for research groups, spin-offs and citizen science
Ethic builds the product around your model, so colleagues, growers or participants can work with it, connected to your institution’s systems. Your method stays yours, and we start small: one pilot of 4 to 6 weeks.
No obligation · 30 minutes · online, with someone who puts AI to work every day
Ethic is the AI studio of Zenjoy from Leuven. At Phonebox, researchers from KU Leuven built the AI model; Zenjoy built the product around it, so growers use it in the field.
At a glance
Route 1 · Your model becomes a product
Your model stays yours. We build what it needs around it.
01
“The model runs. Only we know how.”
Around your model we build an environment with roles, uploads and a screen to review and correct results. Your model itself stays as it is.
02
“Soon growers, participants or partners will have to work with it themselves.”
Then your model needs a product: accounts, a screen that also works in the field, and hosting that keeps running when the project ends.
What Zenjoy built around researchers’ AI
Phonebox · KU Leuven, MeBioS
The researchers built an AI model that recognises insects on sticky traps. Zenjoy built the product around it: web application, cloud and UX, so the model became usable in the field.
Maarallee · Scivil
Zenjoy built the citizen science app Scivil uses to collect Flemish speech, so speech recognition learns to understand Flemish better. The speech model itself is not Zenjoy’s.
Honest advice
Your research question, the analysis, the conclusions and what you publish. We don’t build AI for those.
Route 2 · Less research admin
Field data from uploads and emails, set out ready for review. A publication list that updates itself, a first draft of the report for your funder, or a knowledge assistant for your protocols and your DMP. These are smaller workflows, separate from your model.
Workflow
Documents and invoicesIntegration
Research systemsWorkflow
ReportingWorkflow
Knowledge assistantSystems you already have
You don’t need to replace anything. We connect to what’s already there, with no more rights than needed, and we work with your IT department.
Research
Publication databases, repositories and your research group’s databases.
Office and collaboration
Email, Teams and SharePoint, where your project documents often already are.
Does your institution already offer a tool, a repository or computing capacity? Use that first. An existing research package is often enough. Custom work only pays off with your own model, unusual data streams, fieldwork or external users.
Who builds it
Ethic is the AI studio of Zenjoy, the digital product studio from Leuven. Since 2011, Zenjoy has designed, built and maintained software, platforms and integrations for more than 400 organisations. You can see it in how we build AI: with tests, documentation and maintenance, so it still works three years from now.
We have used AI in our own work for years. With Ethic, we now put that experience to work where it matters, including in research.
15 years
building and maintaining software
400+
organisations
4 years
using AI daily in our own work
No middleman: the people who design and build your workflow sit at the table themselves.
Code, prompts, documentation and access are in your institution’s name, not ours.
If a system or a model changes, the same team adapts the workflow.
01
If a system we connect to changes, we adapt the integration as part of maintenance. Zenjoy has maintained platforms and integrations for years after launch.
02
If a smaller, more efficient or European model comes along, you can switch without rebuilding the workflow. We do test it again first.
03
Code, prompts, documentation and access. If the researcher who led the project leaves, the knowledge stays.
04
Updates, monitoring and small changes are covered by a monthly arrangement. New workflows are planned separately.
Rules that matter in research
This is not legal advice. These are the questions we look at in every research project, together with your DPO or legal department.
Research with personal data needs appropriate safeguards, with as little data as the research allows. Pseudonymisation is one of them: identity and measurement data stay separate, and the researcher sees data, not people.
An AI system you develop and use solely for scientific research falls outside the AI Act. If growers, participants or partners work with it, that exemption may no longer apply.
Anyone who works with AI needs to know what it can do and where the risks are. A short training for your research group is often the simplest start.
Your institution decides what the ethics committee and your data management plan require. We supply the technical description you need for it: which data, which model, where it runs, who checks.
Our ground rules in research
Your research team stays responsible for the method and the scientific validation. We take on the design and the software.
Every result comes with a source, the model version and who reviewed or corrected it. So you can explain it in your methods section.
AI may handle reversible internal tasks, such as labelling fieldwork data. A colleague first reviews anything shared outside the organisation and any decisions about people.
Where you can start
A 1-hour Pitstop or a 2-to-3-week AI Scan
You have an idea or a prototype and want to know what is feasible. You get an honest answer, even if that’s “not yet”.
You get concrete advice or an action plan.
4 to 6 weeks
One model, one data stream or one task, in your own environment. We measure the time per task and how often your team has to make corrections. Then you decide: continue, adjust or stop.
You get the measurements and our advice.
Half a day, on site or online
For your research group: what AI can do, where the risks are and what to agree on about data and sources.
A first step towards AI literacy in your research group.
After a first conversation, you get a proposal with a clear scope and a fixed price.
Frequently asked questions
Question not listed? Just ask us.
You can get a data processing agreement and a security overview on request.
If you work with a grant or a project deadline, we plan around it. A pilot around one workflow runs for 4 to 6 weeks, so in the introductory call we put your planning next to ours. You keep an eye on what the grant conditions allow. We supply the documentation you need for your reporting.
The model only sees the data we agree on in advance. As standard, we choose hosting in the EU and name the provider and country in the AI fact sheet. We choose models and services that don’t use your data to train their models. If it can run on your institution’s infrastructure, we look at that first, together with your IT department.
Yes, AI can make mistakes. That’s why we build in a check: results the model is unsure about go to a researcher who reviews or corrects them. We record the source, the model version and who approved the result.
It happens. That’s why every integration gets checks and error alerts, so we notice a problem quickly. The fix falls within the fixed maintenance arrangement. If you stop maintenance, you get the code and the documentation, so your IT department or another partner can take over.
That depends on who puts the system into use under their own name. Usually that’s your institution, and then it is the provider. Whoever uses the system in practice is the deployer. If it stays purely research, the AI Act may not apply (see above). We record together in the AI fact sheet which role you take on. Your legal department has the final say.
Approval stays with your institution. We supply what belongs in your file, ideally before you submit: an AI fact sheet with which data, which model, where it runs, who checks and what we log. If data needs to be pseudonymised, we build that in from the start.
Show us in an introductory call what already works, who needs to use it and what project deadline you face. One specific question? Request a 1-hour Pitstop.
No obligation · 30 minutes · online, with someone who puts AI to work every day
