Activity
Mon
Wed
Fri
Sun
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec
Jan
Feb
What is this?
Less
More

Memberships

69 contributions to Assistable.ai
Nvidia PersonaPlex - Very promising
@Jorden Williams worth a look. Putting it on your radar... 😎 https://youtu.be/n_m0fqp8xwQ?si=SJOf4K2fhR7BXHUC
0 likes • 19d
The comments on that video are comedy gold btw, and it clearly needs some work, but still cool.
0 likes • 4d
@Halie Bot I will help any way I can.
If you still have issues with GPT-4o may try GPT-4.1Nano
Since last Thursday our voice AI agents went completely off and I’ve seen many of you had the same with GPT-4o. It was perfect before, but suddenly made our AI unusable. I waited until today, then decided to rebuild. After a few hours of refinement I’m now getting the best results with GPT-4.1 Nano. Still not as good as 4o used to be, but workable. And AI no longer speaks tools or actions out loud. Anything above 4.1 Nano didn’t work for me.
1 like • 9d
@Vicente Vercellino The simplest solution is to just remove custom fields unfortunately if you want to be 100% sure it never makes that mistake. But I also tried the following which worked: ## Contact Fields If the value includes "contact." always refer to the key in a casual and natural way when referencing a contact field. This is because you do not have this information present and should not mention it in conversation. If the value includes "contact." you do not have the correct value and the value should remain unknown to the user at all costs. If the user asks for specific information from the value and the value includes "contact." always let the user know that for security purposes you are unable to provide that information since they are calling from a different phone number than the one on file. { "name": "{{contact.name}}", "first_name": "{{contact.first_name}}", "last_name": "{{contact.last_name}}", "call_in_date": "{{contact.call_in_date}}", "lender": "{{contact.lender}}", "mortgage_amount": "{{contact.mortgage_amount}}", "sale_date": "{{contact.sale_date}}", "address1": "{{contact.address1}}", "phone": "{{contact.phone}}" } Field Descriptions: "first_name": "last_name": "name": "lender": "mortgage_amount": "mortgage" "sale_date": "address1": "call_in_date": "phone": In Task 1 under the introduction: - #If the user's name includes "Guest", "Caller", or "Orb", always assume you do not have the user's name and always ask who you are speaking with. -- ##Example Response: "Hello, my name is Lillian. Who do I have the pleasure of speaking with?"
0 likes • 6d
@Vicente Vercellino Anytime! Hope it helps 🙏🏻
Fav AI Model
Hey, what's everyones favourite AI model to use for their assistant? @Jorden Williams what do you recommend?
Fav AI Model
1 like • 11d
@Jorden Williams @Alexa Briggs @Vicente Vercellino @Timo Weber @Simeon Krastev @Robert Boydston Sorry for the late reply, I've been working on some massive rollouts and new TV ad campaigns. It was working amazing with Haiku 4.5 and Andromeda from Rime when I tested it yesterday and I really liked Emily and Grace. That being said... I just did a demo for a massive group live in Zoom of about 60 brokers and the voice on it was super slow... and it also appears that the AI is rambling through tasks again.... Egg on my face... 😔 When it worked though, man was it good. Call examples: Example 1: Model: Haiku 4.5 Voice: (Rime) Andromeda / Default Voice Engine / Multi-Lingual Processor Voice speed 1.55x call_87f897b8-0d95-4e2b-8313-03f849c4dcb8 Example 2: Model: Haiku 4.5 Voice: (Rime) Andromeda / Default Voice Engine / Multi-Lingual Processor Voice speed 1.55x call_a7c7798a-bb6b-423a-934a-4c73e3db5100 Example 3: Model: Haiku 4.5 Voice: (Rime) Andromeda / Default Voice Engine / Multi-Lingual Processor Voice speed 1.0x call_c2645ba8-1b1d-4416-8e20-44f21c27227b Example 4: Rambling call: Model: Haiku 4.5 Voice: Grace / Default Voice Engine / Multi-Lingual Processor Voice speed 1.55x call_57f56571-2dcc-46f1-85c9-5b1cef48e484
0 likes • 9d
@Alexa Briggs Yup. I am also experiencing that as well. It happened in a live presentation where I was demoing the new great voices for about 60 brokers. Super embarrassing.
GPT 5 Nano
Anyone switched to 5 Nano model yet? If so, curious how it runs. It's faster, cheaper and 2.5x the context window. I use 5 Nano in a lot of my other stuff (its awesome), but haven't tried it for chat/voice. If you've used it with Assistable, how did it do?
1 like • 13d
Curious about this as well. I've been on 4o for ages, but thinking about trying some other models.
2 likes • 12d
@Cemre Sancaktar I just tried Haiku after seeing Jorden's post yesterday and ended up switching all of my bots to it. Crazy to think that all this time GPT4o has had such a negative affect on the voice and I didn't know it, but that seems to be the case. I think we need to encourage some more frequent discussions like this regarding the models from time to time. Very big difference, and a noteworthy development.
Inbound Caller Name Issue!!
Why my inbound call agent does not saves the name, email and phone of the lead. It saves it as Guest00909 something like this. How to fix this please help. Thanks!!!
0 likes • 13d
Is your incoming-call webhook setup?
1-10 of 69
Anthony Castiglia
4
15points to level up
@anthony-castiglia-6799
Matthew 19:26

Active 1d ago
Joined Aug 17, 2025
ENTJ
Powered by