Axis Ordo is the platform on which Soft Dev Oy builds software products from AI agents.
The platform was built for our own development: every next product has to be assembled from parts that already exist rather than written again. Axis Ordo holds the agents, the rules for how they work together, and the points where a human makes the decision.
02What a digital agent is
A digital agent is a program that carries out a work task in full: it takes the assignment, does the work by the given rules and answers with a finished result, not with an answer to a question.
That is where the difference from a chatbot comes from: a bot holds a conversation, an agent closes the task and leaves behind a result that can be checked.
| Factor | Chatbot | Agent |
|---|---|---|
| What it does | Answers a question | Does the work |
| Result | Text in a chat window | A document, a record, an action |
| How it starts | Waits for a person | A schedule or an event |
| Memory | The conversation context | Task state and decision history |
| On error | Tries another answer | Stops and calls a human |
| Volume | One conversation | Dozens of tasks in parallel |
03How a product is built from agents
A product is a construction of agents, not a single program.
- Data collection
- Checking and reconciliation
- Document preparation
- Publishing
- Reporting
- The order of steps
- Rules and limits
- Human checkpoints
- Handling failures
- The operator interface
- Access rights and roles
- The work log
- Support
- The platform holds the agents and their versions: one agent works in several products at once.
- The construction sets the order and the limits — what the agent does itself and what goes to a human.
- The product adds the interface, the access rights and the log: the operator sees what was done and when.
- The core handles startup, the task queue and recovery after a failure.
04Why a release every two to three weeks
Each next product costs less than the last.
The first products were released on 17.08.2026, on the same day. They share the same agents: data collection, checking and publishing.
A new product does not start from nothing — an existing set of agents is taken, and a construction and an interface are written on top of it. That is why the time is counted in weeks rather than quarters, and the next product follows two or three weeks after the last.
05How the core is built
What the platform consists of.
- The agent registry: what each agent can do, what its inputs and outputs are, and which version is running now.
- The task runner: the queue, parallel execution, retries after a failure and a log of every step.
- Rules and limits: where the agent decides on its own, where a human is required and what happens on failure.
- Work accounting: how many tasks were completed, where the stops were and what the work cost.