The values are reference values for planning. Review utilization after rollout and adjust the tier if necessary.
Recommendation by headcount
Count all people with access to VARIOS AI. It is assumed that on a working day 40% of them use VARIOS AI and send an average of 8 messages. Load peaks are taken into account.Recommendation by concurrently active users
Use this table if you know the expected concurrency, for example from a pilot phase. A concurrently active user is currently chatting and sends a message about every two to three minutes. Open browser tabs without interaction do not count.
If you measured concurrency at normal times, use three times the value as a reserve for peak times before reading the tier. If the two tables yield different tiers, the higher one applies.
Storage during operation
The database grows with the chats. With mixed usage of text and documents, this averages about 100 KB per question, or about 80 GB per year per 1,000 employees. The exact value depends heavily on usage; pure text chats need considerably less. With a chat retention period of 90 days, it stays at about 20 GB per 1,000 employees. Uploaded files add to this at their full size; their volume depends on usage and should be planned generously. Knowledge bases require about 7 MB per 100 pages of document text. Plan twice the database size for database backups.What shifts the values
- Knowledge bases and file uploads. Large volumes of documents increase the load of the DLP analysis. Load large knowledge bases outside working hours or choose one tier higher. For special requirements, such as DLP scanning of very many or very large documents, the DLP analysis can be moved to a dedicated server. Please contact support for this.
- API integrations. Applications that send requests via API keys do so without pauses and often several requests in parallel. Count each concurrently running API request as three concurrently active users.