We’ve Been Living in AI Hype for Years
The following is a true story with a fake name. It happened just a few months ago.
Susie’s campaign was crushing it on conversions. The dashboard was green, CPAs and CPMs were down, and leads were up. The media buyer was thrilled with herself; the algorithm was giving her gold stars. Meanwhile, she had over 200 units of last year’s model collecting dust on dealer lots while the ad platform hummed with the newly launched models. So, when she went to deliver her report, instead of the glowing review she expected, she walked into the lion’s den. Her CMO didn’t care about aggregate conversion rates. He cared about clearing inventory before the new models arrived. He cared about dealer relationships. He cared about floorplan financing costs eating into margins.
While the rest of the world is buried in generative AI hype, marketers have been living in machine learning (ML) and gen AI for years. And in many cases, we’ve already been burned. We know what a campaign feels like when it’s optimizing for Google’s metrics instead of our business objectives. We’ve watched predictive analytics make technically correct decisions that miss the strategic point entirely. We tried to write ad headlines in November of 2022 when ChatGPT was released and saw the slop it made on day one.
The disconnect between what the tools optimize for and what the business needs can feel like a chasm too big to cross, but the truth is, we have to go there. The future is now. Peter Drucker reminds us that “the bottleneck is at the top of the bottle.” If you’re leading the team, you must lead through this change or risk being left behind.
At Element Three, we’ve been in the workshop building new tools to serve OEs and their dealers, and we’ve learned some things. I’ll tell you more about that at the end, but first, I want to give you a three-step process for your own AI adoption model.
Lots of people have written about AI Councils and organizational AI adoption. I recommend reading Ethan Mollick’s excellent “Co-Intelligence: Living and Working with AI” for a broader perspective if you haven’t yet.
Build Your Roadmap
Start by dreaming. Get a whiteboard and ask yourself: “What would be great?” Not what’s realistic, not what’s approved in the budget—what would actually make your job better?
Make a long list. Ask different teams what drives them crazy. Ask your dealers what they wish they had. Ask customers what they’re trying to accomplish when they interact with your brand. You’ll be surprised how many of their answers have nothing to do with your product and everything to do with their problems.
Then map it on a 2×2 matrix like this:
High Impact, Low Complexity: For most companies, these are things you can buy off the shelf at a low cost or give to individual team members. If you think about internal tools first, or parts of a process instead of the whole thing, you start to find accelerators.
High Impact, High Complexity: These are north star-type projects that could be a big deal but need to be broken into smaller pieces. You might want a fully automated customer journey system, so you should start with automated follow-up emails for abandoned shopping carts. Build toward the vision in steps that each deliver value.
Low Impact, Low Complexity: Don’t spend much time here, but look for credibility builders. Sometimes a quick win can show momentum. Especially as you pick off the low-hanging fruit in the top right, you can look here. You’ll likely find projects that might not be game-changing, but if it takes two hours to build and saves someone 30 minutes a week, launch.
Low Impact, High Complexity: Don’t go there. The hardest thing you’ll do is put your favorite ideas in this box, because the technology is cool, or you’d really like to see it happen. If you’re honest with yourself, and let ideas go here to die, you’ll save yourself and everyone around you a lot of heartache.
You can do this alone in an hour, but it’s better with a cross-functional team. The service manager sees problems you don’t. The dealer relations team knows pain points you’ve never considered. Get them in a room with a whiteboard. You’ll have your roadmap before lunch.
Pick your Pilot
And call it a pilot. It can be part of a “strategic initiative” or a “digital transformation effort,” but calling it something that indicates you’re testing sets expectations. Pilots don’t need to be home runs to be valuable. That’s why we run them.
Here are a few things to keep in mind:
- Establish your risk tolerance upfront. What happens if this completely, spectacularly, embarrassingly fails? If the answer is “we lose some time and learn something,” you’re good. If the answer is “dealers revolt and sales tank,” you’re not ready. Find a way to do some more controlled testing first.
- Resource it appropriately. Talent, attention, and dollars all matter, but not equally. Attention and interest are huge accelerators in this work because the hype is high. Go for “small team that gives a damn” over “large team that’s sort of interested.”
- Define success criteria before you start. “We’ll know it when we see it” is a scary start. You need specific gates: “If we can generate 10 accurate responses without human intervention,” or “If three dealers use it voluntarily for a month.” Make the finish line real and visible.
- Keep cycles tight. We run 90-day milestones because they align with our quarterly planning cycles. Six weeks to see first results. Three months to kill/keep decision. If you can’t show value and feedback in 90 days, you’re probably well beyond a pilot.
I talked to a technology leader last week who spent months building a feature that no one wanted, and they could have found that out in week two if they had started smaller with better loops. Don’t be that team.
The Five-step Test Loop
Once you’ve locked on your project, here are our five steps:
1. Build a proof of concept. Take whatever you’re building, and as quickly as possible, build a proof of concept. Spend $100 on some subscriptions. Use the free trials. Mock it up with tools like Replit or v0, and don’t worry about perfection. The no-code tools out there can get you something working fast, even if it’s not production ready.
2. Get feedback fast and make it public. Show your ugly baby to people. Tell them it’s a pilot. Tell them it might suck. Then shut up and listen. The feedback that stings is usually the feedback that matters.
Show it to internal stakeholders, prospects, and friends. For us, these conversations have turned into partnership opportunities and sales even before the tool was off the ground.
3. Fix what matters, ignore what doesn’t. Just like in any research, you’ll get 100 pieces of feedback, and ten matter. The skill is knowing which ten. Here’s a hint: If multiple people independently mention the same issue, that’s a signal. If one person has seventeen suggestions, that might be noise.
4. Put it in production and measure. After you’ve established it won’t burn down the building, put it in actual use. Depending on what you’re testing, that might be asking someone to use it to write ad copy for the next few weeks or letting it produce initial versions of reports. Dedicate yourself to the test to see how it performs.
Use your established metrics. If you said success was “dealers use it three times a week,” measure that. Not engagement rates or satisfaction scores or other proxy metrics. Measure what you said you’d measure.
5. Have the courage to kill or commit. This is where most organizations fail. They can’t kill the zombies—pilots that shuffle along, not quite dead but certainly not alive. Every zombie project undermines the next pilot’s credibility.
Get Started
The AI conversation in your organization is happening whether you’re leading it or not. Your dealers are getting pitched. Your team is secretly testing ChatGPT. Your CEO is reading case studies.
So just pick something from your 2×2 matrix in that upper right quadrant. Give yourself 90 days. Build it, test it, measure it. Kill it or scale it.
In five years, nobody will care who was first to test AI. They’ll care who was first to make it work. The distance between those two points isn’t measured in technology or budget. It’s measured in willingness to start.
Want to see what’s working? Reach out, and I’ll show you the pilot we’re running right now—it’s helping dealers and OEMs solve an advertising problem they’ve had for decades, and the production costs have just gone from totally prohibitive to almost free.
