What we do
We work with research institutes and innovative companies to build research-grade systems that turn data into decisions and ideas into controlled experiments – not demos.
Our work spans AI agents that sit on top of existing dashboards and data sources, experiment and lab platforms that make complex studies runnable in practice, and the expert engineering that holds both together.
Most teams already have dashboards, metrics, and reports; we build the AI agents that act on them – running structured what-if scenarios, producing the artefacts a decision needs, and acting instead of just suggesting.
Readiness is most of the work
Agents only deliver when the organisation around them is ready: legible data, real APIs, and processes that aren’t trapped in people’s heads. Getting there is the hard part – and it’s the part we handle, not a prerequisite you have to finish before calling us.
We’ve written up how we get you there: readyforagents.ai.
AI agents for decisions and action
Many organisations already track everything: KPIs, funnels, survey results, financial data. What’s missing is an agent layer: a system that can query this data on demand, run structured “what-if” scenarios, and then act on the result instead of stopping at a suggestion.
We design and build AI agents that:
- connect to your existing tools and data sources via well-defined APIs,
- run controlled tool use that stays within guardrails,
- act and produce artefacts – briefings, reports, updated records – not just opaque answers, and
- keep a decision and reasoning log so every step can be audited or revisited later.
This makes it possible to move from “we have dashboards” to “we have AI that can decide and act on what to do next.”
Experiment & lab platforms
We build experiment and lab systems for social science, human-robot, and human-AI interaction research.
Examples include:
- Social robot interviewers – the complete system behind Willi, a social robot that interviews visitors at an innovation lab: on-robot AI skill, backend, tablet app, and management tools.
- Multi-module online experiments – where one system coordinates shopping assistants, chat-based interviewers, and survey components into a single study flow.
- AI coding of open-ended responses – dual independent AI raters, automatic agreement metrics (Cohen’s kappa, Krippendorff’s alpha), and reconciliation that learns from every run. We’ve turned this into a product: qualcode.ai.
- Behavioral data capture on the live web – instrumenting real browsing as structured data, with domain enforcement and a built-in experiment pane, without touching a participant’s own browser. This is the basis of labrowser.app.
- Custom lab tooling – including precise sensor synchronization, logging infrastructure, and control software that ties together robots, screens, sensors, and backends.
Behind each visible system is a set of enabling technologies: tool APIs, sensor synchronization, logging infrastructure, and experiment logic that make studies repeatable, adaptable, and analysable instead of one-off prototypes.
Expert engineering as a foundation
Under everything we build is expert engineering:
- Correctness and reliability – tested components, careful treatment of edge cases, and clear contracts between systems.
- Performance where it matters – from low-level Zig for performance-critical components to efficient Python backends and streaming interfaces.
- Reproducibility – versioned infrastructure, experiment configurations, and data pipelines so results can be revisited and extended.
- Long-term maintainability – code and systems that can be handed over to in-house teams, no black boxes.
That mix of research awareness and engineering discipline lets us take on projects that sit between a research prototype and a production system – exactly where many ambitious AI and HRI experiments live.
Working with the lab
We typically collaborate with:
- research groups running complex online or lab experiments,
- innovation labs exploring new forms of human-AI interaction, and
- organisations that already have serious data and dashboards, but need an AI layer that helps them decide and experiment, not just chat.
If you’re planning a project that needs AI agents, experiment platforms, or custom lab infrastructure, we’re happy to explore whether it’s a good fit for the lab.