Do you have private secure AI?
Or just Copilot for a few people?

No. EP365 AI Assistant calls Azure OpenAI Service hosted in your Azure subscription, in an EU region. Microsoft contractually guarantees that your content is not used to train models and does not leave your tenant. We (Accelapps) only have access to a service account for deployment and support — you can revoke it any time.

Yes — if you want it to. The Company knowledge mode is switched on right next to the question box: the AI searches your SharePoint before answering, replies with citations and shows cards for the sources it used — a click opens a document preview right inside the app. The admin controls the search scope (whole tenant, a site, selected libraries) and you can narrow it further in the question itself. No index and no copy of your data is ever created. Ad-hoc attachments still work alongside this: a user attaches a specific file and the AI works with it within that conversation.

No. The search runs under the identity of the specific user through standard SharePoint search, which trims results by permissions (security trimming). The app has no service account with broader access to content — two colleagues can get different answers to the same question depending on what each of them is allowed to see. For compliance you can optionally enable an audit: who searched for what and which sources the AI used.

Conversations stay stored in your Microsoft 365 and every user only sees their own — nobody can open someone else's conversation by any route. For compliance you can optionally enable admin oversight (read-only, off by default); the user always sees a clear indicator of whether their conversations are private or visible to an admin. Sign-in uses the company Microsoft account, no extra password. Attachment content is never stored — only the filename.

PDF, Word (DOCX), Excel (XLSX), PowerPoint (PPTX), TXT, MD, CSV — we extract text from these. Images JPG, PNG, WEBP. The default attachment size limit is 10 MB (configurable).

Copilot is excellent inside a specific app (Word, Outlook, PowerPoint) and is licensed per user. EP365 AI Assistant covers the whole organisation for one company-wide price: chat on an intranet page and in a floating widget, Company knowledge with citations and source previews, prompt templates shared across the team, daily per-user limits and an admin cost overview. We often deploy both — side by side.

Every user has a daily request limit (default 50, configurable). Admins get a dashboard inside the app with monthly spend in USD, token counts, a daily chart and the top 10 users. Repeating an answer (retry) doesn't count against the daily limit.

The technical setup of the chat itself takes a few hours. A complete Business deployment (installation, configuration, training) takes 5–10 working days from kickoff, Starter (1 SharePoint site / 1 department) 3–5 days. Enterprise is scoped individually based on scale and regulatory requirements.

Yes — in three layers. Personal templates are visible only to the owner. Site templates (created by the SharePoint site admin) are visible to everyone in that team. Global templates are managed centrally by the admin in the EP365 Homepage app and are visible to the whole organisation. Every template has an icon, colour, name and prompt body.

Yes, when the EP365 Homepage app (hub) is deployed. From one place you then manage global prompts, authoritative sources for Company knowledge (three modes — extend the scope, prefer them in ranking, or search only within them) and default AI parameters: answer tuning, token prices and the Knowledge preparation indexing prompt. Each site either inherits these settings from the global configuration or sets them locally — via the “Inherit from global / Own local” switch right in the Assistant's settings. Without Homepage deployed, the Assistant runs on that site's local settings only.

None, but watch out for Azure consumption — LLM tokens come straight out of your Azure subscription and typically run between 1,500 and 8,000 CZK per month depending on volume. The admin dashboard shows the daily curve. In the first month we evaluate real usage and fine-tune the model + caching.