AI software from Hamburg

AI software that sells.

We set up automated workflows and write the applications that go with them, for mid-sized companies and corporations. It starts with a workshop day where we find your biggest revenue lever. Then our IT team builds the software, and servers, operations and data protection stay in-house.

Book an AI workshop First call with Florian

AI Hands-On Day: €2,490 fixed price per group, net of VAT, for business customers only.

Recipe overview of Mein Ernährungsberater, a web app we built
Map with charging points in the VoltExchange app we built
  • Gebr. Heinemann
  • I. Schroeder KG
  • Global Gruppe
  • Terra Ingenieurgesellschaft
  • Hamburg Airport
  • R.Evo
  • Manchester City
  • studiomuc
  • GeheimPunkt
  • Dinoreiten
  • Geo-Concepts
  • Jütro

An AI workshop without software is an expensive company outing.

Between personal productivity and the operating result lie workflows that still run the same way after the training. That's why our workshop is the first phase of the build: by the evening, it's clear which three workflows come first. Then our IT team gets to work, with no steering committee and no handover to another firm.

80% say AI makes them personally more productive.
37% see some of it in their operating results.

Source: McKinsey, The state of AI in 2026 (opens in a new tab), August 2026, 1,719 respondents.

The biggest lever

Where AI brings in the most revenue in your business

McKinsey asked in 2026: revenue gains from AI come most often from marketing and sales. Cost savings tend to come from supply chain, service and production. That's why we start where the money is already on your doorstep: with the enquiries you already have.

  1. Enquiries
  2. Close rateWe start here
  3. Order value
  4. Repeat business
  5. Revenue

Close rate

Every enquiry gets a good answer right away.

An AI agent reads every enquiry from your inbox or web form, pulls in the knowledge base and customer history and writes the draft. Standard cases are answered in minutes, anything unclear goes to your team. You decide what can go out on its own and what a person approves.

In practice: Since studiomuc started using AI in its communication, it has closed 19% more deals (as of October 2026). See the case (in German)

Example illustration, invented data.

Close rate and order value

Your quote is out before the competition has run the numbers.

An email, a PDF or a tender becomes a finished draft: line items identified, prices from your price list or ERP system, wording in your tone of voice. The agent also suggests matching accessories. Your sales team checks, edits and sends it.

In-house: Our own quotes are made the same way, in a quoting engine we built ourselves.

Example illustration, invented data.

Enquiries

More people find you, in other languages too.

Articles, product copy, landing pages and translations run as a routine, with research, review steps and in your tone of voice. On top of that, we keep it all in shape for Google and for AI search like ChatGPT.

In practice: Mein Ernährungsberater and eatmondo, our own product: an AI pipeline for recipes, 14 languages. See the case (in German)

Example illustration, invented data.

Repeat business

Existing customers hear from you again.

An agent reads your customer list, shop or ERP system and notices who would normally have reordered long ago, whose contract is running out and where an add-on product fits. Every week, your team gets a list with messages ready to send.

Typical use case: In the workshop, we use your customer list to work out which signals count for you.

Example illustration, invented data.

Time for selling

The back office helps pay for the rest.

Reading receipts, building reports, maintaining data, writing minutes. None of that brings in a single euro of revenue, but it takes work off your people that keeps them from selling.

Typical use case: We get to it after the revenue levers, unless something is on fire.

Example illustration, invented data.

Check the numbers

What would your lever be worth?

Put in your own numbers. The calculation is out in the open, and nothing is stored.

120
20%
€4,000
30%
10%

Extra revenue per year

€115,200

You keep €34,560 of that. The AI Hands-On Day would pay for itself after 27 days.

€2,490 €34,560

120 enquiries × 12 months × 20% makes 288 closed deals a year. Another 10% means 29 more deals, times an order value of €4,000.

Work it out with real numbers in the workshop

An example calculation with your figures, not a promise. For comparison: since studiomuc started using AI in its communication, it has closed 19% more deals. We'll work out what's in it for you with real numbers.

What we build

Automated workflows and the applications that go with them

AI software means applications and automated workflows in which an AI model takes over tasks that people used to do: answering enquiries, drafting quotes, extracting data from documents. Collective Brain in Hamburg develops this kind of software for mid-sized companies and corporations, connects it to existing software and runs it together with Progressive Mind from the same group of companies.

  • AI agents

    They work in your inbox, CRM, shop and accounting, every day following the same process and the same checklist. Each one gets a measurable goal before we build it, such as response time or turnaround time.

  • Automated workflows

    Reports, data maintenance, receipts and handovers between programs. What runs on copy and paste today runs on its own afterwards, and you can see when it ran.

  • Custom applications

    Web apps, portals, booking tools, iPhone apps and internal tools, when the software for your process isn't available to buy.

  • Integration and operations

    We connect to the programs you already have and keep everything running: servers, updates and the number your team calls when something gets stuck.

32% of McKinsey's respondents have already passed on buying at least one piece of software because they could build it in-house with coding agents. If you have no such team in-house, use ours.

Proof, not claims

Already running

Two of them are our own products. That's on purpose: if you run software yourself, you know what it gets up to at night.

The studiomuc website on a tablet Client project studiomuc AI in communication, automation, website 19% more closed deals since using AI in communication, as of October 2026 See the case (in German) Home page of the VoltExchange charging community A project from our group of companies VoltExchange Web app, iPhone app, AI agents in operation 100+ members in the first weeks after launch, with no media budget, as of July 30, 2026 See the case (in German) Recipe overview on the Mein Ernährungsberater platform Our own product Mein Ernährungsberater and eatmondo Web app, AI pipeline for recipes 14 languages, as of July 30, 2026, just over five weeks after the project started See the case (in German)

From client projects, without names

  • Food company from HamburgContent routine: it finds topics, does the research and creates copy and images. At the end, a finished draft sits in the CMS, and a person publishes it.
  • Engineering firmBuilt in the workshop on real cases from the staff: a plausibility check for lab values with 16 rules and ground investigation planning from the borehole list.
  • Group of companies in the food industryAI-assisted review of 192 pages across five websites, plus an approval tool for everyone involved.

Check for yourself how we work: AI Leaderboard, Claude Skills, Claude cost calculator (in German). All built by us, all public.

The path

From the workshop to running software

  1. Day 1

    AI Hands-On Day

    One day, 9 am to 5 pm, up to eight people from your team, working on your real tasks. By the evening, it's clear which three AI agents start first in your business.

  2. Two working days later

    Fixed quote

    Goal, price and date for each agent on one page. We only get started once you have agreed.

  3. Weeks 4 to 6

    The first software is running

    The first use case is usually in operation after four to six weeks. Tested on real cases, checked for data protection before launch, built alongside your day-to-day business.

  4. Every month

    The next agent

    Three to start, then one more every month. Plus an update call, maintenance and IT in the background.

Work out what could go wrong.

Worst case

You pay €2,490 net and have a plan.

Your team works with AI more confidently, and you keep the handout, a record for your documentation under Article 4 of the EU AI Act, 30 days of email support and the list of workflows worth automating. Even if you never hear from us again.

Best case

In a few weeks, software that sells is running.

The first agent is usually in operation after four to six weeks. With the AI Partnership, the cost of your workshop day is waived, and in the first three months you can leave without giving a reason.

Book the AI Hands-On Day

Up to 8 participants, with us in Hamburg or Herrnhut, or at your site with travel costs on top. All prices net of VAT, for business customers only.

Why us

What holds mid-sized companies back, we cover in-house

In summer 2026, Bitkom asked 603 companies with 20 or more employees what is holding them back with AI. The answers read like our job description.

4 to 6 weeks

We're fast

That's how long it usually takes until the first use case is running. Small routines are quicker: for one client, eight days passed between the first commit and regular operation.

85% lack the know-how

We have the IT team

That's how many companies without AI lack the technical know-how, according to Bitkom. Our IT team connects the AI to your programs, builds the agents and runs them. You don't have to program anything.

Since 2019 IT from Hamburg

Servers and hardware come from us

Progressive Mind from our group of companies takes care of servers, devices and network and maintains everything. If hardware is missing, we supply it, but only if it's really needed.

66% held back by data protection

We keep an eye on the regulations

That's how many companies using AI name data protection requirements as a hurdle, according to Bitkom. Our data protection team checks every AI agent before launch. You get an AI policy with a data traffic light and an approval pack for data protection officers, IT and the works council.

And four things you'd otherwise spend a long time looking for

  • We start from revenueWe come from marketing. Before we write a single line, we ask what the software is supposed to sell.
  • One contact, one invoiceConsulting, software, IT and training from one group of companies. Four providers would mean four culprits.
  • A fixed quote before anything startsAccording to Bitkom, 63% of companies without AI say the costs are hard to calculate. With us, you pay nothing you haven't agreed to beforehand.
  • Not tied to any vendorOur focus is Claude. Depending on the task, we also use ChatGPT, Microsoft 365 Copilot or models that run in Europe.

Sources: Bitkom, press release of September 14, 2026 (in German) (opens in a new tab). McKinsey, The state of AI in 2026 (opens in a new tab).

No sales department

You talk to the people who build it

  • Florian Wessling, CEO at Collective Brain

    Florian Wessling

    CEO

    Advises companies on introducing AI and leads the workshops in Hamburg.

    Book a meeting
  • Arno Hoffrichter, CTO at Collective Brain

    Arno Hoffrichter

    CTO

    In charge of IT, automation and software development, from architecture to operations.

    Book a meeting (German page)
  • Learoy Eichholz, Lead AI Automation at Collective Brain

    Learoy Eichholz

    Lead AI Automation

    Builds agents for sales and service in Hamburg. 15 years of e-commerce experience in mid-sized companies.

Common questions about AI software

How much does AI software from you cost?

You start with the AI Hands-On Day at €2,490 net per group. Depending on scope, a first use case in production often costs a five-figure sum, and for ongoing operations there is the AI Partnership at a fixed monthly price. You always get a fixed quote before anything starts. On top of that come the usage costs of the AI model, which we estimate with you beforehand. All prices net of VAT, for business customers only.

Do we have to start with the workshop?

No. If you already know exactly which workflow is up next, a first call is enough. The workshop is the fastest route if it's still open where it pays off most. It also brings your team on board right away, because software nobody uses doesn't earn anything either.

How soon is the first application running?

The first use case is usually running after four to six weeks. Small routines are quicker, automations across several departments take several months. We'll tell you what's realistic for you after the workshop.

Where is our data stored?

We decide that for each type of data before we start. Websites and applications run in German data centers. For AI models, the data traffic light decides which data may go into which model. Where EU processing is required, we use regional endpoints or models that run in Europe. Honestly: if data must not go into a cloud model at all, we tell you so and leave that step with a person.

Do we need our own IT department?

No. Our IT team builds and runs the agents, and Progressive Mind from our group of companies takes care of servers, devices and network. From you, we need access to your programs and one fixed contact person.

Do you only work with Claude?

Claude is our focus. But we're not tied to it: depending on the task, we also work with ChatGPT, Microsoft 365 Copilot, automation platforms or models that run in Europe. The use case decides which tool you end up with.

Does the works council need to be involved?

Often, yes. If an AI application can record the behavior or performance of employees, co-determination under Section 87 of the German Works Constitution Act usually applies. That's why our approval pack is also written for the works council. It does not replace legal advice.

What happens if an AI agent makes a mistake?

Every agent follows the same process and the same checklist, and you decide what a person approves before it goes out. Mistakes still happen. That's why every agent gets a measurable goal before we build it, and we look at the results together every month.

What happens if we want to stop?

In the first three months of the AI Partnership, you can leave without giving a reason. The handout, the record and the list of your workflows from the workshop are yours in any case.

Is our company big enough for this?

It pays off from about five employees, and it really comes together for mid-sized companies with around twenty people or more. Then there are enough recurring tasks and enough enquiries for a lever to work on.

One day. Then you know what your lever brings in.

Book an AI workshop Let's talk first

More on this topic

AI workshopsAI ServicesAI automation (in German)Claude for teamsWeb Development