Frequently asked questions

We build software. Web services, mobile apps, and business systems, from design through operation, and on top of them, AI agents. By AI agent we do not mean a chatbot; we mean software that takes on a role in your organization, such as writing quotes, triaging support tickets, or processing expenses.

We also help set direction without building. AX/DX transformation assesses which processes to automate first and produces a roadmap, and technology consulting advises on technology choices and team structure.

The four services and their deliverables are in the Services section, and the things we have built are in our Work.

Yes, and we would rather start small. Building a small working result before a big plan, using it for real, and then deciding direction is how we work.

We take on work that finishes in a few weeks: a prototype to test an idea, an agent that handles a single task, or a small feature added to an existing system. Even a half-formed idea is fine. Tell us on the contact page how you work today and what gets in the way.

Usually in four stages.

1. Listening (1–2 weeks): we see how you work today and where time leaks, on site. We look at your existing data and systems too. This stage ends with a proposal stating what we will build, how long it will take, and what it will cost.
2. Building small: we get one core flow working first and run it on real data. For an agent, that means running it first on real cases in an environment kept separate from operations.
3. Widening: the proven flow is extended to the remaining processes and teams, with approval screens, dashboards, and integrations.
4. Handing over: operations handover, staff training, and a guide for adjusting rules. We can keep running it for you if you prefer.

We share progress, cost, and risk openly, and raise the hard conversations first.

When the first stage, listening, is done, you get a proposal stating what we will build, how long it will take, and what it will cost. We do not give numbers before that, because an estimate made without seeing how you work is hard to keep.

As a rough sense: an agent for a single task or a prototype takes a few weeks to two or three months; a system or app used by several teams takes three to six months. Cost is based on duration and team size and is billed by stage. You can stop after the first stage.

Budget range and preferred timeline on the contact form are optional. You do not need to have decided them in advance. Start a project

Delivery is not the end. Keeping deployed software running reliably and improving it is part of our job.

You can choose one of two ways to operate.

• Run it yourself: at handover you get staff training, a monitoring screen, and a guide for changing rules and settings. Agent classification criteria and policy tables are built so your own staff can edit them.
• Have us run it: on a monthly contract we handle monitoring, incident response, and small improvements. For agents, we also review criteria and rules with you on a regular schedule after launch.

Either way, the code, data, and accounts belong to you.

The principle is that data stays in your accounts and systems. Agents work with access permissions to your existing systems, such as the ERP, email, or help desk, and copy no data to our side. Any servers or accounts needed are created in your name, and our access is withdrawn when the project ends.

Where AI models are used, we document at the design stage which data goes to which model, and choose between commercial models under terms that exclude training on your data and models running inside your own infrastructure. For data containing personal information, masking and minimization rules are agreed first.

What an agent reads and produces is logged, and your staff can review it at any time. If you have security policies or audit requirements, we review them together in the first stage.

Not the way we build them. Agents draft and process, but anything that spends money or leaves the company is approved by a person: sending a quote, paying an expense, placing an order, replying to a customer. The seat of responsibility is never left empty.

When confidence is low, the agent does not force a decision; it shows candidates or flags the item “needs review” and hands it to a person. Before rollout, agents run on real cases in an environment kept separate from operations, and enter the workflow only after accuracy is confirmed.

The reasoning behind each judgment is shown on screen, which is why a backup approver can approve as is when the owner is away.