# majortom.ai — AI that works inside your systems > Keep the signal steady. We build mid-sized companies the AI that works inside their own systems. In a view of their own, they see every step and approve what goes out. Ground Control Berlin · reading you loud and clear ## Keep the signal steady. We connect AI to the software your company already uses. We start with one workflow that pays off. Then comes the next. Your people and the AI see the same processes, the same numbers and the same status. You see every step and approve what goes out. 30 minutes for your workflow · See an example The same 14 orders · four desks ## One case, four perspectives. Fourteen orders are stuck. Each team needs a different answer. Everyone asks the same place and sees only what belongs to their work. ### Four perspectives Management — Monday, just before nine “Why are 14 orders stuck? Does this happen often?” - ✓ Eleven are waiting on the same supplier. It was late four times in three months. - ? Three addresses are incomplete. You decide who follows up. - ⏸ Visible only to you. Data from your inventory system, summarised by AI. ### Four perspectives Sales — Monday, half past nine “Which of my customers are waiting? I want to call before they do.” - ✓ Five customers, sorted by revenue. Phone numbers and delivery dates are ready. - ? One customer has an open quote. You decide whether it matters. - ⏸ New dates only after your approval. You make the call. The AI promises nothing. ### Four perspectives Customer service — Monday, quarter past ten “Who has followed up? Draft replies with the current status.” - ✓ Nine e-mails and nine drafts. Six customers are asking for the second time. - ? One customer wants to cancel. That e-mail gets no draft. - ⏸ Nothing is sent without your approval. With your templates, in your tone. ### Four perspectives Accounting — Tuesday, after lunch “Which orders are paid? What happens to the due dates?” - ✓ Eight are prepaid, six are on invoice. - ? The supplier contract includes a late-delivery penalty. You decide. - ⏸ Nothing is posted or chased without approval. Missing information stays visible. What we actually build ## Your people already use AI. Just each for themselves. Three out of four companies that use AI use ChatGPT: in sales, in accounting or on the boss's phone at night. But every window works alone and knows nothing about your orders. That is where AI breaks down in mid-sized companies, not at the models. ### Separate: Every window is an island What sales found out yesterday stays in their chat history. Customer service asks again today and may get a different answer. ### Lost: The good approach goes on holiday too A colleague finds an approach that works. It belongs to her, not the company. When she leaves, it leaves with her. ### Not connected: In the end, copy and paste again The chat does not know your orders, inbox or price list. Someone must transfer the answer into the system by hand. What we make of it ### The sources stay the same. Your team works as usual. We build the bridges between them. The AI provider can be replaced. If you wish, the AI runs in the EU. - **One place, different models.** The interface knows inbox, inventory, shop and accounting. The language model working in the background can be changed. - **Your workflows stay inside the company.** What one colleague finds out becomes a workflow for everyone. The knowledge stays, even when she leaves or the provider changes. - **Every action is individually permitted and visible.** You decide who may read orders, write drafts or book invoices. A log shows what was read and sent. - **Data routes are clear and documented.** For sensitive data, we can run models ourselves in the EU. A table shows which data go where. Trust ## You don't have to trust us. You have to be able to watch. Trust from a brochure lasts until the first mistake. So every step is recorded. You can cut every line yourself. ### Control — “And if it messes up?” You see the mistake first. No e-mail, invoice or customer reply goes out without approval. The AI prepares it. You approve it. ### Data — “Where do our data end up?” Your programs remain the source. A table shows you in advance which data go to whom and why. The provider is contractually barred from training on them. Where that is not enough, we use a self-run model in the EU. ### Understanding — “I don't even understand what's going on in there.” You only need to be able to read it. Every run records the task, what it accessed, what it produced and where it waited, all in plain language. A record for you, not a log file for technicians. ### Way back — “And if we want to get rid of it again?” You switch it off, and your company works as before. Your data stay yours and your workflows run without us. The way back is part of the delivery. ### “Am I even allowed to let this into my company?” The question is right and not technical. Hold us to this: 1. **We automate the repetition, not the responsibility.** The AI types the fourteenth identical apology. The decision stays with a person. 2. **Your team hears it from you, not from the software.** Nothing changes quietly. You announce what is new. 3. **We report no success we haven't checked.** The record shows whether a workflow holds up. We do not publish unchecked numbers. And why you should believe us ## Because we built it on ourselves first. One of us runs two trading companies. AI handles product data, prices, translations, marketplaces, invoices and customer service. The workflows run at night. In the morning, a person says yes or no. - **42** active workflows keeping two trading companies running - **14** hours a week in which one human steers both companies - **10** languages served from a single source - **2** trading companies that run this way. Both belong to one of us. How it goes ## Four stages. After the first, you can get off at any time. ### 01 We come by We spend two or three days on site with your team. After two weeks, you receive the plan: what you use, what gets in the way and where AI pays off first. The plan is yours. audit · approx. 2 weeks · fixed price ### 02 We settle what it may do We decide together: What may the AI do? Who may do what? What stays manual? We write down what you will measure us by: the starting value and the target. Only then do we build. 1–2 weeks ### 03 We build the workflow and connect it to your systems The first workflow goes live early. You use the AI during the project. Every workflow runs as a trial first. The old way stays until you say the new one has proven itself. 4–8 weeks · fixed scope ### 04 You take the controls We train your team, document everything and hand over the keys. If you wish, we support changes and further workflows. handover · support optional Who you'll be talking to ## Two founders. The same two who build for you afterwards. ### Malte Buttjer Over twenty years in technology and entrepreneurship. Malte co-founded Tocsen and sold it to England in 2023. Tocsen crash sensors sit in hundreds of thousands of helmets. Today he runs two trading companies with AI workflows and makes sure technology pays off. [buttjer.net](https://buttjer.net) ### David Friedrich Ten years of web products and visual communication. David builds open-source tools that check AI results. He brings the same principle to your workflows: easy to use and reliable in operation. [durchnull.de](https://durchnull.de) - **Strategy:** We start where the effort pays off first. If it does not, we tell you. - **Technology:** We hook the AI up to your systems: inbox, inventory, shop, documents, accounting. - **Interaction:** We build one interface everyone in the company understands. Questions you will ask ## The awkward ones too. ### Roughly what does it cost? The audit has a fixed price. After that, we work to a fixed scope rather than open-ended hourly billing. The price depends on the number of systems. So we look first, then name the figure. ### Do we have to buy new software for this? No. We connect the programs you already run. Some of them charge for their interface: the access other programs use to fetch data. We tell you beforehand, and you decide. ### Will my people lose their jobs over this? We automate what nobody likes doing: retyping, looking things up, the same e-mail for the fortieth time. Decisions stay with your people. The audit plan shows what changes before you sign anything. ### Do we need our own IT people for this? No. We train your team and document everything in plain language. Later changes can be made by us or your IT. There is no secret knowledge. ### Our chamber of commerce or chamber of crafts advises on AI for free. Use it. It is good for orientation and funding. We build the connection to your systems. Before you call us ## When we turn you down. - Your work happens in a single program? Then you don't need us. We work between systems. - You are looking for a chatbot for your website? We don't do that. We build for your team, not website visitors. - Something is supposed to go out without human approval? We won't build it. The approval is not the brake, it's the point. - Your information lives on paper and in people's heads? Then you are starting at the wrong end. We need a system with an interface. - It has to be running in two weeks? Then we decline. The audit alone takes two weeks. Ground Control, come in ## Tell us about one workflow that annoys you every week. 30 minutes, no slide deck, no pressure. We sketch the workflow. If it is not worth it yet, we tell you in the call, not in the quote. Send a transmission. Four equal ways to the same message: - With ChatGPT, Gemini or Claude. The three buttons open the assistant with a ready-made prompt: it asks which workflow annoys you every week, which programs it runs in and how much time it costs, and writes a short email from that, ready to send. - Write the email yourself: groundcontrol@majortom.ai --- Ground Control Berlin · channel open. AI that works inside your own systems. Built in Berlin. Contact: groundcontrol@majortom.ai · Malte Buttjer (https://buttjer.net) · David Friedrich (https://durchnull.de) This page in German: https://majortom.ai/