second intellect
Book a review
Menu
From the ideas poolNotes from the Innovation Hub

what ideas
become.

In the Innovation Hub every piece of work starts with a question, not a product. Some questions end as a note. A few become tools that people use every day.

We treat the ideas pool as a small think tank: we notice where work gets stuck, write down a hypothesis and build the smallest version that can test it. Only when something holds up in everyday use does it become more.

This page documents what came of it – the starting question, the approach and what we learned along the way. Not as a list of references, but as open working notes.

Four theses that have held up so far
  1. T1

    The best AI is often invisible.

    It lives inside a workflow people already use – not in yet another chat window.

  2. T2

    Built for one business beats built for everyone.

    Software that knows a team's rules and habits saves more time than a tool designed to fit anyone.

  3. T3

    People stay the authors.

    AI sorts, drafts and reminds. Decisions, values and the final word stay with people.

  4. T4

    Start small, use it for real, then decide.

    A working prototype in daily use says more than any presentation. What does not hold up, we deliberately leave behind.

Two ideas that stayed
01

AgencyFlow

“The CRM that doesn't exist anywhere. Except here.”

Starting questionDoes an agency have to fit its work around off-the-shelf software – or can the software fit the agency?

A CRM for agencies and service businesses in the DACH region that brings time tracking, travel expenses under Austrian law, invoicing and tasks into one system. It grew out of an agency's daily work and is used there every day.

Approach

  • Austrian rules as a feature, not an exception: daily allowances, mileage rates and the 12-month rule are built in.
  • AI as an interface: through its own MCP server, assistants such as Claude or ChatGPT work directly – and with controlled access – with the data.
  • Small automations with a big effect – such as emails that turn straight into tasks, or SEPA payment QR codes on invoices.
  • Security from the start: row-level access rights in the database, two-factor sign-in and an external security audit.

What we learned

  1. Modelling a country's and a company's rules precisely takes work off the team that no generic tool can.
  2. An MCP server turns a database into something you can talk to: questions like “What is still open this week?” no longer need their own dashboard.
  3. Deliberate scarcity – by request only – keeps the focus on quality rather than growth at any price.
02

Inner World Picture

“Not a journal. Not a coach. A mirror.”

Starting questionCan AI strengthen self-reflection without replacing it?

A tool that lets people describe, in thirteen fields, who they are and where they want to go – from values and talents to short- and long-term goals. It turns this into a personal audio of about three minutes that they listen to every day.

Approach

  • People write, AI shapes: the content and the words come from the person; the technology turns them into something you can listen to.
  • Repetition instead of overload: one short daily audio and a weekly look back rather than endless journal entries.
  • Trust as a precondition: data is stored in the EU and can be fully exported and deleted.

What we learned

  1. With personal topics, value is decided not by the technology but by whether people recognise themselves in the words.
  2. The most effective form was the simplest: the same short audio every day – not new text every time you open the app.
  3. With intimate content, data protection is not an add-on; it is the reason people answer honestly at all.

Your question could
be the next one.

Many of our best projects began with a passing observation inside a business. If you have an idea that deserves a small, real test, tell us about it.

Share an idea