Skip to content
Butterfly
Use case · Product · Note Taker

Note Taker
in banking

A bank. Meeting data had to stay within the bank’s network.

On-premiseson the bank’s own server
4concurrent meetings on a single server
No GPUon the bot server
0audio recordings sent to an external note service
Problem

The challenges facing the organization

  1. Under the information security policy, meeting audio and decisions could not leave the organization.
  2. So external note services could not be used.
  3. Meeting recordings and transcripts had to be produced on the bank’s own infrastructure.
Solution

How it was solved with Butterfly

Note Taker was deployed fully on-premises, on the bank’s own server. The bot joins the Teams meeting from the bank network, stores the recording on the bank’s server and transcribes the conversation with a speech recognition model on the bank’s own AI infrastructure.

  • The bot joins the Teams meeting from the bank network
  • The recording is stored on the bank’s server
  • Transcription on the bank’s AI infrastructure
  • Native installation on a Linux server, no containers required
Outcome

From the field, measured.

0audio recordings sent to an external note service
4concurrent meetings on a single server
No GPUon the bot server; models on the bank’s infrastructure

The customer name is not shared for confidentiality reasons. Figures are taken from the organization’s records.

Do you have a similar process?