AI for Your Online Store: Where It Pays, and Where It Doesn't
Most retailers I talk to have tried ChatGPT and have a folder of half-finished experiments. I help you work out which two or three uses of AI are worth building, run them as small pilots on real work, and measure what they save or earn.
Where AI pays off in a retailer
These are the six places I look first. Each one is a repeat task with a number attached, which is what makes it easy to test.
Data readiness
AI can only use the data it can read. Product titles and attributes have gaps, customer data sits in the store, the email tool and the ad accounts, and tracking often has not been checked in a year. I start there: clean product data, one agreed definition of a customer and an order, and tracking you can trust. Clean product data also helps search engines and AI assistants understand what you sell.
Email and lifecycle
Subject lines, segment definitions and first drafts of flows are quick wins. The gain comes from testing more variations, more often, on the flows that already earn money: welcome, abandoned cart and win-back.
Customer service
Order status, delivery and returns questions repeat all day. AI can draft answers from your own policies for a person to approve. I have worked on AI-assisted customer service in a live retail business, so I know where it needs a human in the loop.
Reporting and analysis
Weekly trading reports, campaign summaries and questions like "why did conversion drop on Tuesday" can be answered in minutes once the data is connected. The work is in connecting the data and writing down the questions.
Advertising
Meta and Google now use AI to choose audiences and placements, so the inputs decide the result: a clean product feed, working conversion tracking and creative worth testing. AI drafts and varies ad copy quickly. I have run paid media on both platforms, and I focus on measurement so you know whether the spend earned its money.
Being found in Google and AI answers
Buying guides, collection copy and FAQs written from real customer questions help in Google and in AI answers, because both run on the same basics: pages a crawler can read, clear product data and answers to real questions. See AI search readiness.
How I sequence it
No big rollout. Small pilots, a baseline to compare against, and a clear rule for stopping.
- List the workWalk through how the business runs week to week and list the repeat tasks in marketing, service, merchandising and reporting.
- Score and pickScore each task on hours spent, revenue it touches and the cost of getting it wrong. Pick two or three.
- PilotRun each one on real work for two to four weeks, with a named owner in your team.
- MeasureCompare against the baseline: hours saved, conversion, response time, error rate. If a pilot did not move a number, it stops.
- Keep, change or dropKeep what worked, write down how it runs so your team can repeat it, and set a 90-day plan for what comes next.
What I have built, not just advised on
I use these tools myself, so I know what breaks.
At Beer Cartel
Hands-on work on AI-assisted customer service and marketing personalisation in a live retail business, alongside five platform migrations and four email platform implementations.
On this site
This site is set up for AI search: a sitemap, structured data, an llms.txt file and a Markdown copy of every page. The AI Search Readiness Audit checks the same things on your store.
In my own work
An agent system on the Claude API that handles parts of my daily work.
Training
Claude Code in Action (Anthropic, March 2026), plus LinkedIn and Microsoft AI courses (September 2025). These are course completions, not a claim of more.
Options and prices
All prices in AUD. Fixed-price work has a fixed scope, written up before we start.
AI Search Readiness Audit
$1,200
Fixed price
- Written report
- 45-minute walkthrough
- A prioritised list of fixes
Ecommerce and AI Growth Audit
$1,500
Fixed price, one store
- Written report
- 60-minute walkthrough
- Growth and AI opportunities ranked
AI Strategy Sprint
$4,500
Fixed price, two weeks
- AI roadmap for your business
- Two or three prioritised use cases
- A 90-day plan
Monthly consulting
From $3,500
Two days a month, one-month minimum
- Ongoing advice and pilot support
- Monthly review of what is working
- Slack or email between sessions
Questions
Do I need technical staff or a big budget?
No. Most first pilots use tools you already pay for plus a general AI assistant. The audit tells you whether anything needs a developer before you spend on one.
Which AI tools do you work with?
I work mostly with Claude and ChatGPT, and with whatever already sits inside your store platform and email tool. I recommend what fits your store, not what I happen to prefer.
How long until I see a result?
A pilot runs two to four weeks on real work, so you get a measured answer inside a month instead of a strategy document that sits in a drawer.
What if AI is not worth it for us yet?
Then the audit says so. Part of the job is telling you where not to spend money.
What does data readiness mean?
It means your product data, customer data and tracking are clean and consistent enough for AI to use. In practice that is complete product attributes, one agreed definition of a customer and an order, and tracking you have tested. The audit lists what to fix first.
Do I need a data warehouse first?
Usually not. Most stores can start with the reports their platform, Google Analytics and email tool already give them, once those are cleaned up. A warehouse earns its cost when you sell across several channels and have questions no single tool can answer.
Can you build it as well as advise on it?
I build small working tools myself. For bigger builds I write the brief and help you manage the developer or agency doing the work.
Start with a conversation
Thirty minutes, in person in Sydney or on a video call, to talk through your store and where AI might help. No pitch.