AI for E-commerce
Businesses
Our process is pragmatic: identify, clean, and test a workflow that improves customer experience and/or revenue. We’ll be explicit about what works, what doesn’t, and where AI adoption may need additional guardrails.
Join your store data
Find the problem
Approve the change
Your team acts on stock, service and sales problems.
What does this apply to?
We work with bigger ecommerce brands built on Shopify, WooCommerce, or a bespoke tech stack, and look to improve:
Product data: AI-generated product titles, descriptions, attributes. These are generated from your existing, approved product data; we don’t use reviews as evidence.
Search and merchandising: AI-assisted search that understands customer intent, collection pages that can be ranked by new criteria such as availability and profitability.
Checkout and buying agents: This is an emerging space. We’re yet to see many people using AI assistants to buy, but it could affect your business eventually. For now, high-quality product data is the reason to improve this area of your site.
Customer service and returns: AI-assisted customer service that’s grounded in your returns policy, customer orders and product data. Customers will still ask for refunds, exchanges, warranties, and advice on product sizing. AI-assisted responses should make it easier for you to deliver great service and improve satisfaction with the answer.
Retention messaging: We can build, test, and iterate lifecycle email and SMS journeys, grounded in your approved data. We’ll report on our assumptions and provide full visibility into what’s been changed.
How will you build these workflows?
- We’ll start with native solutions where they’re available. We’ll build bespoke solutions where we need to implement specific business rules, sync data across systems, or if there’s no native solution for the feature.
- Agents can be integrated into your current customer service stack. We’ll show you the reasoning behind their actions, which policy/order/carrier tracking/product data they used to respond. We’ll highlight the tricky cases that should be handed to humans, like angry customers or high-value claims.
What the work involves
- Store audit. We’ll take a deep-dive into your ecommerce business, including the webstore, customer service, analytics, email/SMS tools, etc. During this phase we’ll highlight gaps and inconsistencies, gather evidence to understand how they affect customer experience or sales, and raise considerations for consumer law.
- Clean up catalogue data. We’ll create product data updates based on your approved data. Where there’s conflicting data we’ll raise this before publishing. We’ll make sure data is suitable to publish via your channels/feeds.
- Build a revenue loop. We’ll agree a clear focus for the project, whether that’s search performance, lifecycle email revenue, customer service efficiency. We’ll build the first revenue loop for this use-case and test it. We’ll be honest if we don’t see improvement, and explain why.
- Build a learning loop. We’ll create evaluation sets for the use-case to make sure every prompt or knowledge base update is tested, we’ll introduce AI features with rollback and audit trails, and iterate on improvements.
- Throughout this process we’ll consider the implications for the DMCC Act, from displaying prices, incentivised and fake reviews, subscriptions, AI-generated customer messages. This is where we’ll flag what’s allowed, for example incentivised reviews are OK but secretively giving an incentive is not. We’ll point to examples to help you and your legal owner feel confident with sign-off.
How is this different to AI-generated product descriptions?
- We start with an audit to understand your current setup. This is likely to include stores built for blue links, a support inbox filled with "where's my order" emails, a merchandiser manually pinning hero products to the first row of collections. We’ll dig into customer expectations around product information, pricing, reviews, and returns.
- We focus on use-cases that improve customer experience and/or increase revenue. We’ll be honest if AI isn’t the solution. Where there are opportunities for AI, we’ll be explicit about the benefits, limitations, and guardrails we’ve introduced.
- You own the solution. Whether that’s prompts and evals to generate product descriptions, workflows where reviews never stand as evidence for a claim, code to improve collection pages, it’s yours. We’ll use AI via your account with the provider, so there’s no intermediary to worry about.
Get in touch
Book a store audit to find out more.