Beyond chat and voice: Realistic real-time video avatar with dynamic UI

[AbrahamJaffar]

As a software engineer, Imma be homeless in less than a year due to AI. So before AI comes for my job let me join forces with it to create something cool.

I looked around the market of AI and nobody has done the hard job of redesigning the UX beyond a chatbot or voice. Both are replacements, not true innovations. What if the user is too lazy to type? What if the user is too scared to call and give up their number? You have used blackberries of AI till now. Let me introduce you to the iPhone.

Introducing Samantha, (loom video demo) from caravatars.com

Samantha is a real-time realistic video agent that greets visitors on your website. It can dynamically change UI to fit queries from inventory, trade-in, financing and booking. Oh and unlike a voice AI call where the user can't see the info they are giving and chatbots where they have to type it manually. This tech breezes through everything like a BD person would. Except its faster than them and can actually show people their cars and options visually.

And the best part, the user consents clearly to talk to an AI, the expectation of talking to a human and the mistrust when AI replies is flipped directly on its head. Not to mention sidestepping the telemarketing laws entirely. Instead of getting a form filled and nurturing it. Nurturing happens before any information is ever committed with arguably the smoothest UX experience you could ask for. When you hold hand of a buyer through the process from step 1, there is no way they will buy from anyone else.

That was me being super excited about it. But we are at very early stages so skepticism and critical feedback is really needed. What do you think? Will it achieve results or just be another gimmick?

 

[MattCopley]

We built these almost 3 years ago and tabled the project. In the AI space you're going to encounter a lot of moments of "just because we can, does it mean we should?" My sense is shoppers want human connection more than another layer of automation... and it's hard to deliver a value proposition other than replacing humans and reducing institutional knowledge.

 

[AbrahamJaffar]

We built these almost 3 years ago and tabled the project. In the AI space you're going to encounter a lot of moments of "just because we can, does it mean we should?" My sense is shoppers want human connection more than another layer of automation... and it's hard to deliver a value proposition other than replacing humans and reducing institutional knowledge.

Thank you for your feedback, the value proposition for buyer is easier and more intuitive UX and for dealership is more and warmer leads.

Did you achieve this quality of responses and flow? the llm's were quite new back then and their tool calling ability was abysmal. Today you can setup a full audio visual journey of a user. Do you think that's simply not valuable and users won't like it?

 

[MattCopley]

Thank you for your feedback, the value proposition for buyer is easier and more intuitive UX and for dealership is more and warmer leads.

Did you achieve this quality of responses and flow? the llm's were quite new back then and their tool calling ability was abysmal. Today you can setup a full audio visual journey of a user. Do you think that's simply not valuable and users won't like it?

We had ours trained to do some pretty amazing stuff and had them grounded in a wide array of data. We're just trying to remain focused on how we amplify people vs replace them. If you look at engagement analysis broadly across social platforms people tend to prefer humans vs avatars, and we found the closer we got to a truly human experience it drove trust erosion with humans starting to question the synthetic experience. I imagine this will get traction and I'm not trying to dissuade you. These are rolling out in every business vertical right now and fairly commoditized in many... it just came down to an ethics question for us.

 

[AbrahamJaffar]

We had ours trained to do some pretty amazing stuff and had them grounded in a wide array of data. We're just trying to remain focused on how we amplify people vs replace them. If you look at engagement analysis broadly across social platforms people tend to prefer humans vs avatars, and we found the closer we got to a truly human experience it drove trust erosion with humans starting to question the synthetic experience. I imagine this will get traction and I'm not trying to dissuade you. These are rolling out in every business vertical right now and fairly commoditized in many... it just came down to an ethics question for us.

Matt, Thank you so much for your response. I agree that this agent would never replace people. People don't buy 30k cars from an agent. It just curbs the anxiety of talking to a salesman when you don't even know what you want yet and whether it fits in your budget. Preliminary screening stuff.

Maybe I think the tone of my first message was too excited and suggested the replacing of humans. No way, people don't trust AI and they never should. AI is only good to get information easier, faster and more naturally. Grounded by real data it can help a lot of undecided and confused buyers into an informed and convinced buyer. That's about all of it.

Btw, your initiative of open data is very commendable. If I can do anything to support it. let me know.

 

[dealerdata1]

There has been a saying going around for a while that I have been using more of

"People buy from people, not software"

And I think this will continue to remain true unless dealers as a whole continue to erode trust in the buying process while other operators streamline buying (Carmax, Carvana, Amazon, ect.)

Even though I use a massive amount of AI and it is a major feature in the software I built, I despise interacting with is as a customer in other industries. AI Phone systems are just a nightmare now when you call in to businesses

 

[AbrahamJaffar]

There has been a saying going around for a while that I have been using more of

"People buy from people, not software"

And I think this will continue to remain true unless dealers as a whole continue to erode trust in the buying process while other operators streamline buying (Carmax, Carvana, Amazon, ect.)

Even though I use a massive amount of AI and it is a major feature in the software I built, I despise interacting with is as a customer in other industries. AI Phone systems are just a nightmare now when you call in to businesses

Thank you so much for your valuable feedback.

I agree actually, AI is their only for giving information in a low commitment way. Do you know many people dread talking to a car salesman? Sixty-two percent of Millennial women and 52% of Millennial men said they feel pressured by dealership salespeople to buy something right away. These people delay car purchases because of that. People with those forms on anxieties can explore with AI before they are ready to put in their info. That means the dealership that uses such agents can assuage the trust barrier and slowly nurture them into taking the leap to give out their info.That's when humans take over

When you call a business you are expecting to talk to a human so you feel cheated when its AI. With Samantha that expectation is not true. Therefore, you will have a better experience. For closing deals nobody can beat a human. But to keep people engaged and get their foot in the door. AI can definitely help.

I would really love to hear your thoughts.

 

[MattCopley]

Thank you so much for your valuable feedback.

I agree actually, AI is their only for giving information in a low commitment way. Do you know many people dread talking to a car salesman? Sixty-two percent of Millennial women and 52% of Millennial men said they feel pressured by dealership salespeople to buy something right away. These people delay car purchases because of that. People with those forms on anxieties can explore with AI before they are ready to put in their info. That means the dealership that uses such agents can assuage the trust barrier and slowly nurture them into taking the leap to give out their info.That's when humans take over

When you call a business you are expecting to talk to a human so you feel cheated when its AI. With Samantha that expectation is not true. Therefore, you will have a better experience. For closing deals nobody can beat a human. But to keep people engaged and get their foot in the door. AI can definitely help.

I would really love to hear your thoughts.

Yep you're spot on with all of this but the fundamental problem still exists. The real paradox is when we improve these synthetic experiences, it creates an even larger experiential gap when the customer eventually interacts with a live person driving downstream CSI and review issues... It's essentially the Disneyland/Jurassic Park paradox amplified 100 fold... So the needle we've been working to thread is how do we 10x humans, 10x institutional expertise etc, so that when a customer interacts with the dealership across surfaces and in person the experience is consistently good. It also lays the foundation for a dealer to layer in a tool like yours without amplifying the CX gap.

 

[AbrahamJaffar]

Yep you're spot on with all of this but the fundamental problem still exists. The real paradox is when we improve these synthetic experiences, it creates an even larger experiential gap when the customer eventually interacts with a live person driving downstream CSI and review issues... It's essentially the Disneyland/Jurassic Park paradox amplified 100 fold... So the needle we've been working to thread is how do we 10x humans, 10x institutional expertise etc, so that when a customer interacts with the dealership across surfaces and in person the experience is consistently good. It also lays the foundation for a dealer to layer in a tool like yours without amplifying the CX gap.

(Sorry for long response, your post deserved a thorough exploration)

I never thought about the problem of consistent CX problem. Thank you so much for giving me such valuable insight. In any experience people remember the peak and the end. And that's how CSI would actually mostly playout. The question would be what are the variables that can affect the CX and how can we improve it alongside the AI and sometimes use AI to help us.

Take inventory search for example, BDR have to type in and get answers and tell the buyer, that's extremely slow compared to AI which does it in less than a second (mine does, most don't). So if a user's expectation is fast answers they just don't wanna talk to humans. That's the opposite of our intent. We want the AI to lead to humans and consider humans superior.

Because this is a general problem, we don't have a silver bullet solution. We would need to follow the principles of least friction in all the processes also manage expectations by making the mind of user, "this takes a bit of time okay?". The tasks that AI's are better at can be done by AI on the showroom floor. How? Because, Samantha can be commanded by voice. You can have a tab open and it will answer not only inventory questions to BDR but intricate nerdy car stuff that salesmen don't know, without them having to look it up.

But focusing on this we are leaving out a very important metric: first impressions

> 88% of car buyers who start their journey online report that the digital visual experience (detailed photos, transparent vehicle specs) on a dealer's website is the single biggest factor enticing them to take the next step and contact the store.
> 78% of online leads ultimately buy from the business that responds first.
> Dealerships that fail to win customers during their online search and website visit never get the chance to compete on the showroom floor.

As AI becomes standard the dealerships that will have a chance to even compete on the whole experience are the ones that use AI to give the best first impression. Salespeople cannot match AI in speed and clarity of information. We must not lose sight of the forest for the trees.

The second problem you talked about is we delay human interaction. This problem is not as hard as the first one. During BDC hours we can simply route a qualified user who is interested to talk to a real human from inside the agent conversation, because the conversation is already voice based that can happen seamlessly instead of them having to take out their phone and dial. When BDC is closed, we send personalized yet real video of a salesperson to let them have an interaction with a real human. And we update the lead list so that a human must call them back tomorrow.

Please share more objections. I am learning a lot. Thank you.