Industry
E-commerce & Retail
An assistant may soon sit between you and your customer.
Product discovery is moving into AI interfaces, which puts the retailer at risk of becoming a line item in someone else's recommendation rather than a brand with a relationship.
The friction
Where the hours actually go.
Support volume scales with orders
Where is it, can I change it, can I return it — predictable, repetitive, and staffed as though it were not.
Catalogue data is thin
Sparse attributes and marketing-copy descriptions are poorly understood by the systems now doing the recommending.
Abandoned carts recovered generically
One template for every reason a basket was left.
Merchandising decisions lag the data
Reporting arrives weekly for decisions that are daily.
The opportunity
What we would build, and in what order.
Sequenced by return, not by how interesting the technology is.
- 01Customer Operations
Resolve order questions end to end
Status, address changes, returns initiation and refunds-within-policy handled conversationally, with genuine exceptions escalated warm.
- 02AI Search Visibility
Enrich the catalogue for machine discovery
Structured attributes and genuinely descriptive content so both search and AI assistants can match your products to real intent.
- 03Workflow Automation
Recover baskets by reason
Different follow-up for a shipping-cost drop-off than for a payment failure, because they are different problems.
- 04Business Intelligence
Put merchandising data in reach
Margin by SKU, return rates and stock risk answerable in plain language rather than a weekly export.
Sector constraints
What we design around in this industry.
These are not caveats added at the end. They shape the scope from the first workshop.
- Refunds beyond policy thresholds require human approval
- Payment details are never handled conversationally
- Consumer-rights obligations — returns windows, cancellation rights — are enforced as rules, not left to a model
The system
What this looks like running in e-commerce & retail.
A worked example, sized for a business of this shape.
- Orders this month
- 24,800
- Baskets classified
- 6,412
- Recovered
- 1,184
by reason
| Abandonment reason | Baskets | Recovered | Action | |
|---|---|---|---|---|
| Shipping cost at checkout | 2,140 | 486 | Threshold reminder | Automated |
| Payment failure | 890 | 401 | Retry link | Automated |
| Comparison shopping | 1,180 | 72 | No discount sent | Margin protected |
| LM-2210 Canvas Tote | — | — | Margin 11.2%, below floor | Merchandiser |
Refunds beyond policy require human approval. Consumer-rights obligations are enforced as rules rather than left to interpretation.
Playbooks
Detailed implementations for e-commerce & retail.
Each one carries a concrete scenario, what gets built, the metric that tells you it worked, and the boundary that stays with a person.
An assistant may soon sit between you and your customer.
E-commerce & Retail · what is actually changing
Questions from this sector
Fair challenges.
Should we be worried about AI shopping assistants?
Worth taking seriously rather than panicking about. The concrete, no-regrets actions are the same ones that help today: make your catalogue machine-legible, and own the post-purchase relationship so the customer has a reason to come back directly.
We are on Shopify. Is this overkill?
For a small catalogue with manageable volume, often yes — apps will cover it and we will tell you so. It becomes worth engineering when volume, SKU count or channel complexity outgrows what configuration can handle.
Is this your business?
The assessment takes eleven questions and scores where your opportunity actually concentrates. It runs in your browser and shows its arithmetic.
Find my AI opportunities