A women's travel brand on Shopify, building luggage for how women actually travel. Three of its AI employees are live and doing real work. Here's what each one does.
Three AI employees, live for Nori
In three weeks, Shotgun shipped about 120 storefront changes, the kind of list an agency would stretch across three to four months, spanning every layer of the site: content, product pages, data, apps, campaigns, speed and SEO.
Nori is a small, founder-led team. And a Shopify storefront is never finished: a badge to add, a page that loads slow, a campaign that needs to go live, a product layout that isn't converting. Something always needs doing.
Every one of those used to mean the same thing. Find a developer or an agency, write a brief, wait in a queue, review, send it back. So the small things piled up, the bigger things slipped, and the founder became the bottleneck for her own website.
Not tweaks. Real storefront engineering across content, product pages, product data, apps and campaigns. Briefed and queued through a freelancer or agency one at a time, the same list is realistically three to four months of waiting. Here is some of what shipped.
Nori's product images sat in a plain stack. Shotgun rebuilt the gallery as a true carousel: swipe gestures, arrow controls, full keyboard navigation, and a lightbox that stays in sync with the main image as you move through it.
A recommendation carousel on every product page, wired to Shopify's own complementary-products setting. Nori curates exactly what shows, per product, by drag-and-drop in the Shopify admin. No code, full control.
A return-window badge on every product page that reads the product type and shows the right policy by itself: 100 days for luggage, 30 for everything else. No manual tagging, nothing to keep in sync.
Specifications rebuilt across every bundle SKU, the Set of 4, 5 and 6. Dimensions normalised to one consistent format across the whole site, and weight pulled out cleanly from dimensions on 18 bundle products. Slow, exacting work.
"Hot off the Press", a press section built into the homepage and the About page: six publications, from VC Circle to Silicon India, each with an editorial pull-quote.
Twenty-two posts taken end to end onto the Traveller's Notes blog: drafted, copy cleaned, images placed, published. Plus a long-form ingredient feature, and a returns policy rewritten from scratch.
And more on top: a sticky add-to-cart bar, an announcement-bar flicker fixed, the review widget restyled, the WhatsApp button repositioned, wheels renamed across the whole store, a Mother's Day collection page.
A few of the asks, in her own words
"Create a Summer Favourites collection, add three products, put a homepage banner linking to it, and a 15% code for that collection. Expires May 10, once per customer, minimum cart ₹2,000."
"Add a 'Complete the Set' carousel on every product page that I can curate myself, per product, from the Shopify admin."
"Put a returns badge on every product page that shows 100 days for luggage and 30 for everything else, automatically."
"Create a Shipping & Returns page, write the policy, and add it to the footer."
From brief to live. One message in Slack: "when someone adds the Cabin Wheelie, in any of its three colours, show a free-gift banner in the cart, and take it away on its own if they remove the bag."
What shipped was a cart that knows what's in it. Add any Cabin Wheelie colourway and the checkout drawer surfaces the gift; pull it out and the banner clears itself. Built and tested on mobile and desktop, wired to the live campaign. A developer scopes this as four to five days, briefed and queued. Shotgun shipped it in about two hours, with a preview to approve before it touched the live site.
A storefront should be a brand's growth engine. For Nori, it had quietly become a dependency instead. Every change waited on an agency, and that wait taught the team to play it safe: fewer experiments, smaller bets, only the work clearly worth the hold-up.
Now the website moves as fast as the team can think, across every layer: speed, UX, SEO, merchandising. They respond to what customers need the day they notice it, and run far more experiments, because each one is cheap to try and quick to learn from. Faster iteration drives growth, growth makes room for more iteration, and the storefront that used to hold Nori back has started to compound.
In 60 days the team put 968 questions to Hermione, reached decisions roughly 20x faster, and doubled revenue while holding ROAS steady. Nori's numbers, once spread across four tools, now answer back in plain English.
The business was running. The data was not moving with it. Shopify held the orders. Meta held the spend. GoKwik held the checkout funnel. Google Analytics held the traffic.
Answering one question, "why did conversion drop this week", meant logging into all four, exporting from each, and stitching it together by hand. By the time the answer arrived, the week was half gone. The team was reacting. The founder was guessing.
The foundation came first: a single layer unifying Shopify, Meta, GoKwik and Google Analytics, with consistent definitions so the same question always returns the same answer.
Hermione is an AI analyst that sits on top of that layer. Anyone on the Nori team asks a question in plain English, across any of those sources, and gets a real answer in seconds. No exports, no stitching, no waiting on the one person who knows where the numbers live.
She works inside the tools the team already uses, so there was nothing new to learn and no dashboard to remember to open.
The real lever isn't the dashboard. It's that the team can now think with something that genuinely knows their business. That took real work on top of the data layer: teaching the analyst the business itself, what each product is, what every campaign is trying to do, which numbers actually matter, and what "good" looks like for Nori. That is the difference between a chatbot pointed at a database and an analyst who understands the company. So when the team plans, corrects, or jams on a new idea, it answers like an operator who has sat in every meeting, not a chart they still have to read.
A real question the team asked
"If we want to ramp sales to a big step-change month, with most of it coming from Meta, what should I focus on? Build me a few ideas, and back each one with data so I know what's actually worth trying."
Hermione rebuilt the last month's economics from the ground up and came back with two findings the team hadn't seen. First, the recent growth had been almost entirely average order value, not volume: people were spending more per order, but the number of orders had barely moved, so the current playbook was near its ceiling. Second, it followed the paid traffic into checkout and caught a hero product converting almost nobody at the checkout step despite real ad spend behind it, money pouring into a page that couldn't close. It even pushed back on the premise, flagging that leaning most of the target on a single channel was riskier than the plan assumed. Then it laid out six ranked plays, each with an expected impact and a way to test it. A strategy session, in one answer.
The point of the layer is not what gets deployed. It is what the team picks up and runs with once it's there.
"Should we restock this SKU?" "Is this campaign worth scaling?" Questions that used to wait for an analyst now get answered in seconds, all day.
A team member uploaded stock data, asked for forecasts, got them, and built it into the monthly stocking routine. The brand stopped guessing how much to order.
With the data unified, the team started spotting fulfilment discrepancies and operational leaks it could not see before. Time and money recovered.
The team built its own agent on top of the layer. It reads ad performance and proposes new creative angles to test. A workflow nobody specced, running on the same foundation.
The data layer is set up once. What the team builds on top of it never stops.
Nori went from near-zero organic presence on Google to page-one rankings, and from invisible in AI search to being cited as a source inside Google's AI Overviews, on the exact queries buyers use to choose luggage.
Ask an AI assistant where to buy cabin luggage in India and you get a confident shortlist. Nori wasn't on it, around 2% visibility, effectively invisible at the moment a customer is deciding. And it wasn't winning the other side of search either: organically, Google sent Nori's content almost no traffic at all. Worse, where AI did mention Nori, it was repeating wrong information pulled from Nori's own website. On the two surfaces where buyers now make decisions, Nori was either absent or misrepresented.
Shakespeare doesn't start with keywords. It starts by learning the brand, not from a brand deck but from the founder: what Nori actually is, who it's for, and what's genuinely true about the product that no competitor can claim. That understanding is the raw material for everything after.
Then comes the intelligence work. It maps the real questions buyers ask an AI on the way to a luggage decision, "best cabin luggage India", "Mokobara review", "Mokobara vs Samsonite", and audits, query by query, how Nori shows up in each: where it's missing, where it's outgunned, where the AI is repeating something untrue. That audit is what surfaced Nori's own site working against it.
Those gaps get organised into content pillars: discovery questions, head-to-head comparisons, feature and trust questions, real use-cases. The work becomes a deliberate map of where Nori needs to exist in the AI's mind, not a pile of blog posts. For each pillar, Shakespeare studies the competition honestly: how the category leaders are described, what's true about each, and where the real opening is.
Only then does it draft, one fact-checked piece per gap, in Nori's voice, written to be quoted rather than just ranked: a direct answer up top, exact specs and comparisons the AI can lift word for word, and honest coverage of rivals. Where Nori can't credibly be the hero of a question, it's the narrator: the most trustworthy independent answer on that question, hosted on Nori's own blog. The output isn't content marketing. It's a brand being deliberately engineered into the answer.
It's working on both surfaces. The content that used to bring almost no organic traffic now ranks on page one of Google, in the top five on high-intent buying queries, under the sponsored block. And on those same queries, Nori's blog is now cited as a source inside Google's AI Overview.
For a young brand up against names many times its size, becoming the source the AI quotes is the hard part, and Nori is now there. Visibility in AI search has climbed from 2% to 17%, the content now drives 12% of the site's traffic, and with 25 articles live and more shipping, a channel that barely existed is becoming one of the biggest.
Each employee was largely built using Claude, then shaped around Nori's own data, context, and the way the team works, which is what makes it useful instead of generic. How you reach it depends on what's being solved: sometimes it's an MCP connector the team talks to inside Claude or Codex, sometimes a web app. And because it's built for the brand, it works with whichever AI they prefer, Claude, Codex, or others.
For the first time, I'm not the bottleneck. The website list that used to sit for weeks now ships in hours, I can ask anything about the business and get a real, data-backed answer in minutes instead of guessing, and we're finally showing up when customers ask AI what luggage to buy. It's handed me back real time and lifted a huge load off the team. Honestly, it feels like having an elite team right beside me.
Fifteen minutes, no pitch. Tell us what's slowing your business down, and we'll show you what we'd do about it.
Talk to us