Which part of your week repeats, and which part needs you?

Dinobridge is a small company, and its owner, Moha Aghanoori, leads every engagement personally: the person you speak to in the first call is the person who sits at your table, sets up the tools and does the teaching. Nothing is handed to a junior. Behind that: years of professional engineering, building AI systems in production that other people depend on.

Yes, and most of the people we have worked with came this way, none of them running a company at the time. You bring a question — a decision, an idea, a company to start, something to learn, a child to teach — and you get a session or a few with someone who does this for a living, in plain language, at a fixed price agreed in writing before anything starts. The first conversation is free either way, and if AI does not help with your question, you hear that in it. The page for individuals sets out how it runs.

Fairly asked, and the answer is on the table before you ask it. Our owner holds a full-time engineering role building production AI, and that role is where most of the production experience behind this site comes from — it is the reason we can tell you what survives an audit rather than guessing. It also means our capacity is real and limited: we take a small number of clients at a time, and the shape of an engagement — which days, on site or on video, over how many weeks — is agreed with you in writing after the first call and before there is a contract. We will never tell you a day is available and then move it. If what you need is more than we can carry, you hear that before you spend money, not after.

Yes, and for regulated work it is the first thing we design, not the last. Everything we build runs in your own accounts, under your own logins, with your own keys — we are not a hosting provider and we hold no credential you cannot revoke in an afternoon. Before we see anything: a mutual NDA, a data processing agreement under Art. 28 DSGVO, and a one-page scope naming exactly which systems we get access to and for how long. For law firms there is a separate confidentiality undertaking in Textform under §43e BRAO, with express instruction on the criminal consequences under §203 StGB. No training on your content, names removed wherever data has to leave your systems, and a complete record of who did what. Where a workflow cannot meet those conditions, it does not get built. The security page sets all of this out in full.

We talk for thirty minutes, free, with no preparation on your side. You describe your business; we tell you honestly where AI could change something for you, where it would not, and what it would involve. If it makes sense, you receive a written proposal with a fixed price. If it does not, you leave with a clear recommendation instead — which is roughly what happens one time in three.

Yes, that is the usual shape. After the consulting work we take the role on a retainer: direction, experiments, tool decisions in writing, your AI use policy kept current, leadership briefings, and a named window every week your team can rely on. Monthly, and you can stop with one month’s notice any month. Nobody on your side needs technical knowledge: you explain the business in your own words and everything technical stays our problem.

IT keeps the systems running. We ask a different question: what could this business do with AI that it cannot do today, and which of those ideas survives a real test? Then we run the tests with your people. Most IT teams are glad someone finally does that. We work with them, not around them — and if your IT team wants to own it, better still: we spend the first weeks with them, they keep the systems, and you stop paying us sooner.

Good — that tells us your people are willing. It also usually means company material is sitting in a tool nobody approved, with no record of what went in. A licence is not a capability. Copilot is good at the sentence in front of you; it cannot see that your contracts arrive by email and get lost, that seven quotes need normalising before they can be compared, or that every deadline in your business lives inside a document. Those are not writing problems. We work on the shape of the work, and on the written rules for what may and may not be pasted into anything.

You could, and if you can already specify the role you probably should. Most businesses at this stage cannot — and an engineer without a brief builds what they find interesting. Take the consulting work first. If the answer is “hire someone”, we will say so, help you write the job description, and be glad to lose the retainer.

No, and that is the point of how we build. The advice stays with the partner, the valuation with the agent, the direction with the founder, the priorities with the head of department. The tool takes over the repetitive work and prepares the material. Your people decide. We have refused to build the “accept all” button more than once.

Usually, yes. A manufacturer, a logistics group, a private clinic. In every field so far the engagement has run the same way: we learn the business, we map where AI opens something new and where it must stop, we run the first experiments. What we build changes completely from field to field. How we find out has not had to — and if your field turns out to be the exception, you hear that early rather than at the end.

Some will, and usually the ones who are right to. Resistance is almost always one of two things: fear about the job, or a previous tool that made the work worse. So we build only where the person doing the work agrees the work is repetitive, we never build anything that reports on individuals, and we refuse to build the button that takes the decision away. The team-leaders example on this site is a system that organises what a manager observed and explicitly never rates anyone — we were asked for the version that watched people, twice, and said no both times.

We agree the numbers before we start. At the beginning we write down two or three measurements with today’s values — hours on a named task, days from A to B, the count of the thing that keeps going wrong. At the end we measure the same things again on whatever we built. If they have not moved, that goes in the write-up too. And the honest arithmetic is simple: count the task for one week, multiply by what an hour of that person’s time is worth to the business, then halve it. If that number is small, we will tell you this is not worth buying.

No, and it is designed out rather than promised. Everything runs in your accounts, under your logins, with your API keys billed to you at cost. The code sits in your repository. Each system has a written handover and a named owner on your team who has already changed a rule in it. Nothing needs a licence, account, server or subscription of ours. If we disappeared tomorrow it would all keep running — put that in the contract and we will sign it.

The retainer ends with one month’s notice, any month, without a reason. Everything we built keeps running: your accounts, your logins, your repository, your keys. You get the handover for each system and, on request, a full export of the logs and written confirmation that we hold nothing of yours. The consulting work has a natural stopping point built into it — at the end you decide whether anything continues, and you keep the plan and the working systems either way.

Then you hear it in the first conversation, before you have committed to anything. It happens often: what looked like an AI problem turns out to be a process nobody owns, or a tool your team already has and does not use. We say so, and we say it early. If we only work it out once we are inside the business, we say it then, we stop, and we bill only for the days already spent.

No. The first conversation is a call either way, and the Head of AI role runs partly over video once we know your business. Clients so far include Leipzig and Brandenburg. Berlin simply makes the on-site part easier to do in person.

Most of it is in the consulting work: time in the building with us, plus an afternoon of training for the people who will own what we build. After that, the Head of AI role costs your leadership a short meeting a week and your team the occasional hour to try something new. What we put in place takes seconds to use: ninety seconds of dictation after a site visit, one photograph of an invoice.