I think the team is being pro-active as Jorden and Mike, and Bernie showed up couple of times already stating they are aware of the issues and actively working on them. I'm with you that timelines would be great, but building technology, especially in the AI world leads to unexpected, unprecedented scenarios especially with so many moving parts - GHL making changes, elevenlabs making changes, LLM models getting sunsetted and new ones emerging, tool calling etc. I'm yet to use a technology piece that doesn't break, let alone a fully autonomous AI agent platform that works across all channels. And I get it as I'm also getting pushback from clients, complaints and ticket daily, and these things are normal. I've been getting it in a regular Facebook ads agency, in GHL SaaS clients, or any other capacity. The support team has been helpful and jumping in to even update prompts in our accounts to address the recent changes in the booking protocol from GHL so the team does deserve credit @Assistable Ai @Assistable Team I'm sure they are pushing hard to make sure we can fulfil on the promises to our clients. In the world of AI we just need to accept that things will be shaky, and that's where the opportunity is at. If everything was stable, easy, and didn't require input and heroic efforts we wouldn't be needed as mediators. I'm saying all of this not to diminish your frustrations (I feel your pain as I experience the exact same things), but just giving you a perspective to look at. Even today with all the things that need polishing and updating, the platform works as intended 95% of the use-cases and most clients are extremely happy.

No, unfortunately with all the models Claude/gpt etc and all prompt variations it still comes through on % of the calls. Proper prompting and stronger reasoning model helps reduce the occurrences but is by no means fixed. Looking forward to stable long term solution, as for brick and mortar appointments it’s crucial.

Just redirect your personal cell to the AI phone number after X beeps. Your provider can help you do that. You have to dial a number to set it up. Check Verizon, A&T, call redirect rules.

@Anthony Castiglia For me the latency goes up a bit. And I'm also curious on the tool calling accuracy. We have a few dynamic Rag databases and wonder if Haiku would pull as good as GPTs. Also just got the booking date / weekday resolved with 5.2 gpt wonder if Haiku would be the same.

@Anthony Castiglia Just tested and exactly the same thing for me. When asked for Friday says I have availability for 7th and goes on to book you for Saturday.

With all of the adjustments and models it keeps happening for me. It's not 100% of the time, but on 10% of the calls/texts it'd choose a wrong date/day of the week. Added all things to the prompt. Here are some call IDs: call_27d95a05-02c8-4b6a-a238-2030fe06e918 call_97eb9c32-f27a-4f44-8b47-dc0641b0485b call_e1d4ba63-3f61-4f50-8e5b-4db2c5146b60 call_db57ec9a-7d8a-47ac-99e1-e5b813301bd7 Maybe it's something with the internal tooling pooling last years calendar dates or something @Jorden Williams @Mike Copeland

@Jorden Williams Yes, probably something to do with that. As it happens accross channels, accounts, with different prompts, etc. Thought the 4o being sunsetted February (not the API) has something to do with it, but happens on 5.1 and 5.2 as well.