• This thread is just the tip of the iceberg.The people ahead of the curve aren't Googling for answers — they're already in here, having the conversations you haven't found yet. DealerRefresh is free.Get the full picture →

The AI Gold Rush: When Everyone Can Suddenly Do Everything

Dan Sayer

Nerd
Staff member
Dec 4, 2009
581
652
Awards
7
First Name
Dan

The AI Gold Rush: When Everyone Can Suddenly Do Everything​

Is AI creating a new era of innovation in automotive retail, or are we heading down a familiar road?

I've been noticing a trend lately, and I'm curious if others in the industry are seeing the same thing.

It seems like everyone can do everything now.

AI vendors, established technology companies, independent developers, dealership employees, and entrepreneurs are all discovering just how quickly they can build software with AI. And suddenly, the conversation has shifted from "Here's what our product does really well" to "Yeah, we can build that too."

Don't get me wrong. I think we're witnessing one of the most exciting periods of technological innovation in our industry's history. The barriers to developing software have dropped dramatically, and that's opening doors that simply weren't available a few years ago. The runway is short.

But I'm also starting to wonder if we're heading down a slippery slope.

Just Because You Can Build It Doesn't Mean You Should​

Some of the best conversations I've had with AI developers lately have been with those who recognize what they don't do.

They understand their lane. They recognize the specific problem they're trying to solve, and they're focused on becoming exceptionally good at solving it.

I find that refreshing.

On the other hand, I've had plenty of conversations where nearly every question about functionality, integrations, reporting, or capabilities is met with some variation of "We can do that."

And maybe they can.

That's the tricky part. With today's AI development tools, it's entirely possible they could create a working version of what you're asking for in a surprisingly short amount of time.

But there's a significant difference between building a feature, developing a product, and supporting an enterprise software solution.

Especially in automotive retail.

The Part That's Easy to Overlook​

Our industry is complicated. Really complicated.

It's not just about creating a CRM, a website, an AI agent, or a reporting dashboard.

New car dealers operate within an ecosystem of manufacturers, DMS providers, CRM platforms, OEM programs, certified vendors, compliance requirements, data integrations, and third-party organizations like Shift Digital.

Something that looks relatively simple on the surface can become considerably more complicated when you factor in OEM requirements, approved integrations, program eligibility, co-op reimbursement, security, and the operational realities of a dealership.

A developer might be able to build a functional replacement for a piece of your technology stack. But can that solution satisfy the requirements of your manufacturers? Will it qualify for applicable OEM programs? What happens to co-op eligibility? How does it handle the integrations, security requirements, and support expectations that dealers take for granted with established providers?

And perhaps most importantly, does the developer even know which questions to ask?

I'm not suggesting established vendors have all of this figured out. Far from it. Many of the frustrations dealers have with existing technology are precisely why these new opportunities exist.

But I'm concerned that we may have two parties entering these conversations who don't fully understand the complexity of what they're agreeing to.

A dealer who doesn't know what it takes to develop and maintain a complete software solution, and a developer who doesn't fully understand what it takes to operate within the franchised automotive retail environment.

That's where things could get interesting. And potentially expensive.

The Fear of Missing Out Is Real​

Here's where I'll admit something.

I have FOMO.

Every day, I see another AI tool, another development platform, another use case, or another person building something that would have required an entire development team just a few years ago.

And I catch myself thinking, "Should I be doing more of this? Should we be building our own solutions? Are we going to fall behind if we don't?"

The possibilities are almost overwhelming.

But like most people working inside a dealership group, I have a day job. Actually, several responsibilities competing for the same hours.

I can't spend every day experimenting with new AI tools or building applications. And yet, there's this underlying feeling that I should be.

But I've recently started looking at that FOMO a little differently.

The development cycle for AI is moving so quickly that the shelf life of FOMO might be 30 days instead of several years. The groundbreaking tool or capability we're afraid of missing today could be commonplace, dramatically improved, or even obsolete a month from now.

Ben Hadley at AutoGenius shared an analogy on a call yesterday that really stuck with me. "AI is like a continuously moving train, but it's still moving slowly enough that we can all jump aboard."

That aligned with my perspective. Maybe the question isn't whether I'm falling behind, but when it makes the most sense to jump on.

I've started to realize that I don't necessarily need to be on every train. I just need to recognize when it's time to get on one.

And perhaps there's an advantage to letting the technology mature a little before committing significant time, money, and resources to something that could look entirely different in another 30 days.

I suspect developers and technology companies are feeling that same pressure, perhaps even more intensely.

When the time required to develop new functionality has shrunk so dramatically, there's a real fear that someone else will build your next big idea before you do.

So instead of focusing on doing one thing exceptionally well, the temptation becomes trying to do everything. Everyone's racing to build the next feature, expand into the next category, or claim the next capability before someone else does.

And now we're hearing conversations that almost sound like, "Why do we need all these vendors? We can just vibe-code our entire technology stack."

Maybe someday that becomes a realistic option.

But I don't think we're doing ourselves any favors by confusing the ability to rapidly develop software with the ability to responsibly replace business-critical systems.

And perhaps that's the irony of all this. The same speed that's creating our fear of missing out might actually be giving us the luxury of waiting.

A Familiar Pattern With a New Twist​

There's another part of this that feels awfully familiar.

I've been in automotive retail for 25 years, and I've watched countless technology companies discover our industry.

They see the revenue generated by dealerships, recognize the amount of money spent on technology and marketing, and identify an opportunity.

And somewhere along the way, a product that might cost a fraction of the price in another industry becomes a substantial monthly subscription because it's being sold to a car dealer.

We've seen it with websites, lead management, digital marketing, equity mining, reputation management, chat tools, and just about every other technology category that's come through our doors.

Now AI is creating an entirely new wave of opportunity.

And I can't help but wonder how many companies are entering automotive because they've identified a legitimate problem they're uniquely qualified to solve, versus how many simply recognize an industry that's historically willing to spend money trying the next big thing.

Let's face it. Dealers will try just about anything if they believe it might sell another car, generate another lead, or improve efficiency and reduce expense.

That's not necessarily a bad thing. It's part of what makes our industry entrepreneurial and adaptable.

But it also makes us an attractive target for companies that are better at selling a vision than delivering a sustainable product.

And when AI makes it easier than ever to demonstrate something impressive, distinguishing between those two becomes even more difficult.

Where Do We Go From Here?​

I'm not advocating that dealers slow down their adoption of AI. Quite the opposite.

I believe AI will fundamentally change how our dealerships operate, how we interact with customers, and how we manage our businesses.

I'm excited about what smaller development teams and independent entrepreneurs can accomplish. In many cases, I think they'll challenge established vendors in ways that are long overdue.

But I also think we need to become more disciplined about separating what's possible from what's practical.

We need to ask better questions, understand the dependencies, and recognize that a compelling demo isn't the same thing as a production-ready solution.

And maybe we need to be a little more skeptical when every company we talk to suddenly claims it can solve every problem we have.

For the developers and vendors reading this, I'd be interested in hearing your perspective as well. Is the pressure to expand your capabilities as intense as it appears from the dealer side? Is specialization becoming harder to justify when everyone around you is promising everything?

And for my fellow dealers, I'm curious:

Are you seeing this same trend?

Are you finding it increasingly difficult to separate legitimate AI innovation from overpromising? Have you encountered situations where a solution looked great in a demo but couldn't navigate the realities of dealership operations, integrations, or OEM requirements?

Or am I being overly cautious, and we're actually witnessing the beginning of a much-needed disruption to the traditional automotive technology model?

I don't pretend to have the answer.

I just have this growing feeling that when everyone believes they can build everything, we might be overlooking what it actually takes to build something worth relying on.

Curious if anyone else feels the same way.
 

The Part That's Easy to Overlook​

Our industry is complicated. Really complicated.

Dan,
AI is so damn good at hallucinating, that a dealer decision maker can give AI complex challenges (outside the decision makers domain), then AI delivers highly conficent and robust piles of shit and our dealer goes all in on the hot pile of :poop:

AI HAS ZERO WISDOM.
That's our job


**Triple 9 = your best player (better than 99.9% of the people on your team).**

Thinking about using AI to build an internal tool? Put your best person on it.

If you’re the decision maker, whoever is the smartest, most experienced, and most business-savvy person in the store needs to own this and report directly to you. That’s your Triple 9. If your smartest person lacks business sense or experience, then *you* lead the project and keep it grounded.

Once you assign your Triple 9, expect them to disappear down rabbit holes for days. They’ll come back with a pile of other things that “need to be fixed or built.” Most of those are forks that lead nowhere fast. Protect the scope.

After your Triple 9 gets their sea legs, have them evaluate vendor tools (both new and existing). Start with small, clear wins — something like “Make the GA4 reports actionable.”
 
Hey Dan - Thoughtful post to read over coffee this morning. You make a ton of observations here worth exploring and I suspect a handful of them will get a decent amount of conversation. Sir, if you were a Vendor, you would have broken this into 16 posts with teasers at the end of each one and called it a series to maximize your reach. ;)

I was fortunate to be on the early cars of the AI train at CarStory. The team was reverse engineering build data and using computer vision to identify missing and inaccurate trim, trans, and features. We built that "point your phone at any car and we'll show you where you can buy it" tool in 2017. There have been many, MANY, days I wish we hadn't been purchased by Vroom. I can only imagine the things that team would have accomplished with Claude code bots, N2N, and LLMs.

Something Dr. Franke and I discussed on a Refresh Friday during that time comes to mind. Dr. Franke had been working in AI since the 1980's. I'll paraphrase because I am not on the same intellectual plane as Dr. Franke... "None of what we are doing today are NEW ideas... we had all of these ideas or derivations of them 20 years ago... we lacked the storage and processing power to execute against them. This is the cycle... we build and iterate until we hit the technology wall... and we enter an AI winter waiting for the hardware to catch up."

This is what I've been thinking about a lot lately. What will be the AI Winter of this cycle? What is the future limiting factor??? I don't know that we can know until we hit it. I do think that there are speedbumps today that may be early warning signs.
  • We have immense storage and processing power (I'm sure that the 3.5 floppy disk felt like that too at some point ;) ), but much of that is currently subsidized.
    • Will the rapid development continue when the costs increase?
    • Will current profitability models hold up for these startups when their operating costs increase as subsidies decrease?
  • Ideas are cheap and plentiful.
    • I don't think this will ever be a limiting factor, especially with Dr. Franke's commentary on the pioneers of these technologies.
  • Over reliance
    • This may be it as counterintuitive as that may sound. I think we are seeing the beginnings of this today.
      • AI audits, we call them "Claudits," are all the rage. Dealers get the AI lead with 18 paragraphs of demands that aren't really relevant from consumers... we get that on the vendor side from well-meaning dealers with website and GA. Most of the demands aren't really relevant either, and some of the reccomendations are really bad.
      • When it takes bots to respond to bots I'm not sure the humans will see the benefits.
  • Apathy
    • I absolutely think this is a risk for AI based communications. Ubiquity isn't an asset of the technology and the differentiation a human provides may become of great value.
      • Raise your hand if you have mashed the "seems like AI slop" button with a sense of righteous indignation. I certainly have.
        • I'll admit it. I can't help but think, "c'mon man, you are a smart guy. I want to hear from you, not a bot disguised as you. I can tell the difference."
        • There is nothing worse in my opinion than "prompt and post." The original post is scholarly, grammatially correct, and linear. They generate an interaction and the response reads like a third grader wrote it. Different tone, different voice, different entity entirely.
  • Inability to "inspect the work"
    • I'll include an anecdotal reference at the bottom.

Here is just one vendor perspective...

I think you are going to see an abnormal number of flashes in the pan and probably a lot fewer technology startup acquisitions. A great idea that birthed a company 10 years ago is likely to be quietly rebuilt by an entrenched provider in the next feature release. Budding tech entrepreneurs reading this, I'd really challenge yourself before you put that lien on the house... "what is truly unique? what is truly proprietary? what has undeniable market-fit and ROI? what about my product or service is going to take more than 20 tokens to rebuild better than my version with no tech debt?... And MOST importantly, HOW am I going to go to market?" GTM in this space is REALLY hard.

I'm going to be an old guy and shake my walking stick at the clouds for a minute... ;)

Experience in this industry is hard won. On the vendor side, we've all seen that guy or gal with a ton of experience outside of the industry that accepts a big title and flames out fast. Not all experience translates to auto. It is different here.

Anecdote:

I think the next AI Winter may be due to an inability and or unwillingness to inspect the work.

This could be another 1,000 words nobody is going to want to read. ;)

I was plagiarized a few weeks ago by an AI content tool. The content of a post I wrote here about 3P in the VLAs was lifted word for word, including a unique analogy. No citation was given. Imagine my surprise when I read a post from a direct competitor on LI that I wrote... literally word for word! I reached out to this person directly and I believe that they were as surprised as I was when I showed them my original from two weeks earlier.

Their AI content tool scraped Refresh, grabbed my post verbatim, and drafted a post they approved and posted as their own. ChatGPT isn't having original thoughts! It is borrowing from a source. It is incumbent upon the user to understand who it is borrowing from, how much it is borrowing, and whether a citation is required. But practically speaking, how are they to do that?

Was this person to have read the 20+ years of posts on Refresh to ensure they were borowing themes not paragraphs? I don't have the answer, but I felt a little of what the artists are complaining about. I'm no artist, that isn't false humility, I just invested a lot of time to write my thoughts. Those thoughts were lifted with a prompt that took a tenth of a second to type. That isn't equitable and unchecked, what benefit is their to the creator in a creator driven economy if the AI steals the material they created? It's all interesting to think about. What is "created?"

I don't think this person did this maliciously, but I'd argue that the use of AI without thresholds without guardrails sure set them up for an embarassing situation if I'd pursued that course. To their credit, they edited the post and tagged me and we both learned something.

Good Stuff Dan! Really glad to see some longform discussion back in the forums. Not everything can be unpacked in 255 characters.
 
Will current profitability models hold up for these startups when their operating costs increase as subsidies decrease?
Thanks, Ryan. My other concern, although maybe it's less of a concern and more of a side-eye, "I know you probably have an exit strategy, so I'm going along for the ride with that expectation" kind of understanding, is that I fully expect the larger, established companies to start acquiring the smaller startups that survive and carve out meaningful market share.

I've never had the stomach for risk that entrepreneurs do, and I genuinely respect those who take that leap. But I've also watched just about every entrepreneur I've known, whether as a vendor or someone I've worked for, eventually take the check when it gets big enough to move them off the hill they once stood on proclaiming, "I'll never sell."

And there's nothing wrong with that. They've earned the opportunity. But from the dealer's perspective, there's a risk in becoming dependent on a solution that could eventually be absorbed by the very companies it set out to disrupt. What happens to the product, pricing, innovation, and support we've come to rely on once that check gets cashed? That's the part I think we need to keep in mind as we embrace all these new startups. Oh, wait, we've already experienced that over and over and over again...
 
  • Like
Reactions: ryan.leslie

The AI Gold Rush: When Everyone Can Suddenly Do Everything​

Is AI creating a new era of innovation in automotive retail, or are we heading down a familiar road?

I've been noticing a trend lately, and I'm curious if others in the industry are seeing the same thing.

It seems like everyone can do everything now.

AI vendors, established technology companies, independent developers, dealership employees, and entrepreneurs are all discovering just how quickly they can build software with AI. And suddenly, the conversation has shifted from "Here's what our product does really well" to "Yeah, we can build that too."

Don't get me wrong. I think we're witnessing one of the most exciting periods of technological innovation in our industry's history. The barriers to developing software have dropped dramatically, and that's opening doors that simply weren't available a few years ago. The runway is short.

But I'm also starting to wonder if we're heading down a slippery slope.

Just Because You Can Build It Doesn't Mean You Should​

Some of the best conversations I've had with AI developers lately have been with those who recognize what they don't do.

They understand their lane. They recognize the specific problem they're trying to solve, and they're focused on becoming exceptionally good at solving it.

I find that refreshing.

On the other hand, I've had plenty of conversations where nearly every question about functionality, integrations, reporting, or capabilities is met with some variation of "We can do that."

And maybe they can.

That's the tricky part. With today's AI development tools, it's entirely possible they could create a working version of what you're asking for in a surprisingly short amount of time.

But there's a significant difference between building a feature, developing a product, and supporting an enterprise software solution.

Especially in automotive retail.

The Part That's Easy to Overlook​

Our industry is complicated. Really complicated.

It's not just about creating a CRM, a website, an AI agent, or a reporting dashboard.

New car dealers operate within an ecosystem of manufacturers, DMS providers, CRM platforms, OEM programs, certified vendors, compliance requirements, data integrations, and third-party organizations like Shift Digital.

Something that looks relatively simple on the surface can become considerably more complicated when you factor in OEM requirements, approved integrations, program eligibility, co-op reimbursement, security, and the operational realities of a dealership.

A developer might be able to build a functional replacement for a piece of your technology stack. But can that solution satisfy the requirements of your manufacturers? Will it qualify for applicable OEM programs? What happens to co-op eligibility? How does it handle the integrations, security requirements, and support expectations that dealers take for granted with established providers?

And perhaps most importantly, does the developer even know which questions to ask?

I'm not suggesting established vendors have all of this figured out. Far from it. Many of the frustrations dealers have with existing technology are precisely why these new opportunities exist.

But I'm concerned that we may have two parties entering these conversations who don't fully understand the complexity of what they're agreeing to.

A dealer who doesn't know what it takes to develop and maintain a complete software solution, and a developer who doesn't fully understand what it takes to operate within the franchised automotive retail environment.

That's where things could get interesting. And potentially expensive.

The Fear of Missing Out Is Real​

Here's where I'll admit something.

I have FOMO.

Every day, I see another AI tool, another development platform, another use case, or another person building something that would have required an entire development team just a few years ago.

And I catch myself thinking, "Should I be doing more of this? Should we be building our own solutions? Are we going to fall behind if we don't?"

The possibilities are almost overwhelming.

But like most people working inside a dealership group, I have a day job. Actually, several responsibilities competing for the same hours.

I can't spend every day experimenting with new AI tools or building applications. And yet, there's this underlying feeling that I should be.

But I've recently started looking at that FOMO a little differently.

The development cycle for AI is moving so quickly that the shelf life of FOMO might be 30 days instead of several years. The groundbreaking tool or capability we're afraid of missing today could be commonplace, dramatically improved, or even obsolete a month from now.

Ben Hadley at AutoGenius shared an analogy on a call yesterday that really stuck with me. "AI is like a continuously moving train, but it's still moving slowly enough that we can all jump aboard."

That aligned with my perspective. Maybe the question isn't whether I'm falling behind, but when it makes the most sense to jump on.

I've started to realize that I don't necessarily need to be on every train. I just need to recognize when it's time to get on one.

And perhaps there's an advantage to letting the technology mature a little before committing significant time, money, and resources to something that could look entirely different in another 30 days.

I suspect developers and technology companies are feeling that same pressure, perhaps even more intensely.

When the time required to develop new functionality has shrunk so dramatically, there's a real fear that someone else will build your next big idea before you do.

So instead of focusing on doing one thing exceptionally well, the temptation becomes trying to do everything. Everyone's racing to build the next feature, expand into the next category, or claim the next capability before someone else does.

And now we're hearing conversations that almost sound like, "Why do we need all these vendors? We can just vibe-code our entire technology stack."

Maybe someday that becomes a realistic option.

But I don't think we're doing ourselves any favors by confusing the ability to rapidly develop software with the ability to responsibly replace business-critical systems.

And perhaps that's the irony of all this. The same speed that's creating our fear of missing out might actually be giving us the luxury of waiting.

A Familiar Pattern With a New Twist​

There's another part of this that feels awfully familiar.

I've been in automotive retail for 25 years, and I've watched countless technology companies discover our industry.

They see the revenue generated by dealerships, recognize the amount of money spent on technology and marketing, and identify an opportunity.

And somewhere along the way, a product that might cost a fraction of the price in another industry becomes a substantial monthly subscription because it's being sold to a car dealer.

We've seen it with websites, lead management, digital marketing, equity mining, reputation management, chat tools, and just about every other technology category that's come through our doors.

Now AI is creating an entirely new wave of opportunity.

And I can't help but wonder how many companies are entering automotive because they've identified a legitimate problem they're uniquely qualified to solve, versus how many simply recognize an industry that's historically willing to spend money trying the next big thing.

Let's face it. Dealers will try just about anything if they believe it might sell another car, generate another lead, or improve efficiency and reduce expense.

That's not necessarily a bad thing. It's part of what makes our industry entrepreneurial and adaptable.

But it also makes us an attractive target for companies that are better at selling a vision than delivering a sustainable product.

And when AI makes it easier than ever to demonstrate something impressive, distinguishing between those two becomes even more difficult.

Where Do We Go From Here?​

I'm not advocating that dealers slow down their adoption of AI. Quite the opposite.

I believe AI will fundamentally change how our dealerships operate, how we interact with customers, and how we manage our businesses.

I'm excited about what smaller development teams and independent entrepreneurs can accomplish. In many cases, I think they'll challenge established vendors in ways that are long overdue.

But I also think we need to become more disciplined about separating what's possible from what's practical.

We need to ask better questions, understand the dependencies, and recognize that a compelling demo isn't the same thing as a production-ready solution.

And maybe we need to be a little more skeptical when every company we talk to suddenly claims it can solve every problem we have.

For the developers and vendors reading this, I'd be interested in hearing your perspective as well. Is the pressure to expand your capabilities as intense as it appears from the dealer side? Is specialization becoming harder to justify when everyone around you is promising everything?

And for my fellow dealers, I'm curious:

Are you seeing this same trend?

Are you finding it increasingly difficult to separate legitimate AI innovation from overpromising? Have you encountered situations where a solution looked great in a demo but couldn't navigate the realities of dealership operations, integrations, or OEM requirements?

Or am I being overly cautious, and we're actually witnessing the beginning of a much-needed disruption to the traditional automotive technology model?

I don't pretend to have the answer.

I just have this growing feeling that when everyone believes they can build everything, we might be overlooking what it actually takes to build something worth relying on.

Curious if anyone else feels the same way.
There's a lot of great insight and wisdom in this!! Thanks for the post and well said. I couldn't agree more and I've been a dev for the last 10 years.

Something that I've learned and been telling people lately:
1. There is a lot of noise and static in AI every single day. About 1% of it is helpful.
2. Humanities tools will keep changing and getting better. What won't change is our need for people to interface with people. The greatest skill you can build right now is relationships, learning sales, and learning how to build trust human to human.

My 2 cents but I believe in that.
 

✨ AI Highlights

Automotive retail professionals are debating whether the AI-driven surge in software development represents genuine innovation or a familiar boom-bust cycle. Dan Sayer observes that lowered barriers to building AI tools have created a 'everyone can do everything' dynamic, raising questions about startup sustainability and acquisition outcomes. The key insight from the thread is that AI lacks real-world wisdom, meaning dealers need their sharpest, most experienced people guiding AI projects to avoid confidently wrong outcomes.

Replies Views 5 169 Started Last Reply