V20 U2 Alpha - AI, Microsoft Teams & Admin Control

Does that mean we will have to have an Enterprise license to continue hosting on Windows?
Yes, they've posted that here a few times and for instance the "3CX | V20 Final Released" email in March.
 
Is the sentiment value 1-5 with five being the best? (I couldn't find documentation on this, so I apologize if this has been shared elsewhere)

Also, for ChatGPT is it possible for us to modify the prompts you are sending to chatgpt?

And finally, I saw a feature that a 3rd party was offering that also would e-mail a user(s) based on criteria on the call. So, for example if someone started using profanity, a manager could be notified. This seems like something that would be relatively easy for 3CX to add in since you are already capturing and grading all the information. Is this on the roadmap?
 
  • Like
Reactions: WGIT
Just got set up on the OpenAI stuff and it's really impressive. I've started modifying the summary and sentiment parameters too with success. What would be good, is if this can be limited to extensions/departments so we can keep a handle on cost/usage.
 
Newbie end user here, just migrated from Avaya IP office. It’s great to see this update as it’s certainly getting better every time.
The status bit is a bit confusing, is there any control over the fact no calls will be received if the user has a calendar event in teams? Sometimes people have an event they may not attend, or they can still receive calls while at it. Sharing the status is interesting but not sure how good this part is. Is there no way to replicate “on a call” in 3CX like Teams and outlook shows?

Secondly, with M365 integration working so well, it seems incredible we can’t turn off the traditional username and password login and force M365. Is there any way to do this? Even per user to stop them getting the automated 3CX email to set a password when we only allow M365 for authentication? It would also be helpful to confirm if a user that’s blocked from signin on M365 (ie a leaver or a compromised account) will also not allow them to sign in via 3CX?

Thanks again for the updates.
 
  • Like
Reactions: stumac
I just tried the new AI-features and called the customer service of Ziggo (just the menu)
You get a computer that asks you for your postcode.
It's just a simple test of course, but I am amazed about the summary of that conversation.
'The Ziggo-agent welcomes the customer and asks them to say their postcode in a certain way for further service' (but then in Dutch).
It's just amazing how well this works!

AI is not very good yet at answering general questions (see all the crazy examples you can find online), but for summarizing and/or sentiment, voice recognition etc. it is really great!
 
Oh, and thanks a lot for bringing back the Activity Log!
That's just a really big life safer!
 
Even per user to stop them getting the automated 3CX email to set a password when we only allow M365 for authentication?
Edit the welcome template and remove that section. Doesn't stop someone from going to the webpage and doing a forgot password though.
It would also be helpful to confirm if a user that’s blocked from signin on M365 (ie a leaver or a compromised account) will also not allow them to sign in via 3CX?
Yes, 365 won't pass a claim back to 3CX. So the login with MS button won't work, but they could still login with 3CX native credentials.
 
Just a question about OpenAI and privacy.
Voicemails aren’t that sensitive, but phone calls usually contain privileged information. Customers can give their account number, date of birth, IBAN and things like that. The entire recording is then uploaded to OpenAI for the transcript and also for the summary and sentiment.
I’m assuming the transfer between 3CX and OpenAI is encrypted with TLS.
But what does OpenAI do with the data? Are the recordings deleted after processing, but more importantly, is the transcript also used for training? It would be undesirable if such privileged information is used for training because you never know what happens with it and if OpenAI (ChatGPT) might ever use that information in an answer it generates.
I didn’t really find a definitive answer to these questions in the OpenAI privacy policy.
 
Edit the welcome template and remove that section. Doesn't stop someone from going to the webpage and doing a forgot password though.

Yes, 365 won't pass a claim back to 3CX. So the login with MS button won't work, but they could still login with 3CX native credentials.
Thanks, good info. I guess my question/suggestion is that we really should be able to disable native 3CX credentials totally on an enterprise system, even on a per-user basis and force SSO of whatever variety we prefer.
 
@speedbird001 @SweetAction If I recall correctly the ability to force SSO for certain users was mentioned at a VIP meeting. Hopefully someone from 3CX can confirm here.

I'd want to force most users to use SSO, but still allow 2FA for some. A global default would be nice. If they try to login using 3CX creds it should identity them as an SSO user and switch them to that login method and no resetting of the local password allowed for those users. We would still need 2FA accounts with local creds for users without access to the SSO provider, break glass accounts, and for managing client systems.
 
Can the transcribed texts be read from the database for further use via API?
 
@speedbird001 @SweetAction If I recall correctly the ability to force SSO for certain users was mentioned at a VIP meeting. Hopefully someone from 3CX can confirm here.

I'd want to force most users to use SSO, but still allow 2FA for some. A global default would be nice. If they try to login using 3CX creds it should identity them as an SSO user and switch them to that login method and no resetting of the local password allowed for those users. We would still need 2FA accounts with local creds for users without access to the SSO provider, break glass accounts, and for managing client systems.
Yes, I remember that as well from the meeting in December
 
  • Like
Reactions: accentlogic
Hi team
A mini update

Multi Edit - Coming soon See video of how it will work.
https://drive.google.com/file/d/1gmluPFeeAHAZrMKnvGLOlPp8PJ7w6gQv/view

Teams dual sync - Under investigation.

Checking posts on this update. Any bugs or reports, now would be a great time as we have the last remaining space allocated for technical debt in the update 3 sprint.

Have a great day ahead
 
Can the transcribed texts be read from the database for further use via API?

For the time being they are not exposed. But we are planning DB mart and re-org so let's park this discussion for the time being because many things will change.
 
  • Like
Reactions: Evolute IT
Hi team
A mini update

Multi Edit - Coming soon See video of how it will work.
https://drive.google.com/file/d/1gmluPFeeAHAZrMKnvGLOlPp8PJ7w6gQv/view

Teams dual sync - Under investigation.

Checking posts on this update. Any bugs or reports, now would be a great time as we have the last remaining space allocated for technical debt in the update 3 sprint.

Have a great day ahead
Looks great Nikolas!

One question, that video is using either Pro or Ent. Will the same be available in SMB?

Cheers
 
Just a question about OpenAI and privacy.
Voicemails aren’t that sensitive, but phone calls usually contain privileged information. Customers can give their account number, date of birth, IBAN and things like that. The entire recording is then uploaded to OpenAI for the transcript and also for the summary and sentiment.
I’m assuming the transfer between 3CX and OpenAI is encrypted with TLS.
But what does OpenAI do with the data? Are the recordings deleted after processing, but more importantly, is the transcript also used for training? It would be undesirable if such privileged information is used for training because you never know what happens with it and if OpenAI (ChatGPT) might ever use that information in an answer it generates.
I didn’t really find a definitive answer to these questions in the OpenAI privacy policy.

Yes comms between 3CX and openAI are encrypted. Everyone needs to prevent unauthorized access.

Read here https://openai.com/policies/eu-privacy-policy/

OpenAI say that the data is deleted after the processing is complete. But I have trouble believing this. Because to have the best model in the world, you need the data. So the data became priceless. The metric that indicates success is today DATA

Why is lama3 of Meta so good for example? Because Meta has unlimited data to feed in it's model.
Why are the other models not so good? Because the data they can get their hands on is limited.

Look at this clause:
1718772519938.png

So on transcript use, OpenAI's policy states that data sent to their API can be used to improve their services.
There is also an opt in to share data to allow them to use the content you give them to train their models but I have not seen this yet.
 
Looks great Nikolas!

One question, that video is using either Pro or Ent. Will the same be available in SMB?

Cheers

Hi Ray
Yes, there are no restrictions on this feature. It is designed to assist and save time for administrators. When it comes to enhancing administrative efficiency, we do not impose limitations.
 
Yes comms between 3CX and openAI are encrypted. Everyone needs to prevent unauthorized access.



Look at this clause:
View attachment 42042

So on transcript use, OpenAI's policy states that data sent to their API can be used to improve their services.
There is also an opt in to share data to allow them to use the content you give them to train their models but I have not seen this yet.


After my original post, I have also done some additional digging.
On https://help.openai.com/en/articles/5722486-how-your-data-is-used-to-improve-model-performance they state this:

Services for businesses, such as ChatGPT Team, ChatGPT Enterprise, and our API Platform

We don’t use content from our business offerings such as ChatGPT Team, ChatGPT Enterprise, and our API Platform to train our models. Please see our Enterprise Privacy page for information on how we handle business data.

So it would seem that at least it's not used for training. That's how I believe it should be. And there is some more info here
https://openai.com/enterprise-privacy/

Our commitments​

Ownership: You own and control your data​

  • We do not train on your business data (data from ChatGPT Team, ChatGPT Enterprise, or our API Platform)
  • You own your inputs and outputs (where allowed by law)
  • You control how long your data is retained (ChatGPT Enterprise)

So it would seem everything is covered by these policies?
 
  • Like
Reactions: NatalyS_3CX

Forum statistics

Threads
111,874
Messages
589,503
Members
164,723
Latest member
aman-pandey