Ahead of the Curve: A Rare CEO Who Isn’t an AI Fanboy When It Comes to Local Enterprise Marketing

An interview with Lastmile CEO and Co-Founder, Michael Carini, by Local Search SME, Miriam Ellis

Introduction by Miriam Ellis

Local search marketing predictions are hard to make but here’s one I’ll stand by on the basis of having analyzed the industry for over two decades: within the next few years, all the well-funded hype over AI will have dwindled and, if your large multi-location enterprise invested heavily in a marketing platform because of its AI wow factor, you’ll start looking for a new partner.

Or, you can write a different narrative for your brand by getting ahead of this curve now. Instead of wondering where all your marketing dollars went, you can choose a platform that isn’t going to try to sell you on the “magic” of AI. I personally know (and consult with) just a handful of CEOs in the digital marketing space who have refused to run with the herd when it comes to artificial intelligence, and Michael Carini is one of them.

In today’s interview, Michael will say the quiet part out loud: that AI isn’t all it's cracked up to be, and that, in fact, it represents a ton of risks to your brand. If you’re already feeling jaded by claims of AI’s greatness, prepare to be refreshed.

‍

Miriam:

Michael, let’s start with your overall take on AI and why it’s problematic in the enterprise marketing scenario.

Michael: While the technology doubtless holds great promise, my unpopular opinion is that today it’s only product ready for enterprise business use cases in some limited and specific ways. Putting today’s inconsistent and error-prone AI up as a viable replacement for critical enterprise-scale business processes is often disingenuous at best, and at worst, simply delusional.

Last year, AI startups received a record amount of almost $50B in fundraising. How do I know? Google’s AI Overview feature told me. But, as it turns out, the AI Overview got it wrong:

The $50B dollar number doesn’t actually come from Crunchbase. It comes from a site I’ve never heard of (edgeDelta.com?), citing a Crunchbase news article, citing a dataset behind a paywall that the original author didn’t bother to verify. An old-fashioned manual check on a source I trust (CB Insights), puts the number closer to $40B, while Stanford’s HAI group puts the number closer to $23B.

Do I care what the accurate number is? Not really. Lived experience has already made me painfully aware of the fad-chasing crisis in the venture capital sector. What I do care about is that this well documented trainwreck is the best AI product we could get out of one of the world’s most prominent technology companies.

So if Google can’t get it right, why should an enterprise scale business trust anyone else to, be it a well-funded startup or mid-tier technology company?

They shouldn’t.

But it’s hard to find a tech vendor these days who will speak truthfully to potential clients about AI because everyone is feeling an artificial market pressure to look hip. Private exchanges in boardrooms about the massive problems with AI don’t come to light in client conversations because SaaS and marketing platform brands are afraid of looking unsophisticated if they aren’t hyping and pitching AI.

Miriam:

Let’s dig into specific problems with AI. Michael, you’ve been partnering with some of the largest multi-location enterprises in North America for many years. What is your take on the risks AI poses to customer experience and brand reputation?

Michael: AI interactions are often horrible for customers because they’re not interacting with an entity that’s actually empowered to solve their problems.Ultimately, this is also highly damaging to brands. Why? Because feedback is a gift. When customers bother to communicate a bad experience or problem via a review, chat or call, they do so with the hope that your company will take the feedback seriously and, if possible, work to correct their underlying issue.

Unfortunately, the best today’s AI chat and response bots can do is say “sorry” and hallucinate a promise no one at the company knows about. This fundamentally violates a customer’s trust and robs the brand of an essential piece of insight on their product or service in a way that no amount of “aggregate rollup insight data” can replicate.

While I’m sure there are some clever use cases out there that use AI to tag a customer problem and appropriately escalate it to a person, almost all of the examples I’ve seen in the wild generally end in the same level of frustration and muttered curses that automated customer service interactions do (except now they take longer).

On a philosophical note, if your enterprise doesn't care enough about your customers to have an empowered employee respond to them, why would you bother to respond at all? And why should the customer care enough to give you feedback, hoping for resolution that earns their loyalty? Why shouldn’t an unhappy customer who realizes they’re being “handled” by AI instead of respectfully served by trained staff at your brand just walk away?

Miriam:

What happens to enterprise quality control in the AI scenario?

Michael: The best brands are maniacal about branding quality, creativity and consistency in a way AI can’t replicate. Great brands hire great people who have an inimitable creative spark and obsessive eye for detail. They agonize over minute details like single word choices, specific shades of orange, and everything in between. As painful and messy as the associated creative process might be, the results are undeniably impressive – great brands become synonymous with a common brand voice, aesthetic, and set of values in every medium where they appear.

AI fundamentally lacks the creativity, vision, or critical eye needed to take content from good to great. So while it might have limited applications in certain low-stakes scenarios, it will never replace the human marketers enforcing brand quality and honing a creative vision.

One claim I’ve seen is that AI is good at helping overworked brand marketers ideate drafts. I can see how that can be a legitimate use case, but it’s also indicative that perhaps enterprise priorities have become lopsided when the creatives in the organization are being overworked instead of carefully supported and rewarded for their essential contributions to brand goals. Creativity and authenticity should be seen as imperilled when your best writers – people who entered their field because they love to write – are turning to machines for ideas. It creates a factory scenario of low-quality inputs and outputs. It’s not a recipe for enterprises that need to shine.

Miriam:

Brand liability surrounding AI is very murky at the moment. What is your take on this for large enterprises?

Michael: AI is a legal nightmare. Does a company owe a customer for failing to deliver on a promise it has no idea it’s even made? Does a company have liability for missing clear signs of abuse or crimes in a customer interaction? What happens when a customer is injured due to inaccurate instructions provided by AI (anyone remember Apple maps sending drivers into lakes)? Does a company retain full ownership of content created by an AI it licensed that was trained on data that wasn’t properly licensed?

Even putting the issue of hallucinations (AI generating nonsense) or prompt injection (getting the AI to do something outside its purported scope) aside, these are just a few of the types of questions AI raises that don’t have good established answers yet. Entrusting the integrity of your brand to an unpredictable and unaccountable machine that is still in its infancy opens up an entirely new attack surface that no one is yet equipped to defend. Large brands aren’t being told this by the AI-fueled platforms they’re contracting with, and it’s a very serious concern.

Miriam:

Is AI up to enterprise standards of consistency or reliability?

Michael: Definitely not. Most Enterprise software is held to extremely stringent service level agreements (SLAs) mandating several “9s” of uptime and reliability. According to GPT itself, its answers are only accurate around 70% of the time.

At Lastmile, we’ve confirmed this in our own testing across large sets of data, where AI inexplicably fails to perform a basic task such as tagging a review by concept as much as 25% of the time. While some parties might find this acceptable for low-stakes applications in which even a high failure rate is better than doing the task manually, it shouldn’t be considered a catch-all replacement for the work of intelligent humans. Not if your brand values its customers and reputation.

So if all AI can be trusted for is to inconsistently assist in low-stakes tasks, it’s hard to argue that it's much more than a distraction from better-developed technology and processes that demonstrably deliver brand growth.

Miriam:

I see signs that AI is being used to “paper over” cracks created by bad processes and tech. Would you agree?

Michael: Unfortunately, yes. I won’t name names, but this question immediately brings to mind an industry peer in our local digital marketing space that has gone “all in” on AI. Their latest offering is an AI chatbot, usable across multiple local web properties. It’s being sold as a solution that answers consumer questions about local businesses, including questions about open hours, inventory, special offers, etc.

This sounds great until you look at the underlying data model powering it. The questions the bot can answer are limited to a rudimentary set of information about a location, such as its name, hours, address, and some basic related attributes, plus extremely basic product and service text entries. In other words, all information that users can already obtain extremely quickly and easily in existing platforms like Google Maps and store locators.

So, there’s nothing new to see here. Why the hype? Why should an enterprise be excited about this and invest in it, because someone has stuck the “AI” label on the product to make it look ultra-modern?

Instead of addressing the underlying issue most customers face – not enough rich local content to answer key questions needed to make a local purchase decision – the AI Chatbot repackages the same content in a less convenient and much more error-prone interface.

Given all of the other potential issues with AI interfaces such as hallucinations, false promises, off-brand responses, etc., smarter enterprises will see beyond the fads and instead choose to invest in better content and search functionality if they want to improve their ability to answer customer questions at scale.

Miriam:

Finally, what about claims that AI is a competitive-difference-maker for the enterprises that invest in it? Will it actually help them stand out from the crowd?

Michael: I’m afraid not. AI features have no moat. The hidden truth is that most technology vendors' AI features are little more than a prompt wrapper and fancy UX over a generic call to a GPT API. This means that, in almost every case, the AI features being offered to you by a potential vendor are not unique. Any strategy you build on top of these features has no moat. Any competitor can and will replicate what you do - and quickly.

To prove the point, we decided to run the experiment of adding the full complement of AI features advertised by our own competitors to our local review management product. This included on-demand summaries, auto-tagging, AI auto response and draft responses. The work took two developers six weeks.

As proud as I am of our great developers, the experiment succeeded in proving to me that AI-driven features are comically easy to replicate and no amount of investment in AI will differentiate our product from the competition. AI is a broad river to a sea of sameness – not a secret gateway to brand stardom.

If decision-makers at enterprises get anything from this interview, I hope it’s the courage to look beyond the multi-billion-dollar AI hype and demand solutions that help them excel at connecting with their customers.

Shortcuts are risks. Service is golden.

In Conclusion

If you are currently in the process of investigating a new local search marketing platform partner for a large enterprise and would like to run the AI claims you are hearing from vendors past a third-party, you’re welcome to contact Lastmile. Our clientele consists entirely of major multi-location brands, including T-Mobile, U.S. Navy Recruiting and Hyatt Hotels and we will be happy to share the facts of what we’ve learned about most current AI products on the market.

We may or may not be the ideal partner for your needs, but at the very least, a chat with us could save your enterprise from investing in a product that puts your quality control and customer service standards at risk.

‍