I have seen this moment before.
Early in my career, I was the founding engineer at a healthcare technology company. Hospitals and insurers were making complicated payment decisions through processes that depended heavily on people. We built decision engines that automated much of that work. I spent the next ten years as CTO, and the company was eventually acquired by Hearst.
Years later, I encountered a similar challenge at Amazon, on a very different scale.
Amazon was growing rapidly, introducing new businesses and services, and pushing the boundaries of what its technology could support. As VP of Technology, I built AI systems that made complex, real-time decisions across Amazon’s global commerce and fulfillment operations.
Much of that work involved taking ideas that hadn’t been tried before and turning them into systems the business could depend on. We started with small teams, proved what was possible, and then worked through the harder challenge of deploying those systems at Amazon’s scale.
That meant understanding how the business actually operated and making new technology work with existing systems, people, and processes. Some of those capabilities now operate worldwide.
Both experiences taught me something about how industries change.
There comes a point when technology makes it possible to do something fundamentally differently. At first, companies use it to improve the processes they already have. But the bigger opportunity comes when they recognize that the old way of working is no longer necessary.
I believe we’re approaching one of those moments in retail.
And this time, the change is coming from two directions at once.
The way customers discover products is changing
For most of the history of ecommerce, retailers have operated within a relatively familiar model.
Customers discover products through search, advertising, social media, and other channels. They visit a retailer’s website, browse its assortment, compare options, and decide what to buy.
Retailers have spent decades learning how to compete within that model. They’ve invested in customer acquisition, search visibility, merchandising, personalization, and conversion.
Now AI is changing the journey.
A customer looking for running shoes might ask ChatGPT for recommendations instead of starting with a search engine. Someone shopping for a gift might describe the recipient and ask an AI assistant to narrow down the options. A shopper researching a major purchase might rely on AI to compare products before ever visiting a retailer’s website.
These are still early changes, but they matter.
AI is beginning to influence which products customers consider, which brands they discover, and where they decide to shop.
And as personal shopping agents become more capable, some of those decisions will happen without a customer visiting a traditional storefront at all.
For retailers, this introduces a new problem.
They’ve spent years optimizing how to attract customers and convert them once they arrive. Now part of that decision-making is moving into AI systems they don’t control.
A retailer can have a great product, competitive pricing, and an excellent website, yet still miss a customer if its products aren’t considered when an AI assistant makes a recommendation.
That changes what it means to compete for demand.
I don’t believe traditional search or ecommerce websites are going away. But retailers can no longer assume that the methods they’ve relied on to reach customers will be sufficient as AI becomes a bigger part of shopping.
The question is how quickly they learn to compete in this new environment.
Retailers need a different way to respond
The immediate challenge is visibility.
Retailers need to understand how their products appear in AI-driven discovery, where they’re missing opportunities, and what they can do to become more relevant when customers are making purchase decisions.
But visibility alone isn’t enough.
Being recommended by an AI assistant is valuable only if it helps the retailer win a customer.
That means connecting what happens during discovery to what happens afterward: which customers arrive, what they’re looking for, which products they consider, whether they purchase, and what the retailer can learn from the outcome.
Today, much of that work happens across separate tools and teams.
A marketing team identifies demand. Another team works on search visibility. Merchandisers decide how products should be presented. Ecommerce teams optimize the shopping experience. Analytics teams measure what happened.
Each team has useful information, but the decisions connecting them still depend heavily on people.
And the number of opportunities is growing.
A retailer might have thousands of products, many different customer needs, and an expanding set of places where those customers discover what to buy.
No team can continuously evaluate every opportunity, decide which ones matter, make the necessary changes, and learn from every result.
This is where I believe AI can make a much bigger difference.
It can help retailers understand demand as it emerges, decide where they have an opportunity to win, take action across the channels where customers are shopping, and improve based on what actually produces sales.
The opportunity isn’t simply to give retail teams more information. It’s to let intelligent systems do more of the work required to turn customer demand into business.
What my experience taught me
At Amazon, I learned that building an intelligent system is only part of the challenge.
The harder work is making it useful inside a real business.
An AI system can identify an opportunity or recommend a course of action. But unless it understands the business it’s working for, that recommendation may not be particularly useful.
The same is true in retail.
Suppose AI identifies growing demand for a particular kind of product.
That’s useful information. But it doesn’t tell a retailer whether it can win that demand, which products it should put forward, or how to turn that interest into a purchase.
A decision that works for one retailer may be completely wrong for another.
The intelligence has to understand the particular business it’s working for: its products, customers, available inventory, commercial priorities, and what has worked before.
It also has to be able to act.
At Amazon, we learned how difficult it was to make new technology work reliably within existing operations. You couldn’t simply build a model and expect the business to change around it.
You had to understand the systems already in place, the people responsible for the work, and the consequences of getting a decision wrong.
I see the same challenge in retail.
Retailers need more than AI tools that produce recommendations for people to interpret and implement. They need intelligence that can work with their existing systems, make decisions within their priorities, and improve through experience.
That’s the kind of technology I know how to build.
And it’s the opportunity that led us to start Spangle.
Why we started Spangle
My co-founder, Fei Wang, came to this problem from a different direction.
Fei spent twelve years at Amazon, becoming a Principal Engineer. He was a founding engineer on the Alexa team and the founding architect of Amazon’s customer support agent chatbot.
He spent years building conversational AI systems that could understand what customers needed and help solve their problems at enormous scale.
Then he became CTO of Saks OFF 5TH.
He went from building some of the most sophisticated customer-facing AI systems in the world to running technology inside a retail business.
There, he experienced firsthand how difficult it was to connect information across systems, coordinate work between teams, and turn customer and business data into decisions that could actually be executed.
He knew what AI was capable of because he had built it. He also understood why putting that intelligence to work inside a retailer was so difficult.
When Fei and I started talking, we saw an opportunity to bring our experiences together.
I had spent years building AI systems that could make complex business decisions and operate at global scale. Fei understood how to build AI that could interpret customer intent, and he knew the challenges retailers faced from the inside.
We wanted to build intelligence that understood both sides of a commercial decision: what the customer wanted and what made sense for the retailer.
That’s why we started Spangle.
Our early work focused on shopping experiences. We built AI that could understand the context of a customer’s visit, determine which products and information were most relevant, and create an experience around that intent.
As we worked with retailers, we saw that the same intelligence could help much earlier in the customer journey.
It could identify demand worth pursuing, help retailers become more visible across search and AI platforms, bring customers to relevant shopping experiences, and learn from what they actually bought.
That learning could then inform what the retailer did next, across channels.
Today, Spangle puts AI agents to work helping retailers acquire customers, improve shopping experiences, and grow sales as commerce moves into AI.
We’re starting with the parts of retail where the way customers discover and choose products is changing most rapidly.
Over time, I believe the same approach can help retailers make better decisions across more of their business. But the immediate opportunity is clear: help retailers compete for customers in a shopping environment that’s already changing.
The window is open
There’s a lot of attention on AI agents that shop for consumers.
I understand why. An agent that can find products, compare options, and eventually make purchases on someone’s behalf could change how commerce works.
But I think there’s another side of this shift that deserves just as much attention.
What happens to the retailer?
A retailer still needs to understand what customers want, decide which products to offer, make good commercial decisions, and earn the customer’s business.
Those responsibilities don’t disappear when the customer uses an AI agent.
In some ways, they become more important.
That’s one reason we’re building Spangle Halo: to give retailers an AI merchant agent working on their side as consumers increasingly rely on AI agents to shop.
But I see agentic commerce as part of a larger change.
AI is changing how demand is discovered. At the same time, it’s making it possible for retailers to respond to that demand in ways that weren’t practical before.
Those two changes are happening together.
That’s what makes this moment interesting to me.
I’ve seen what happens when technology reaches a point where it can take on work that previously required enormous amounts of human coordination.
The transition takes time. Companies have to experiment. The technology has to earn trust. New systems have to work alongside the people and processes already in place.
But there is a period when the companies that recognize the change early have an opportunity to learn faster than everyone else.
I believe we’re in that period now.
Retailers don’t need to replace everything they’ve built. They do need to start learning how to compete as AI changes where customers discover products and how they decide what to buy.
The opportunity is to build intelligence that understands each retailer’s business, helps it find and win demand, and gets better with every outcome.
Consumers are getting AI working on their side.
We’re building Spangle to put AI to work on the retailer’s side.