How does Oh My Cream's Head of E-commerce use AI in her daily work?

The context

Keridwen Pierret is Head of E-commerce & Digital Projects at Oh My Cream. It's a busy period and the digital team is small. That is exactly when AI helps her most. Her tools: Claude first, the MCPs of her tools (Shopify, Meta…), and Shopify Sidekick to build a report or a customer segment in one sentence.

This is not a one-off initiative. At Oh My Cream, AI is driven by senior management: the whole company is encouraged to automate its most time-consuming tasks with AI, and every team reports on how it uses it. The question isn't just “are you using it?”, but “what does it save you?”. That framework explains the rest: Keridwen measures her gains task by task.

If your company doesn't allow these tools (yet), read on to see what is possible.

1. Reporting: cross-referencing data, not just compiling it

Plenty of teams already automate their reporting. What's interesting at Oh My Cream is what they get out of it.

• Cross-referencing sources. Claude pulls from Meta, the BI tool, Shopify and Google Analytics, then cross-references everything. That's where it shines: it runs comparisons that would take ages by hand, and flags inconsistencies, for example between Shopify and GA sessions.
• Making up for expertise you don't have. Meta isn't usually part of Keridwen's remit. A skill analyses each campaign and suggests an action (raise the budget on this campaign, and so on). The team then adds the business context.
• Working around limits. GA doesn't connect directly: the team automates an export to Google Sheets, which Claude then reads.

The result: at least two hours saved per person per week, and Monday morning shifts from filling in spreadsheets to analysis, and above all to action.

Same logic for one-off questions, like “how many orders for this limited edition contain a single product?”. Before, an export and some sorting, sometimes tedious. Now, one question.

2. The catalogue without copy-paste

Oh My Cream adds 5 to 20 product pages a week, up to a hundred for a big launch, across two sites. Claude now fills in the pages from the copywriting document, through the Shopify MCP.

Same for bulk changes: adding a gift offer to a brand's 40 product pages, in French and English, goes through a skill. And for the 404s that were hurting SEO, Claude produces the redirects file.

3. Testing before calling in the agency

Keridwen doesn't code, but she prototypes with Claude. A “new shade” label on makeup product pages: coded in half an hour, A/B tested, approved, then built by the agency.

Before, most ideas were built with the agency, then tested. Now, the agency only builds what has already proven itself. “We're more rational about what we spend: we put live features we know are useful.”

4. Everything outside the core job

This is the part that surprised me most. A Head of E-commerce spends a good share of her time hiring, managing people and handling suppliers. That's where Keridwen finds her least expected gains.

Hiring faster. She is currently hiring several people at once. Claude analyses the CVs, prepares an interview guide tailored to each candidate and a tracking sheet she doesn't have to update.

Managing suppliers. A supplier billed by the hour sends dozens of lines of activity every month. Claude sorts them by topic, so she can track her budget project by project. Not to police them, she points out, but to manage her budgets better. Time saved: about a day and a half a month.

Making the team's work easier. Take a newsletter that is almost identical for each store, bar three details. Instead of editing the code by hand, a skill asks for the ID, the visual and the name, then generates clean code. When someone flags a painful task, her reflex: suggest turning it into a skill, or help them build it.

Her advice

1. Add “do you have an MCP?” to your tool selection criteria. At Oh My Cream, every tool gets asked. Without an MCP, it's copy-paste, and the Claude browser extension is heavier and less effective.
2. Measure the time saved, task by task. Keridwen keeps a running inventory. That's what lets her put a number on it and report to senior management.
3. Ask Claude to explain its work. On an analysis, she checks how it reached the result. On code, she has each part explained to her: she knows what she is putting live.
4. Turn every repetitive team task into a skill. Someone non-technical can run it, and human error disappears.

What's next?

Her dream: a real personal assistant.

Did you enjoy this format? Would you like us to open up another e-commerce leader's toolbox? Reply to the newsletter and let us know. We read every reply.

Gautier
CMO @ Agentic Commerce Club