If you’re new, welcome! If you’re a regular, thanks for being here as we navigate this AI shift in eCommerce. I share what I’m testing and what’s actually moving the needle. Where’s the vault you ask? Every previous edition is saved here.

In this edition (shortcuts):
Alexa Is Now in Your SMS (US Only)💥
Claude Moves Into Your Google Docs 🧰
MCP vs API, Explained With a Grocery Run
Voices from Thought Leaders
News and Events Worth Following

Alexa Is Now in Your SMS (US Only)

Did you get a text from Amazon this week?

I did. And I spent a minute deciding whether to report it as spam.

It was real. Alexa for Shopping now lives in your Messages app, and you can search, compare, add to cart and buy without opening Amazon at all.

Amazon turned it on for a small group on November 6. On November 13 they opened it to everyone in the US.

I switched it on the same evening. It took about 20 seconds. Here is the whole thing, screen by screen.

Step 1: Find the sign-up page

First, go to this link on your mobile phone: amazon.com/text-alexa-shopping

Open the Amazon app on your phone and go to the "Shop by text" page. The headline reads "Shop by text? Now you can."

Tap the yellow Text Alexa button.

The sign-up page inside the Amazon app. One yellow button.

Step 2: Agree and text

A sheet slides up with your phone number already filled in (the one on your Amazon account).

Read it. Two lines matter.

🔶 Alexa Shopping can place, change and cancel orders using your saved payment methods and addresses. This is a shopping agent with your card on file, not a chatbot.

🔶 It works over SMS and RCS. SMS comes from the short code 23606. RCS messages show up as "Alexa Shopping". On my iPhone the thread was labelled RCS.

Tap Agree and text.

The consent sheet. Note the line about placing orders with your saved payment methods.

Step 3: You are connected

A green bar confirms it: "Connected to your number. Check your text messages for a message from Alexa."

The button changes to Go to chat. Tap it.

Connected. The page now offers Go to chat, and a Disconnect link if you change your mind.

Step 4: Say hi

Messages opens on a new thread with a contact called Alexa Shopping.

I typed "Hey Alexa!" and hit send.

The new Alexa Shopping contact in Messages. Nothing to install.

Step 5: Alexa replies

The first message back is a menu of four starting points:

🔶 Reorder essentials

🔶 Start my grocery list

🔶 Snap my fridge to shop

🔶 See what others ask

Alexa's first reply, with four suggested starting points. The Report Spam button is what Apple adds under any new business thread.

That third one is the tell. Take a photo of your fridge, text it, and Alexa builds the reorder.

Amazon's post lists what else you can do from the thread: send a photo to find similar products, track a price for up to a year, change an order you already placed ("add broccoli to my last order"), hand off a to-do list with several items in one text, and set standing preferences like organic only.

It also says a request started over text will carry over to the Amazon app, Amazon.com and Alexa devices, and back. That part is still rolling out.

Why this matters for sellers

Here is the part I keep turning over.

Amazon says more than 400 million customers used Alexa for Shopping in the last 12 months. And the number using its agentic features (the ones where Alexa acts, not just answers) has grown more than eight times since January.

Now that agent sits in the one app nobody uninstalls.

Think about what a text thread does not have.

No search results page. No row of sponsored products. No "customers also bought." No scrolling.

A customer texts "running low on vitamin D" and Alexa comes back with a pick. Maybe two. The agent does the comparing, and the customer approves.

Nobody outside Amazon knows yet how the agent chooses. But we know what it reads: your title, your bullets, your reviews, your price history, your stock. The same inputs Rufus / Alexa has been reading for a year.

So the work has not changed. It has just become harder to skip.

🔶 Say what the product is in plain language. An agent summarizing to a text bubble cannot use a keyword-stuffed title.

🔶 Answer the questions a shopper would text. "Is it unscented?" "Does it fit a 2019 model?" If the answer is not in the listing, the agent cannot give it.

🔶 Watch your price history. Alexa will quote a year of it on request. A deep discount followed by a hike is now one text away from the customer.

🔶 Stay in stock. An agent that says "that one is unavailable, here is a similar one" is a competitor you cannot outbid.

I wrote about this shift in January in The Invisible Storefront. Back then the agent was a browser plugin a few early adopters ran. Now it is a contact in your phone, and Amazon is the one who put it there.

Try it this week

If you sell on Amazon, sign up as a customer first. Text Alexa for your own product by its plain description, not its brand name.

See what comes back. See whether it is you.

Hit reply and tell me what Alexa picked. I am collecting the answers for a follow-up.

Know anyone who wants to work with an agency that has AI and tech in their DNA? Point them to [email protected].

COOL TOOLS

Claude Moves Into Your Google Docs 🧰

Most of my week lives in Google Sheets, Docs and Slides.

And most of my AI work used to mean copying a table out of a Sheet, pasting it into Claude, then pasting the answer back.

That back and forth just got shorter.

Anthropic launched Claude for Google Workspace: one install that puts Claude inside Docs, Sheets and Slides.

It reads the file you have open. And it edits that file in place.

What It Does in Each App

🔶 Sheets: writes formulas, builds pivot tables and native charts, adds tabs. For messier jobs like joining two tables or cleaning data, it runs the work in Python and writes the result back into the sheet.

🔶 Docs: fixes a sentence or restyles a heading without touching the formatting around it. For bigger rewrites it shows suggestion cards you apply or dismiss.

🔶 Slides: builds slides that match your deck's existing layouts and theme, then checks for overlapping elements, anything hanging off the slide, and hard to read text.

Why It's Different

It is not a separate, stripped down Claude.

The sidebar uses the same connectors and skills you already use in Claude. So if you have taught Claude how your team formats a monthly report, that skill comes along into the Sheet.

Why this matters: picture your search term report open in Sheets. You ask for a pivot by campaign and a chart of wasted spend, and it appears in the file you were already working in. Nothing to paste back.

How to Get It

  1. Install Claude for Google Workspace from the Google Workspace Marketplace.

  2. Open any Doc, Sheet or Slide.

  3. Go to Extensions > Claude > Open Claude.

Pro Tip: leave the default setting, "Ask before edits," on while you learn it. Every change shows up as a card and waits for your OK. Switch to "Accept all edits" once you trust it on a file.

The Catch

🔶 It is in beta. Available on all paid Claude plans, but expect rough edges.

🔶 It sees what you can see. Claude's access follows your Google sharing permissions, so a file you can't open stays closed to it too.

🔶 Team and Enterprise plans: an owner has to switch on the Google connectors before you can create or edit Google files from a Claude chat.

Who Is This For?

Anyone whose real work happens in Google files. Think search term reports in Sheets, listing copy drafts in Docs, and the monthly brand review deck in Slides.

Tried Claude inside a Sheet yet? Hit reply and tell me the first thing you asked it to build.

NERD BYTES

MCP vs API, Explained With a Grocery Run

Every AI agent you connect to a tool uses one of two doors.

An API is the tool's own door. It is how software talks to software: a fixed set of commands like "get order", "update price". Precise, but the AI has to be told it exists.

MCP is the manual. It is the standard that lets an AI discover a tool's commands and call them in plain language. Plug in Slack, GitHub, your Sheets, and the agent knows what each one can do.

Sounds perfect. Here is the catch.

The agent reads every manual before you say a word. Cloudflare's API has over 2,500 commands. As MCP tools, that manual is about 1.17 million tokens. More than most models can even hold.

Then it calls home for every step.

A short I watched this week put it well. Send a friend to the store with MCP and he grabs the milk, then calls you. "What next?" Eggs. Calls again. Bread. Calls again. Forty items, forty calls, and he reads you every label.

Watch another reel on why MCP (in its current form) sucks and why.

That is direct tool calling, and it is how most agents use MCP today.

The fix is called Code Mode. Instead of one call per step, the AI writes the whole shopping trip as a short script, runs it once in a sandbox, and sends back only the answer.

🔶 Cloudflare rebuilt its MCP server as two tools, search and execute. About 1,000 tokens instead of 1.17 million.

🔶 Anthropic measured one workflow drop from 150,000 tokens to 2,000.

🔶 Block ships the same pattern in goose.

So MCP vs API is the wrong fight. MCP tells the agent what is possible. Code lets it do forty things in one trip.

The quiet advantage: connect only the tools the job needs, and ask your agent to write the script instead of narrating the steps.

NEWS AND EVENTS WORTH FOLLOWING

What happens when 18 seasoned speakers get together on one hackathon? They spill nuggets of hard-won wisdom based on years of experience, failures and successes. And you get ALL of these from the comfort of your couch in a distilled format that is easy to action right away.

When: November 18th, 1PM EST
Signup: https://lnkd.in/gyZbdk-G

I am especially excited for this because it is PPC Ninja's 6th birthday as well 🎁 🎊

Other News 📡

🔶 ChatGPT Wallet lets the agent check out for you OpenAI launched it on November 13 for a limited group of US Pro subscribers. You add a card once, the agent pays with a token, and neither the agent nor the store sees your card number. OpenAI also published a playbook for sites that want agent orders to go through instead of getting rejected.

🔶 Amazon drafts a "World Product Graph" Reuters saw an internal plan to collect data on nearly every product sold anywhere, including by crawling other retailers, so that outside shopping agents end up sending people to Amazon. Read next to the main feature above: Amazon wants to be both the agent and the catalog.

🔶 Review Requests are live in Amazon Ads The new ad unit asks verified buyers for a one tap rating from busy pages like the homepage. Amazon says early tests brought about three times the weekly ratings. The same post dates Branded Conversations for January 2027: a sponsored, labelled answer inside Alexa for Shopping, fed by the brand. That is the ad slot inside the agent.

🔶 Claude Fable 5.2 and Mythos 5.2 arrive Anthropic shipped its new top models on November 9 at a lower price than 5.1: 8 dollars per million input tokens and 40 per million output. Fable 5.2 is on every platform today; Mythos 5.2 stays limited to vetted organizations.

🔶 Claude joins your meetings On November 10 Anthropic added Claude Meetings (beta, Team and Enterprise, Google Meet first), plus Claude Whiteboards and Claude Sheets in beta on paid plans. Claude Sheets is Anthropic's own spreadsheet, separate from the Google Sheets add-on in Cool Tools above.

🔶 Jassy: Alexa for Shopping will be the primary agent Amazon's CEO said it on the October 29 earnings call, and Amazon has blocked Meta's Muse agent from its site while Walmart and Shopify let outside agents in. Two different bets on who owns the shopper.

We hope you liked this edition of the AI for E-Commerce Newsletter! Hit reply and let us know what you think! Thank you for being a subscriber! Know anyone who might be interested to receive this newsletter? Share it with them and they will thank you for it! 😃 Ritu

Reply

Avatar

or to participate