 ecosystem, a much closer connection to the existing AI roadmap and technical foundations is essential for my point of view. Otherwise, we risk creating yet another isolated solution instead of building something truly sustainable and integrated.

I’d be happy to help in establishing a connection with the relevant people in the core team where your initiative could be a valuable part of the broader effort.

Best,
Oli

Thank you Oli, for your observations and time to respond, these are really helpful.

The approach of an extension that also runs in v13 has to do with current initiatives. Thanks to your explanation I understand that it’s not the best part of my proposal. On the other hand, we aim for a fast implementation and v14 will not see the light before halfway 2026.

So from busines sperspective, it would be great to be able to include it sooner. It may be a killer application for enterprises that consider TYPO3.

For us it is not really an issue to create the solution as an extension or right away for the core. The first option seems most safe to start with, it wouldn’t be the first time that an extension later on is included in the core. Or compare it to for instance scheduler, which is also loaded separately. I think these are details in this stage of making decisions.

And I surely understand the concerns about the Python based solution for the MCP component. Without this component, I don’t think a useful solution can be created, as every organization wil have its own AI stack. And I am aware of the fact that the latest versions of PHP can replace Python in AI solutions. The only thing is that Python has a known ecosystem. But this is something we can investigate. I prefer a PHP solution.

There are some parts we need to align before we start. I just hoped that phase could be part of this project. Are we in a hurry?

Let me copy a part of a response by my companion Bart from Outpacr, to some questions raised following this budget application:

“I really appreciate your input, and I totally understand that we need visible next steps.
But I also want to echo Vincent’s position: content creation alone won’t cut it long-term. Within 2 years, most content in enterprise settings will be AI-personalized at delivery. That turns content management into some kind of classification and metadata management.
So while buttons to generate text are helpful short-term, the strategic shift imo is this:
TYPO3 as a semantic operating system, not just a CMS.
That means:
structuring internal knowledge and context
enabling agents to act across structured records
providing explainable recommendations within workflows
That’s where we see the biggest opportunity for TYPO3: not competing with content generators, but leading in controlled, contextual AI, especially for organizations who want to own their stack.
We’ll think about some lower-level UI examples to make that bridge visible, but just wanted to share this framing as well.
My discussions with Vincent were primarily on intranet use, which I think is a different game of content creation. More serious, and a perfect fit with locally run (open source) AI.”

Cheers, Vincent

Hi Vincent,

thanks for your detailed reply and for sharing more of the broader thinking behind the proposal — I really appreciate the vision you and Bart are aiming for.

That said, I still feel that some of the key points I raised haven’t been fully addressed — and I’d like to highlight one of the most important again:

There is already a dedicated team within the core, actively working on this topic as part of the official roadmap.

To me, this is a crucial factor. Regardless of whether the idea starts as an extension or aims for core inclusion later, developing such a strategically relevant feature without direct alignment with the core team poses a serious risk — both in terms of long-term sustainability and technical direction. If we want TYPO3 to take a leadership role in this space, we need to coordinate efforts from the start, not merge visions after the fact.

Also, while I understand the reasoning for targeting v13 for earlier adoption, a short-term implementation shouldn’t come at the cost of architectural fragmentation. If the plan includes developing critical components outside of TYPO3’s core technologies (e.g. Python-based MCP components), this needs deeper discussion and alignment before moving forward.

I’d really like to see your ideas evolve as part of the bigger picture — ideally in close collaboration with the team already working on these topics within the core. I’m happy to support that exchange.

Best,
Oli

Hey Oli, thanks again!

We can replace Python by PHP as I already wrote. Python is de facto associated with AI, but I shouldn’t have mentioned the technique. PHP is just as good (and perhaps even better).

Apart from that, I see that alignment is important. Specifications may be decided on in a larger committee. I addressed that a bit earlier.

If this results in a situation that does not fit in this budget application round, let me know. Then I withdraw.

Some other project ideas also aim for implementation in v13, so I don’t see exactly why my particular request needs to aim at v14 alone. But that explains why my first take is at an extension for v13 that can later be part of the core in v14.

There are many ways in which a project like this can be managed for now and for the future, in a lifecycle, in collaboration with a dedicated team around it. A clear roadmap will help.

In my vision: start in v13, learn and improve, and incorporate solidly in v14. That way, quick wins can be achieved and TYPO3 can prove its innovative character (I read many complaints on LinkedIn from German origin, so in my opinion it is becoming a necessity).

Working in another set up than via a budget round is also okay of course. It is about the result.

We can dive into every detail soon, but we should have some confidence in getting support or budget. Because it takes much effort.

Cheers,
Vincent

Hello,
In addition to the questions from others:
As I remember from other contexts you are not a developer but using a freelancer for backend jobs in your project (correct me if i am wrong).

My question: I am wondering who would realize your idea?

just a FYI: drupal’s AI initiative just has been launched https://dri.es/files/drupal-ai-strategy-june-2025.pdf

Hi,

Not completely true. I play different roles.

I am a developer from origin (C++ at Ericsson and then JAVA at a bank, from 1998 tot 2005) but now I prefer a role in the center, that’s correct. For TYPO3 and PHP I sometimes include a freelancer and I take the architectural, test and controlling parts of the project (obviously next to the management tasks).

That being said, on this particular project we set up the project with Outpacr, an AI startup of very experienced people in this area and for instance block chain. There is no actual TYPO3 knowledge there, it’s the AI programming and PHP. For the technical connection to TYPO3 (how to build an extension) the project relies on me, and if necessary a developer from outside (might be you).

I think we’re well equiped.

Also I am a law graduate specialized in labour law and intellectual property, art painter and gallerist. Seems a lot but I’m 56 so it’s all possible.

That’s great, it’s progress at the neighbours :wink:

Want to do the exact same? Do you have a question or a remark?

There is a difference with this project I am proclaiming. We take a marketing stand with the AI Powered Intranet.

I see advantages in automated website creation, but really, is it the first thing people think of on enterprise level?

Like I said before, we considered a few of the topics this Drupal project addresses. I, for instance, would use one of the bootstrap package extensions as an entry hook to create and enroll websites based on let’s say a well organized prompt.

But we want to turn it around and make TYPO3 stronger where it is already powerful.

Perhaps the situation is different from Germany but in The Netherlands we are not in a strong position with TYPO3. I’m also afraid that TYPO3 will lose territory in Germany. I’m also afraid that waiting is not in the spirit of the founder, he’s an inventor (in my opinion).

What your opinion may be, this is part of the motivation for me to drop this project here.

In general sense, I feel strongly that the contribution of me and the team I work with, is not much appreciated. Assumptions are being made. I work for a long time in ICT, and with TYPO3 since 2003. I’m not joking here.

It is normal to have a different opinion, but the latest posts suggest that I have no business here.

On June 8th in this thread, I already popped the question, shall I withdraw this application. Hereby I repeat that question.

Hi Vincent,

I’m really sorry to hear that you feel your contribution isn’t appreciated — that’s absolutely not the message I intended to send. Quite the opposite: I truly value the ideas you’ve brought in and the long-standing experience you and your team have, both in ICT and with TYPO3.

Personally, I’d really like to see you involved — your perspective is important, and there’s a lot of potential in the vision you’ve outlined. At the same time, my intention was simply to point out that we already have related work happening in the core. I believe that closer coordination with the existing team would be the most effective way to bring all of this together and make the best use of everyone’s energy.

But in the end, that’s just my personal view. The TYPO3 community will ultimately decide through the budget voting process. I just wanted to share how, from my role, I think we could collaborate most efficiently to reach the common goal.

So no — I don’t think you should withdraw the proposal. I’d much rather see this turn into a productive conversation and possibly a strong joint effort.

Best,
Oli

Hi Oli,

Your posts relate to the matter of the budget proposal, so that’s more than okay and very necessary to calibrate the idea and project, and make it meaningful.

There are useless posts here, however, that do not add anything to the ideas. I wonder why. These contributions make me feel like I’m in the wrong place.

Obviously, the votes decide! No problems with that in a true democracy.

Meanwhile we proceed with this project, but without backup from the community it will look different. If there are teams working on this, as you say, let them reach out to me (ai@typo3cms.nl, forwards to vincent@vibis.nl).

Cheers,
Vincent