What does AI change for a retailer that is not another recommendation engine?
The back office. Vendor and purchase order handling, markdown and allocation analysis, store communications, and customer service volume. Personalization gets the attention and is usually already bought with the ecommerce platform; the unstaffed work is in merchandising operations and store support, where the hours are and where nobody has looked.
- Increase revenue
- Improve operational efficiency
- Reduce cost
Retail
Personalization is already in the platform
Most retailers asking about AI already have recommendations, and they came with the ecommerce platform. Spending again there is the least differentiated move available.
The work nobody has staffed is in merchandising operations: purchase order exceptions, vendor chasing, markdown decisions argued from incomplete data, and store communications written five times for five formats.
What we build
Vendor correspondence and purchase order exception handling, which is high volume, low risk, and currently done by people you hired for judgment.
Markdown and allocation analysis across stores that genuinely differ, with the reasoning shown so a merchant can disagree with it. The goal is a better argument, not an automated decision.
Store communications drafted once and tailored per format and region, then tracked for acknowledgement. Most chains cannot currently tell you whether a communication was read.
Customer service automation for the deflectable contact types, with escalation designed in.
What we will tell you
If your product data is inconsistent across channels, the merchandising analysis will be wrong in ways that are hard to see. Fixing the data is unglamorous and it comes first.
Typical projects
- Vendor correspondence and purchase order exception handling
- Markdown and allocation analysis across stores that perform differently
- Store communications drafted once and tailored per format and region
- Customer service automation and agent assist for the deflectable contact types
- Product data and attribution maintained across channels
The same work, by business function
How it starts
A three-week assessment scoped to this area: where the hours actually go, what is worth building, and what to fix first. You keep the findings whether or not there is a next part.