Consulting
We help you decide where software and AI belong in the business: what is worth building, what can wait, and how to move in steps you can reverse.
Typical first stepA short review of where you are: risks, options and a plan.
A Nordic software company for AI-driven digital transformation.
We help you decide what is worth building, build it end to end, and stay with it after launch.
Celebrating 30 years of our founder's award-winning digital work: hundreds of campaigns, websites, apps and core systems.
What we do
AI adoption
We help make AI part of the work, not a loose experiment: processes redesigned around it, usage tied to the right context and tools, and people and agents working together.
Typical first stepAn open discussion session or mapping of where AI fits your processes and tools.
Development
We build the solution end to end: the systems, integrations and automation that stay in daily use after the project ends. The technology is a means, not the message.
Typical first stepOne bounded first delivery with clear acceptance criteria.
Hosting
Running it is part of the work. We operate and maintain the environments so the software stays up after launch, on servers in Finland or in the cloud that fits.
Typical first stepIncluded with a delivery, or agreed on its own for an existing system.
Situations we are called into
AI writes code fast. Wrong architecture decisions cost more than the speed saves.
A prototype that works as a demo
You built a prototype quickly, often with an AI or no-code builder. It works as a demo, but a lot remains before enterprise use: sign-in and roles, security, data protection, backups, monitoring and maintainability. Many find that gap hard to close, and doing it badly creates new problems.
The right thing is to decide what to keep, then rebuild the architecture so it is secure, scalable and maintainable.
A SaaS that no longer fits
You pay monthly for a product that does not bend to your process. You work around it with spreadsheets and manual steps, the price grows with every user, and your data stays in someone else's system.
AI has made custom software much cheaper to build, so the break-even has moved. The right thing is to count both in euros, side by side: a custom application plus its upkeep against the licence plus its limits, including what you would have to replace, such as security and support.
A website that stays yours
AI makes building fast, but not everything belongs in an AI company's site builder. Those tools can tie you to a platform, limit changes and keep charging.
For example, a plain HTML, CSS and JavaScript site on S3 and CloudFront fits many cases: it costs very little to run, with no subscription or lock-in, and is edited as files. It is fast by design, has little to attack (no database, plugins or admin login) and can still be built the AI-driven way.
The right thing differs: architecture in one case, an economic decision in another, a platform choice in the third. We start by working out which kind of problem you have.
How we work
Some clients need one problem solved, others a technical partner who stays. Both are welcome.
A defined project
A clear scope with a clear end. We work alongside your development team as technical lead or builder, or deliver a bounded complete solution if you have no team.
A continuing partnership
Ongoing development, technical leadership and upkeep, agreed as a fixed monthly arrangement or as hours used. Terms stay flexible: early-stage companies can start small with no long lock-in, and we can take a role in your team for a period, for example as technical lead.
How it goes
-
Scope
We define the problem and what done means.
-
Proposal
A short written plan with options and a clear next step.
-
Build
Delivered in steps you can review and reverse.
-
Run
Handover, documentation, and operation if you want us to run it.
What you pay for
You pay for the work, not for the layers around it. You work directly with the people who do it. Small projects are welcome.
- A project carries no obligation to continue with us after it ends.
- Handed-over work comes with enough documentation to carry on.
- A partnership is either a fixed monthly arrangement or hours used. You choose.
How we use AI agents
Anyone can build an agent. Making it reliable, and making it work with people and other agents, takes understanding and planning: clear roles, written rules and review.
This is not a future promise. Here is one example of how we work today, for everyday tasks.*
-
Start
You talk or write to the AI tool you already use, or a scheduled agent run begins on its own. The task and its comments go to your task board, assigned to an agent.
-
Context
Before acting, the agent reads your written rules, documents and history, so it works with your facts and your way of working, not general guesses.
-
Work
The agent does the work itself: it reads, analyses, writes and updates your systems, with only the access the task needs, and hands parts to other agents.
-
Check
A separate agent reviews the result against your written rules before it reaches you. The maker never checks its own work.
-
You decide
Anything irreversible or outward-facing, such as sending, paying or publishing, waits for your approval. You can stop any step, and every step is logged on the task board.**
* Demanding software development runs in a proper development environment, where a person directs the agents more closely.
** The people who direct and review the agents are trained to use AI and to recognise its limits, such as made-up answers and over-trusting the output.
Want something like this in your business? Ask us about the options.
Selected work
An admin tool for access management, built for a customer who found nothing ready-made.
15 daysfrom first technical question to full production
10–20×faster and cheaper than a traditional build, based on our previous projects
Only tens of €a month in cloud running costs
- 26 Oct · day 0First technical question
- 28 Oct · day 2We start coding
- 6 Nov · day 11First live use
- 10 Nov · day 15Full production
From first question to production in 15 days
The customer needed access control for their own customers, who use it on the web and with their phones: for the existing system and for new products. Locking brings its own security requirements, nothing ready-made fit, and the systems in use were hard and expensive to develop further.
They built a clickable mockup with AI. We made the architecture decisions, took the mockup as material for design and coding, and built the core: all the code was written by AI agents under our direction. From their first technical question to full production took just 15 days. Based on our previous projects, building a service of this scope the traditional way takes ten to twenty times the time and the money.
Their own developer now carries the development forward with AI tools. We operate and maintain the environments in their own cloud accounts and answer questions as they come up.
The customer is very pleased with how well and how versatilely the new system works.
- A web admin interface for several organisations in one service, with superadmins, organisation admins and managers.
- Sign-in by invitation with passkeys, and people, groups and doors with schedules and time-limited access.
- APIs in every direction for integrations, and the admin itself can be configured by AI tools without the UI: both designed from the start for AI tooling.
- Open and change logs with export functionality and the retention you choose, system and error logs for superadmins, and time-limited access codes that are shown once and never stored.
- Enterprise basics covered: designed for GDPR from the start (data in the EU, minimal data stored, a data processing agreement accepted in the service) and a security review with every finding fixed before launch.
- Serverless with versioned releases, automated database backups and release checks, and running costs in practice only a few tens of euros a month for the cloud services.
Clients and partners across sectors
Contact
We are happy to tell you more. Just reach out.
Tell us the problem, not the solution. A few lines is enough. If someone recommended us, say who.
We read it and reply with a short proposal for the next step. If we are not the right fit, we will say so.
Toni Saikkonen
CEO, Norso Oy
Helsinki, Finland
LinkedIn