Digital Personal Assistants (DPA)

[joe.pistell]

The Beast is knocking at the door.
Digital Personal Assistants (DPA), powered by AI are going to disrupt EVERYTHING. It will come to our industry in multiple waves.

WTF am I talking about?
Wave 1: I explore this in a LI article: Dealer's Tech Planning in 2024: AI+CRM = AI personal assistants. In a short period of time (24 months?), the entire Auto Shopping workflow -as we know it- is TOAST. 2 types of dealers will exist. Those with DPAs and those without.

WTF is a DPA?
Digital Personal Assistants are 100% task oriented. They are a Digital Personal Assistant with a highly structured job description that can help anyone with a job to do.

DPAs need two building blocks.
1. Structured data
2. APIs to the structured data.

Structured data.​

Automotive retail is famous for it's cesspool of semi-structured data. IMO, this alone is THE reason why Wall Street has not been able to 'roll-up' the automobile industry. Automotive is a small universe highly repetitive uses. There is a great reward to bring order to the semi-structured data, LLM's strength is using text, words & language to solve complex problems. I expect mega-vendors in our space to attack this opportunity.​

​

APIs to Everything.​

DPAs need APIs to solve domain specific tasks (e.g. build data, marketplace data, vehicle valuations, loan values, reviews, and on and on.) Once the DPA has a stask to complete, it'll know which APIs to fish from. More APIs = smarter DPA. API owners that have data that is in demand will see API traffic as a core business.​

Does the DPA sell cars?
In Wave 1, No. But, it will help everyone in dealership align to sell MORE cars with less people.

Wave 2: DPAs galore.
Your store will have a DPA for nearly every employee in the store. Sales rep's DPA will assist them in digital workflow (product knowledge, lead management, goal setting and feedback, paper work, compliance, etc). Nearly every employee will have their DPA. Therefore your store will have a 'swarm' of DPAs, all aligned -and- all aware of each other. The GM and DP will have their own DPA that will oversee 'the swarm' of DPAs. If you dont think this is real, it's already in use (google: "multi-agent swarms" or "Microsoft's Auto-Gen" product)
Next, the majority of your 'high value' customers will have 'domain-specific' DPAs that are fully aware of the human owners entire life. They will own Automotive DPA that will become their trusted automotive SME (Subject Matter Expert). There is nothing out of bounds here, right down to the DPA negotiating with multiple dealers on behalf of it's owner (yes, this means consumer DPA to dealer DPA discussions)

Think about it.
Losers: Google as we know it is TOAST. Gone are the days you'll go to Google and 'pogo-stick' back and forth to get answers. ALl Classified sites as we know it... toast.
Winners: Anyone that the DPA needs or serves. Data sources that feed the APIs. CAR DEALERS that own the franchise or hold the inventory.

Wave 3: Krakatoa
Domain specific DPAs are no longer needed. SuperIntelligence is here. I will NOT speculate on this. I am an optimist. I follow David Shapiro for genius level insights.


Takeaways:
ALL VENDORS: Beware incrementally solving today's digital workflow frictions. Digital Personal Assistants (DPA) will blow up everything that has a digital workflow.

Example: @JDHigginbotham's Vonage is a CPaaS (Communications Platform as a Service). IMO, adding AI to the CPaaS platform is perfectly positioned for the 1st wave (they have to hurry find ways to not let the AI wave blow by them).

 

[joe.pistell]

This article took me 3 hrs to compile. If you like it, plz hit the like button, or, toss in your comments for some Q&A.

 

[JDHigginbotham]

Great insight, @joe.pistell. It is context like this that has drawn me back to Auto. You've poked the bear, so hopefully a long reply is of value.

Background
Much of the activity I've had within this community has focused on the basics of using individual communication channels. The following is that next step.

As I follow more content speaking to the evolution of retail automotive, I see many of the first steps you mention in Wave 1 happening now.

Current State
Everybody wants to be a CDP. Some are opening up data transfer via API. Only few are leveraging LLMs and conversational channels to activate the data. As result, Auto's current state appears to be doing great work with data but falling short with highly targeted "notifications" instead of "conversations".

My intention is to live in that activation piece by combining the flexibility and capability of Communication APIs and AI.

Problem
When it comes to data activation, Auto seems to be focused on integration. "I have the data, you have the conversation. Lets integrate and make that conversation smarter!". I'm not against this approach, but it isn't always the best route.

1. You need to justify the integration...dev lift, partner value, commercials, quality/fit of end product
2. Even a justified integration still depends on fairly static, fully-baked products...further iteration requires further justification
3. Gaps will still exist in moving the customer from one part of the journey to next.

The current perception of OWNING an experience rather than borrowing another with integration may be overestimated in it's difficulty - leaving the power of DPAs to less than it could.

Solution
You mentioned Vonage's need to combine AI and CPaaS in their offering. Enter Vonage AI Studio. It's sole purpose is to meet the market with the need you are describing with DPAs being omnipresent.

Right now, anyone with basic coding knowledge could leverage AI Studio to build/train virtual assistants in a low-code manner within their application. You can bring your own LLM or lean on us to give you access/guidance to OpenAI. Our Natural Language Understanding (NLU), Automatic Speech Recognition (ASR), and Text-to-Speech (TTS) capabilities are proprietary and mature.

So, if you have powerful data organized in a manner that nails identification...you no longer need to depend on integration to activate. The effort to own both the channel and the intelligence has been drastically reduced.

Placed well, AI Studio could democratize journey building with Conversational/Generative AI on engaged channels for platforms, OEMs, and even tech-savvy dealers.

I have this pipe dream that connectors in AI Studio could help bring data together that otherwise would be silo'd.

 

[Carsten]

Google once people start to value chatGPT for search will be in danger and bing will grow.
But ... what is AI really?

it's a great search tool, it's great at creating somewhat human writing, it's still can't do fingers, and apparently sucks at music.

What really is an API? When you click the blue post reply button, you have clicked on the early generation of an API. If the button triggers a posting to another website then it would likely be using an API. It's like using a cell phone in a way to make a call.

Can AI code? Not really. It can give you steps and it can give you the boiler plate to the code but it's still up to you to create the function and know why it's broken. it can help you install your alpine radio but Bing search results will only be able to help you so much to solve why the display isn't showing.

WE are still a ways off from the LLM actually being useful.

This idea of DPA is likely going to happen and the sales man is going to be a glorified valet. Carvana is already showing this happening. THe towers are pretty impressive.

Why do I need a salesman if everything is no haggle and all they do is relay delayed information on a load the financing guy is going to jack up my payment?

Customers used to play the game for a few hundred in "discounts" but will they continue?

 

[Todd Thompson]

This direction has a lot of promise.

rabbit — home

r1: your pocket companion. Order now: $199, no subscription required.

www.rabbit.tech

Can see it on youtube in action.

View: https://www.youtube.com/watch?v=22wlLy7hKP4


The concept is: you train (show) rabbit how to navigate your desired applications by adding them to the 'rabbit hole' dashboard once and from then on just ask it.
LAM (Large Action Model) which is the Rabbit direction - can actually do stuff
LLM great at understanding what you want, can't do anything really

So uses both above

It can combine various apps think "create uber request to airport, book your flight and hotel in paris (expedia), play some travel music while you are going (spotify), email your work to let them know you quit and brag about your trip on Facebook.

Just simply ask it to do that in one shot.

Easy DPA

 

[JDHigginbotham]

I started scanning once he presented the idea I would throw away my iPhone for a voice activated device.

I completely agree with him that consumers will depend less on individual mobile apps. I am in support of the idea that we could all bail on iOS/Android native apps for web-based experiences...this, though, doesn't require adopting a new device.

His entire demo could have been done either on a voice call or messaging thread.

A much more realistic approach would be to build fluid journeys that can reference/action data anywhere from a preferred communication channel.

 

[Todd Thompson]

I started scanning once he presented the idea I would throw away my iPhone for a voice activated device.

I completely agree with him that consumers will depend less on individual mobile apps. I am in support of the idea that we could all bail on iOS/Android native apps for web-based experiences...this, though, doesn't require adopting a new device.

His entire demo could have been done either on a voice call or messaging thread.

A much more realistic approach would be to build fluid journeys that can reference/action data anywhere from a preferred communication channel.

Time will tell for sure...
Fun fact regarding ("we could all bail on iOS/Android native apps for web-based experiences")
When the original iphone was released, the intention was to provide developers with a means to create a hybrid web based app (Progressive Web App), which at the time Steve Jobs was showing. He mentioned it would give us access to all the same API's etc they would use on the phone but be web based. Something like 'Web clippings' it was referred to. Can't really recall.

Soon thereafter, apple discovered if they created a store, they'd make a fortune forcing devs to build native apps and then skim 30% for subscriptions/purchases through the apps the devs developed.

So... PWA's fell behind and are more or less the 'gimp in the basement'. Over the years apple has so slowly.. I mean slow.. improved their user experience on PWA's but still intentionally hobble them.

Not sure we will see web based anytime soon.

 

[JDHigginbotham]

Over the years apple has so slowly.. I mean slow.. improved their user experience on PWA's but still intentionally hobble them.

Bastards!

Though, there is still so much that can be done with what is available today. If the basic goal is to replace manual tasks with strong conversational experiences at scale using AI, these experiences can live on trusty old SMS or web chat.

Staying on the adoption of DPAs in Automotive...if there is structured data that can be accessed on a webhook, it can be brought to a channel leveraging AI to activate that data in a conversational manner. That conversation actions the next step.

The above isn't necessarily a brand new idea, but it being seamless to be built and owned by anyone is.

 

[Carsten]

Why isn't this Rabbit just an app on my phone?

Why do I still have keys, and not just an app on my phone?
Why do I even need a wallet? I have my phone.
I hate carry things ...

 

[Carsten]

Where have I seen this rabbit logo / icon before ... it's so familiar ...