I’m Aniket Deosthali—Co-founder & CEO at Spiffy.ai, where we’re helping brands clear the pathway to purchase through turnkey and ultra-safe AI solutions. I love solving hard problems, and currently I’m focused on bridging the gap between how the right marketing can vastly improve how customers discover products and make purchases. This article outlines how we’re helping growth leaders do that.
The way humans make decisions is messy. We deliberate for hours about dinner plans and debate for weeks about vacation destinations. And some of the messiest decisions of all take place on our phones and computers when we shop online.
When was the last time you visited an online store, instantly added something to your cart, and checked out right away? If you’re like most people, the answer is “pretty rarely.”
Shoppers want to weigh their options, get social proof, and figure out whether a product is tailored to their unique needs—all without feeling pushed by an intrusive pop-up or a clueless chatbot.
This chaotic stage between discovery and making a purchase is the “messy middle” of online shopping—and it’s where marketing budgets get the most scrutiny and customers are won and lost. But it’s also one of the most misunderstood aspects of ecommerce. Thousands of brands are turning to AI to eliminate friction in the buyer journey in an effort to meet their sales goals. But just slapping a generic model onto your shopping experience can create more problems than it solves.
In this article, I’ll explore how growth leaders can navigate this messy middle of the shopping experience with the right kind of AI, rather than generic customer service tools.
Here’s what you’ll learn:
- How addressing the messy middle can help you make the most of your marketing dollars
- Why generic LLMs are falling short in solving the messy middle
- Three steps to start gaining a competitive advantage and testing how AI can improve the path to purchase
The Messy Middle of Ecommerce, Explained
The actions someone takes between a trigger and purchase decision-making are nonlinear. There’s a complex web of touchpoints where customers can be won or lost—and that space is what we’ll call the “messy middle” of ecommerce.
As people explore and evaluate a category’s products and brands, they weigh factors such as price, social proof, how quickly they need it, and brand value until they’re ready to make a decision. It would be easy to optimize the messy middle if every shopper’s journey was the same, but that’s not the case. Every single shopper is on their own journey. Different people have different browsing behaviors, levels of knowledge, need for social proof, and more. At the same time, ecommerce brands are under heavy pressure to customize and improve user experiences to ramp up conversion rates, cross-sell, and drive repeat purchases.
There are countless AI tools that promise to solve this challenge. But today’s shoppers are highly sophisticated and can easily sniff out when it feels like they are talking to a robot instead of a human being who understands context and the environment around them.
The goal, then, shouldn’t be to hammer people over the head with cookie-cutter content and generic product recommendations. It should be to get a granular understanding of what each shopper needs and provide them with relevant, supportive, on-brand information to guide their decision-making.
The technology to do that in a scalable way hasn’t existed until recently, which has left many brands taking a “peanut butter” approach to AI: smearing an LLM solution across the end-to-end experience without accounting for shoppers’ individuality.
Why Generic LLMs Can’t Solve the Messy Middle
Assuming generic AI models can solve the messy middle is like assuming a trip to the grocery store can get you a Michelin-star dining experience. The raw materials are there, but it requires considerable guidance to get the results you need.
Almost every ecommerce website has an “AI-powered” chatbot that can answer basic questions about return policies or product specifications. But when it comes to giving hyper-personalized recommendations or crafting thoughtful responses, they can’t perform anywhere close to a seasoned sales representative.
Consider Amazon’s new shopping chatbot, Rufus, which was dubbed “mostly useless” in a scathing Washington Post review. “When the Amazon bot responded to my questions, I usually couldn’t tell why the suggested products were considered the right ones for me. Or, I didn’t feel I could trust the chatbot’s recommendations,” said Shira Ovide. “When so many AI chatbots overpromise and underdeliver, it’s a tax on your time, your attention, and potentially your money.”
Yikes.
Out-of-the-box LLMs don’t know where else a customer has been shopping, what pages they’ve visited on your site, or what’s going to empower them to make a purchase. As we all know, no two purchase journeys are the same. If two people are on the fence about buying a set of golf clubs, one might prefer what their favorite Instagram influencer uses while the other person wants to read a detailed blog post about the features.
Solving the messy middle with AI is 100% possible, but not with a “peanut butter” approach that treats every shopper the same. If we want to deliver better shopping experiences, customers deserve solutions that accommodate the entire mosaic of shopping journeys.
That technology is here and now available. It uses your brand data—such as product pages, social media posts, and brand guidelines—to create recommendations, comparisons, and more to elegantly guide your shoppers to make confident purchase decisions.
Here’s how we do it:
Every Shopper Deserves a Personalized Experience
Generic LLMs are fine for simple back-and-forth exchanges. But that’s table stakes. The real value lies in owning your own model—one that understands your shoppers’ behavior holistically and, from there, engages with them using the same level of nuance as your best sales representative.
That’s where Spiffy’s outcome-oriented models come in. While most consumer-facing AI products rely on prompt engineering (see here for more on that), our hyper-personalized models are trained on millions of unique data points from your brand. Better yet, these models continuously improve as more training data (AKA what’s actually happening on your brand’s website and the interactions customers are having) is accumulated.
TL’DR: You can scale the abilities of a seasoned sales expert instead of replacing them with an everyday chatbot. The proof is in the (digital) pudding. Here’s how Spiffy’s output stacks up against generic models:
In other words: In order for retailers to both scale their business and build enduring relationships with customers, they need AI and LLM tools that act, look, sound, and think like their best salesperson.
Cleaning Up The Mess: 3 Steps to Take Before the End of 2024
If you’re a growth marketer who’s curious about using AI to clean up the messy middle of your buyer journey, here are three things you need to do before the end of this year.
1. Commit to Being in Test Mode
The earlier you start testing AI, the better. Everything happening on your website right now—from customer support inquiries to abandoned carts to product reviews—is high-quality insights that you could be feeding into your model.
If you’re reading this, chances are your competitors are building an AI strategy. The longer you wait, the harder it will be to catch up.
2. Get Picky About the Tools You Buy
Consumer trust is the most important asset you have. Don’t jeopardize it with an AI tool that you don’t fully understand or that you can’t control. If you have any concerns that your model might generate misleading or dangerous outputs, pump the breaks.
Ask yourself: How confident am I about the quality of the results? How easily can I imbue new skills or information? How quickly will I know if something goes haywire? What’s the process for course correcting?
If you’re using a generic LLM, the answers to those questions are probably unclear.
3. Choose the Metric that Gets Mentioned Most in Meetings
Your AI strategy has to be more sophisticated than slapping a chatbot on your website and hoping for the best so you can check it off the CEO’s list of directives. Teams that succeed with AI start by choosing a specific outcome that stakeholders agree upon—whether that’s increasing conversion rates by a certain percentage or decreasing abandoned carts—and then buying a tool that’s oriented toward that outcome.
To solve the messy middle, you have to test AI solutions that drive specific business outcomes rather than overhauling your entire customer journey using peanut butter AI.
At Spiffy, we’re helping growth leaders at major brands clean up the messy middle with hyper-personalized AI that helps improve key metrics like conversion, basket size, AOV, LTV, and brand loyalty.
Want to learn more about Spiffy? Schedule some time to chat with our team.
