FAQ
Short answers to the questions that come up most often.
Which AI model does Elicira use?
An open-weight model (GLM-5.2), out of a sovereignty conviction: no dependency on a closed proprietary model.
Where is my data hosted?
Your account and work data are hosted in Canada (Montreal). Prompts, however, are processed by an inference provider located in the United States, under standard contractual clauses and with no reuse of your data for training. Details in privacy.
Is my data used to train the AI?
No. The model is used without reusing your data for training.
Which languages are supported?
French and English, for both the interface and the assistant’s answers. Set it in preferences.
Can I export my deliverables?
Yes: copy the text in one click, and BPMN 2.0 export for process models. See the deliverables library.
How do I delete my data?
From Settings → Danger zone: deletion is complete and overrides the retention period.
What happens to my data if I do not come back?
From 1 December 2026, an account left twelve months without a sign-in is deleted, and every sign-in resets that clock. You get an email 30 days before. Details in the privacy policy.
What if I stop paying?
Your deliverables do not vanish with the subscription: you keep read access and export for 30 days after the paid period ends, without being able to generate more. Access then closes, but the data is still kept and you find it again if you resubscribe. Exact periods are in the privacy policy.
Why does the assistant run a module on its own?
Because it reads your request before answering it: if it recognises a deliverable in there, it runs the matching module and shows you its progress in the thread. You can still open a module by hand from the composer chips. Details in how the assistant works.
Can I ask for two deliverables at once?
Yes: “prepare my meeting then write the user stories” produces both, in that order. Beyond two modules in a single message, better to split it up: too long a chain risks being cut short.
Does the AI replace the business analyst?
No. It proposes a structured draft; you make the call at every step. The principle is filtering > generation.