AI Integration in Vicidial

[speedmaker]

Irrespective of the current legislation, is it still possible to develop an AI agent and integrate it with ChatGPT?

This could potentially be one of the most significant milestones or upgrades ever.

I would donate a five-figure sum.

regards Speedmaker

[mflorell]

There are sure a lot of companies that are trying, I have yet to hear about any true success stories with anyone using AI agents on any call center platform at any kind of scale though. The biggest issue is that AI doesn't follow a script, and it often makes stuff up. Some recent examples of big companies that have thrown large amounts of money at AI are McDonalds(which killed their AI order-taking project prematurely because it went so badly) and Air Canada (which killed their AI chat support because it made stuff up and they lost a court case because of it). There are more examples like these two, but for some reason I can recall these off the top of my head.

We've helped about a dozen VICIdial clients integrate with AI services companies in the last couple years, none of them are using AI agents at this point.

[speedmaker]

thx Matt for quick answer ! well may it takes 1-2 Years more

[carpenox]

not to mention its illegal for outbound auto dialing.

[williamconley]

Note that "AI" is an oft-misused term similar to "cloud".

You CAN train AI to be accurate where you need it to be, or you CAN use an AI that is merely grand if/then statements that are never fake and/or have limited outcomes.

Note that the existing SoundBoard built in to Vicidial could be ... enhanced to both allow a wide variety of outcomes, but still limit those outcomes to what is in fact legal/viable/available.

It is not illegal to use AI for outbound calls on the one condition that you have permission to call with an automated voice for those people you are calling.

It's also not a "legal problem" if nobody complains. It's not like speeding: The penalties are potentionally in the millions and could go higher if you get good at what you do. But if you control the system well, and train it to perfection before releasing it to the general public by having it handle inbound calls first (to perfection!) then you have a shot at making it technically legal after demonstrating its value (welcome to America where cognitive dissonance is handily embraced IF you can make enough money to graduate to the level of political Lobbying).

Can it be done? Absolutely. Will it likely run into six figures? That depends ENTIRELY on your approach, but mostly on your ability to train the AI without paying someone else to do so. The Vicidial end would not be six figures. Likely under half that.

Of course, you could offset the costs greatly if your mid-phase includes someone else's call center profiting from the limited, small scale test launches. You pay for their minutes and they keep the sales ... there will be some inbound call centers (and honestly some outbound ones) that would jump on that bandwagon and take the heat as long as there is a kill switch and monitoring involved. Just don't go the way of Derek Bartoli (look him up, I refer to the Scale of his offenses in robodialing). Keep tight controls on risk (perhaps limit them to "none"). And the timing/completion would be based entirely on your ability to manage the AI.

[Cyberwolf2]

I just looked him up. Sad the case was dismissed. More fuel to TCPA/FTC violators. I'm in sales requiring a professional license and it's regulatory body would throw the book at me for much less...

[bbakirtas]

AI technology is not suitable for use as an active agent. At least for now. AI can be used more for performance evaluation. For example, listening to agent voice recordings, catching words that should not be said in the evaluation recording, documents that can help the agent based on product, quick answers to questions, etc.

[carpenox]

very true Bill, I am using AI for quality control just as you have mentioned. great feature.

[williamconley]

very true Bill, I am using AI for quality control just as you have mentioned. great feature.

I can see how AI could be useful for flagging purposes in bulk. Definitely.