EXPERTISE / ROBOT SOFTWARE

AI-Assisted Software Development

Let AI coding agents take on useful robot-software work - package porting, test generation, log triage and documentation - while your engineers stay responsible for the requirements, tests and release decisions.

Explore the approach ↓

AI-generated documentary image selected for the page hero; OEM engineer reviews a robot-software change beside a test setup.

01 / THE OPPORTUNITY

Make AI useful on the work around the robot

Coding agents can speed up well-defined engineering work. They need more than a source repository to do that safely on a robot:

Engineers need clear ownership. Someone must decide what the software should do, which changes are acceptable and what evidence is needed before release.
The best tests are often tied to hardware. Simulation, recorded data and hardware-in-the-loop tests need to be easy to run and trusted by the team.
A successful build is not enough. The code still needs to dock the AMR, avoid the table or handle the sensor dropout it was designed for.

That is an opportunity to improve the engineering process as well as introduce new tooling.

02 / HOW WE SOLVE IT

From specification to independently checked result

We start with a small workflow that matters to your team, then make it repeatable:

Start with the user specification. Your team sets the intended robot behaviour, acceptance criteria and constraints before work begins.
Let agents do the implementation work. Coding agents produce a proposed change and its tests in a repeatable development environment.
Run an immutable test harness. Simulation, recorded-data replay or hardware-in-the-loop scenarios check the change. The harness is versioned and protected: the agent cannot adjust the test to make its own work pass.
Produce a clear report. The workflow records what was requested, what changed, which scenarios ran and the evidence for each result.
Validate independently. A separate agentic run checks the work against the original specification and test evidence, rather than accepting the implementation agent's own conclusion.

Your engineers remain responsible for the specification and release decision. We coach them through the first workflows and make the evidence useful in review.

03 / IN PRACTICE

In practice: bringing coding agents into an OEM workflow

Situation. A European robot OEM wanted to use coding agents without weakening the review and test discipline needed for a complex product.

What we did. We helped structure controlled access to development tools and data, automated the build-and-test workflow, and connected simulation scenarios to the review process. The team could use agents to produce and improve modules, test them against protected scenarios, and check the evidence against the intended behaviour.

Outcome. A practical route to introduce agents without creating a second, unreviewable way of changing robot software.

Also: in our own customer projects we routinely let agents port packages, generate test suites and triage logs under exactly these rules; recommending test datasets and harnesses before scaling agent contributions is now part of every architecture review we do.

04 / WHAT YOU GET