I can finally tell you what we've been building

You've been reading our newsletter for a while now, and today I finally get to tell you what we’re building.

Today, we're officially launching Lumian, the world’s first AI-native Amazon agency, with $3M in funding from Flybridge, Bowery Capital, and Fika Ventures.

Before any of this, I built and sold a seven-figure eyeglasses brand on Amazon. Along the way I worked with the best agencies in the space, hired in-house teams, and wired together every tool I could get my hands on. None of it solved the actual problem. 

The agencies, the teams, the software stacks were all built for a slower game than the one Amazon is running now.

After the exit, I completed my MBA at Harvard Business School and spent a year deep inside the world of AI agents. What became clear is that AI doesn't just replace the work of running an Amazon business. It changes when and how that work gets done. Continuously, at a level of coverage no human team can sustain on its own.

That's what we built Lumian around. 

One brand manager who actually knows your account. A team of specialists behind them. And AI agents running 24/7 across listings, ads, inventory, and account health, flagging issues the moment they appear.

We're already live across 10+ categories and inbound even outpaced our ability to onboard since beta. We're early, but we're building this for the exact operators this newsletter has always been for.

If Lumian has been useful to you over the months, do check out our launch post. And if you're running a brand yourself and wondering how to implement AI in your business, book a demo here.

Thank you for being here. None of this happens without you.

In this week’s issue:

1/ Amazon launches Scheduled Actions for Rufus

Amazon just added a feature called Scheduled Actions to Rufus, its AI shopping assistant. Instead of only answering live questions, Rufus can now carry out recurring shopping tasks on its own once a shopper sets them up: things like reminding them about a birthday and suggesting a gift, restocking coffee every month, alerting them when a product drops below a target price, or reordering pet food based on pack size. The shopper defines the intent once, and Rufus picks the product and triggers the action on schedule, with no ad auction or second prompt in the loop.

Why it matters:

The product Rufus selects in those scheduled moments is decided by Amazon's catalog logic and not by bids, which means listings that aren't cleanly tagged for occasions, replenishment cycles, gift contexts, pack sizes, or audience fit will quietly get left out of a growing slice of demand. This turns structured attributes and backend data into a real growth lever and it's worth auditing your top SKUs now for this.

2/ California AG exposes Amazon's price parity emails

California's Attorney General unsealed court evidence showing Amazon repeatedly pressured brands like Levi's and Hanes to raise prices on competing sites such as Walmart, Target, Home Depot, Chewy, and Best Buy. The filings document more than 15 exchanges where Amazon flagged the exact price gap, asked the brand to "fix" the lower price elsewhere, and warned of reduced visibility or other consequences on Amazon if they didn't comply. The AG is asking the court to block this practice through a preliminary injunction at a July 23 hearing, while the broader case heads to trial in January 2027.

Why it matters:

Most of these examples involve first-party vendor relationships, but the ruling could reshape what's allowed for everyone, including third-party brands who have quietly seen their Buy Box suppressed after running a lower price on Walmart or their own DTC site. If the injunction goes through, you may have meaningfully more room to use other channels for promotions, clearance, or DTC-exclusive pricing without Amazon retaliating.

3/ Amazon opens China distribution center for US-bound seller inventory 

Amazon opened a new bulk storage facility in Shenzhen, China where sellers manufacturing in the region can store US-bound inventory close to the factory before shipping it to US fulfillment centers through Amazon Global Logistics. Storage at the Shenzhen location runs up to 45% cheaper than Amazon's US-based AWD service, and the combined origin-storage plus cross-border setup can get inventory to US FCs up to seven days faster. Amazon has flagged this as just the start, with expansion planned for other regions of China as well as Europe and Japan.

Why it matters:

For anyone manufacturing in China, this is a real cash flow lever. You can hold inventory near the factory at a meaningful discount versus parking it stateside, then drip it into the US only as demand builds, which softens carrying costs and trims your exposure to tariffs and demand swings during long lead times.

Amazon is now showing two opposing return-rate badges on product detail pages: a green "Customers usually keep this item" badge for listings with low return rates, and a red "Frequently returned item" badge for listings with high ones. The red badge has been live since February 16, 2026, and affected sellers are reporting conversion drops of 20 to 40 percent, with no formal dispute process. The only way to clear it is to bring the underlying return rate down. Rufus is also pulling this data into its AI shopping answers, so even shoppers who never see the badge itself can get steered away when they ask about product quality.

The signal here is worth paying attention to. Return rate has quietly become a compounding metric rather than just a margin one: a lower rate earns the green badge, which lifts conversion, which feeds Amazon's ad relevance scoring, which tends to pull your CPC down on the same budget, and the red badge runs that loop in reverse. Worth pulling your return reports by ASIN this week and splitting the reasons into listing fixes like sizing, photos, and unclear expectations versus product fixes like defects and packaging.

Best,

Robin Lobo

Co-Founder | Lumian