AI at Orveon
Born in the shift to AI, and built to keep up with it.
Orveon was founded as AI changed how software is made. We follow its developments closely, adopt what proves itself, and use AI in everything we build, starting with our own work. It lets us release a first working version early and improve it in short cycles.
01Our view
AI is changing how software is made, and what it can do.
Business software has usually been slow to change and built once for everyone. With AI, it can be built faster, shaped around one business, and take on work that only people could do before.
Many organisations have tried AI. Getting it to work in production, inside the systems they run, with people still in charge, is harder. That is where we work.
02How we use it
We run our own work on AI first.
What we offer clients, we use ourselves every day. It keeps our own processes lean and our delivery cycles short.
Agents in our own processes
Agentic software takes on routine work across our own business processes, and our people approve what matters. Routine work does not wait in a queue.
Development with AI
We design, write, test and review software with AI at every step. Our engineers stay accountable for every change, and every change is reviewed before it ships.
Efficient use of AI
We match each task to a suitable model and keep usage in proportion to the work, so usage costs can be set out before the build.
03Agents
We find where an agent belongs, then build it.
An agent is useful only in the right place in a process, with the right limits. Placing it well is most of the work.
- 01
Find the right place
We map the process with your team and find where an agent saves real time, and where a person must stay in charge.
- 02
Set its limits
We agree which data and systems the agent may use, what it may do on its own, and what needs a person's approval.
- 03
Build, test and release
AI Studio builds the agent, tests it against the agreed quality targets and releases the smallest version that works, then grows it.
04Keeping up
We follow developments closely, and adopt what proves itself.
New models, tools and methods arrive every week. We follow them closely and try what is promising on our own work first.
What holds up there reaches our clients' projects quickly. What does not, we leave out, so clients get what is new only when it is also reliable.
05On our agenda
What we are working on now.
Agentic workflows
Agents that take on a whole task within agreed limits, and hand every decision that matters to a person.
AI-assisted engineering
Building software with AI from the first draft to the final review, without lowering the bar for quality.
Choosing the right model
Testing models against the actual job, and matching each task to the model that does it best for its cost.
Interfaces people make themselves
Our founders' belief that fixed interfaces will give way to ones people shape around their work.
AI in the systems you already run
Connecting models and agents to existing data and workflows, so AI works where the business works.
People in charge
Approval, review and clear limits, so the people responsible for a decision keep it.
06Questions
Questions about AI and agents.
What is an AI agent?
Software that uses an AI model to carry out a defined task over several steps, with the tools, data and systems it is allowed to reach. In the agents we build, the decisions that matter go to a person. See the glossary for related terms.
Will an agent make decisions without us?
Only within limits you agree. We set with you what an agent may do on its own and what needs a person's approval, and the decisions that matter stay with your people.
How does Orveon use AI in its own work?
Agentic software takes on routine work across our own processes, and we design, write, test and review software with AI at every step. Our engineers stay accountable for every change.
Where should an agent sit in our process?
We map the process with your team to find where an agent saves real time and where a person must stay in charge. Placing it well is most of the work.
How do you keep up with new models and tools?
We follow developments closely, try what is promising on our own work first, and bring to client projects what holds up there.
Bring us a process you want AI to take on.
We will show you where an agent belongs, what it would take, and how we would release a first working version.