AI Commerce Lab
Field notes from the AI commerce frontier.
Engineering perspectives, decision frameworks, and post-mortems from the intersection of AI and commerce. Written by the people shipping the code.
Featured
What we'd read first this month.
All posts
115 of 115 posts
Forecasting Holiday Demand When Your History Lies
Sales history records what you did, not what customers wanted. Label promos as regressors, censor-flag stockout zeros, borrow history for cold-start SKUs, reconcile the hierarchy, backtest on last peak per-slice, and score judgmental overrides after the season.
Inventory Is a Stream, Not a Table
A quantity column oversells in the gaps between updates. Model inventory as an append-only event log with a projection per consumer, and ATP, allocation, and finance each read a view they can rebuild, while oversell and undersell become measurable projection lag.
UCP Is an Integration Contract, Not a Traffic Channel
Salesforce put Commerce Cloud checkout inside Google Search on September 15, with GA in October. The protocol underneath is public and versioned. Reading it as a marketing channel misses the real work: a catalog contract, a checkout API, and a liability model your engineering org now owns.
The Human in the Loop Is a Role, Not a Checkbox
A working loop is a staffed role: a named reviewer with domain ownership, a review surface built for verification, an explicit split between gating and sampling, triggers that tighten and loosen it, and a verdict wire back into the eval suite.
The Peak-Season Code Freeze Is a Symptom, Not a Strategy
A blanket peak-season freeze prices every deploy as equally dangerous, which is only rational when you cannot tell them apart. Harden rollback drills, canary analysis, feature flags, and dependency pinning by October, then freeze only the genuinely irreversible changes.
Inheriting a Codebase You Did Not Write: The First Two Weeks
Two weeks to observability, not to improvements. What to read first in an inherited codebase, what to baseline before touching anything, the 5 stabilization items to land, and an evidence-based test for rescue versus rewrite.
Returns and Refunds Are an Engineering Problem
A return is a long-running distributed transaction that moves goods one way and money the other. Policy as code, classification as a routing problem, the 5-state refund machine, and an explicit line between what to automate and what stays human.
Evals Before Agents: The Regression Suite Is What Makes an AI Feature Shippable
An AI feature without a regression suite is a demo with a deploy pipeline. What a golden set looks like for real retail workflows, the 4 grader types in cost order, the 3 kinds of drift, and where the human approval gate belongs.
One Event Backbone Beats Point-to-Point Integration
With 6 systems there are 15 possible pairs, and every new system you add costs another handful of point-to-point links. A backbone changes the slope: the Nth system costs 1 connection. Here are the source-of-truth rules, cadence tiers, retry and dead-letter mechanics that make that hold in production.
The Performance Work That Actually Moves Revenue
A lab score describes a fast laptop. p75 field data describes the customer you are losing. Which Core Web Vital maps to which shopping moment, why LCP is a delivery problem before it is an image problem, and how to hold the gains with a budget that fails the build.
MACH-Aligned Without Being MACH-Certified: What Composable Actually Costs
MACH certification is a vendor membership programme, not a property of your architecture. Composable commerce is worth the money for the right retailer, and the difference is whether the team budgeted for the seams: optimistic concurrency, version conflicts, eventual consistency, and the operational surface you inherit.
Why AI Pilots Die Before Production
The demo works, the stakeholders are pleased, and 7 months later it is quietly switched off. The model is almost never the reason. Five failure modes account for most of it: no evals, no data contracts, no cost ceiling, guardrails scoped too late, and no approval path.
Browse by topic
What we write about
Get insights delivered.
Field notes from the engineering frontier of AI commerce — direct to your inbox, monthly. No spam, no AI-written filler.