My AI team now works independently: Notion Custom Agents in practice test

As a TYPO3 integrator and trainer, you wear many hats at the same time. Customer projects, course preparations, new TYPO3 versions, community engagement, content, administration. And mostly as an individual or in a small team, with no one in the background to coordinate everything.

I have been using Notion as my central working tool for a long time. Databases for projects, tasks, content, customer contacts. Everything in one place, structured and accessible.

A few months ago, Notion AI changed this significantly once again. The AI has access to the entire workspace, knows my documents, my projects, my style. For me, that was the point at which I almost completely left ChatGPT and the like behind. Notion AI can do most things better because it knows the context of my business.

Since February 24, 2026, there has been another level: Custom Agents. And that changes a lot.

What is Notion?

For those who don't know Notion yet: Notion is an all-in-one platform for knowledge, tasks and projects. At its core, it is a combination of note app, database tool and wiki. Once you have a clean Notion setup, you hardly need any other tools for your day-to-day business. In my case, the workspace contains all active customer projects with tasks and status, my course and training documents, content pipelines for blog and social media, notes on TYPO3 versions and technical topics as well as contact and partner information.

What is Notion AI?

Notion AI is the integrated AI in Notion. The decisive advantage over ChatGPT or other AI tools: Notion AI has direct access to all content in the workspace. No copy-paste actions, no context building in every new chat. The AI already knows what you're working on, who your customers are and what your style is.

I work in Notion with specialized AI roles that I set up as a virtual team. Each role has its own personality, focus and clear areas of responsibility:

• A business advisor who critically scrutinizes new product and course ideas, points out weaknesses and suggests alternative approaches.
• A copywriting specialist who is familiar with various methods and techniques and writes texts for landing pages, sales emails or webinar invitations.
• A blog specialist who knows my target group, my style and my topics and turns them into readable, substantial articles.
• An editorial assistant who checks and smoothes finished texts for style and brand conformity.
• A prompt specialist for image-generating AIs who creates ready-made image prompts from blog articles or descriptions for tools such as Midjourney or Nano Banana.

These are just a few examples. Overall, I have built up significantly more specialized roles, including SEO, content strategy, product decisions, design briefings and external quality assurance. Each role has its own job description, clear areas of responsibility and access to the information relevant to it in the workspace.

That works well. But there was a catch: I had to address each of these roles manually. "Hey, analyze this." "Hey, write me something about this." All initiated via a chat, all started manually.

What are custom agents?

Custom Agents are a new feature in Notion that has been officially available since February 24, 2026. The difference to the previous AI team is easily explained:

The specialized AI roles wait for me to address them. Custom agents do not wait.

A custom agent can start automatically when a certain event occurs: a new entry in a database, a status change, a message in Slack. Or it can run based on time: every day at 6 a.m., every Monday at 9 a.m., whenever it makes sense.

In short: the previous AI team works on call. Custom agents work independently.

My four custom agents

On the day of the launch, I immediately set up and tested four agents.

Transcript auto-analysis

I regularly transcribe interesting YouTube videos that could be relevant for my business. These can be videos from the TYPO3 community, but also content on business, marketing or product development for the self-employed. I use TurboScribe, a tool that delivers very clean transcripts. I import these transcripts into Notion.

Previously: import the transcript, switch manually to the video analyst, start the analysis, wait.

Now: import transcript, done. The agent recognizes the new entry, analyses the transcript automatically and creates a summary, key messages and concrete ideas that I can use for my business.

This agent already works very well and actually saves me several manual steps.

Morning Briefing

Every morning in Slack: a structured overview with all tasks and content entries that are due or already overdue today.

Sounds simple, but it's practical. I already have linked database views with preset filters and sorting that give me a good overview. But the Slack message comes to me, I don't have to open Notion for it. This is a different mode. The potential is there, I'll see in the next few weeks whether it proves itself in everyday use.

Tally assistant

Tally is a form tool that I use for surveys, feedback forms and registration forms. The Tally Assistant is connected to Notion via MCP integration.

This means: I describe in Notion what a form should do, which fields it needs and what the goal is. The agent then creates the form automatically in Tally. Without me having to work in Tally.

This is still in the testing phase, but the first test was convincing.

Slack Assistant

This agent is the most interesting for me in my everyday mobile life.

I write or dictate unstructured messages into a specific Slack channel: an idea, a task, a note, a link. The agent analyzes the text, recognizes what is meant and creates the appropriate entry in Notion: Task, content idea, note or resource.

The advantage: I don't have to open the Notion app. Short voice message in Slack, phone in my pocket. Notion is updated in the background.

Whether I will use it every day remains to be seen. But the initial reaction was: amazingly good.

What's important when setting up

One major difference to normal Notion AI: custom agents do not have automatic access to the workspace.

Each agent starts without authorizations. You have to explicitly define which pages and databases it is allowed to see and whether it is only allowed to read or also write. This requires some preparatory work and careful consideration.

This is not a disadvantage. On the contrary: it forces you to think carefully about what an agent really needs. A morning briefing agent needs read access to the task database, but no write access. The transcript agent needs write access to the resource database, but not to all other areas.

The authorization concept makes custom agents more secure and controllable.

Ideas for TYPO3-specific custom agents

For the sake of completeness: What could be useful for TYPO3 integrators and freelancers?

• An agent that automatically creates a checklist of known breaking changes and upgrade steps in Notion when a new TYPO3 version is released.
• An agent that reports due customer projects daily in Slack, including the next open tasks.
• An agent that automatically creates a project framework in Notion based on customer briefings (tasks, milestones, relevant contacts).
• An agent that processes new support requests from a form tool and creates them directly as a task in Notion with priority and context.

These are just initial ideas. What actually makes sense depends on your own workflow.

Is Notion worthwhile for TYPO3 freelancers?

Short, honest answer: Yes, but not overnight.

Notion AI has been a real turning point for me since the last major update. I no longer use ChatGPT on a day-to-day basis, and I rarely use other AI tools. The decisive advantage is the context: Notion AI knows my business, my projects, my customers, my style. This makes every chat more productive than an external AI that I have to explain everything to first.

Custom agents are the logical evolution of this. An AI team that works on demand becomes an AI team that thinks and acts independently.

Training takes time. A functioning setup takes a few days to weeks. And it requires the Notion databases to be properly structured. If you invest in this, you get a tool that grows with your business.

At the beginning of this article was the picture with the many hats. Customer projects, course preparations, new TYPO3 versions, community, content, administration. All at the same time, usually alone.

Custom agents don't solve this completely. But they take over some of the coordination that used to depend on me. This is not a revolution. It's a further step towards a setup that no longer depends solely on my attention.

I'll see how this works in practice. I'll report back.

Do you use Notion or are you testing custom agents? Let me know what you experience.