Retail organizations make thousands of consequential merchandising decisions every week, yet very few can clearly explain why those decisions were made, what tradeoffs were accepted or whether the outcome achieved the intent.
Ask a merchandising team why a pricing decision was made last quarter. A promotion. A shelf reset. A category expansion.
What evidence drove the decision? What tradeoffs were weighed? What was the expected outcome, and how would you know if it worked?
Often, the reality is: “We did it last year. We tried to do 2% better. We hoped it would work.”
That’s not a people problem. Category managers are operating under extraordinary pressure. Consumer demand is shifting in an instant, competitor pricing adjustments are happening faster than any planning cycle was designed to track, tariff volatility is cascading through supplier relationships and supply chain disruptions are arriving without warning. The tools category managers have been given weren’t built for this environment. They’re being asked to operate with the analytical depth of a stock trader and the reaction time of a market maker, using retrospective spreadsheets designed for annual planning cycles.
So they do what anyone would. They rely on instinct. They lean on what worked before, betting that this year will be just like last year. They get through the day.
We call it a process. We sign off on it in reviews. This isn’t accountability. It’s accountability theater.
Margin erosion behaves like compound interest: the accumulation of hundreds of small decisions compounding over time; a bad shelf placement; a missed promotion window; a pricing decision made without visibility into what a competitor changed two weeks earlier. Individually, none of these impacts appear material. Combined, they create structurally thin margins. I attribute a meaningful share of retail margin performance today to luck rather than deliberate optimization. The underlying tools, processes and planning models were built for a slower, more predictable world than the one we now operate in.
Automation Without Accountability Solves the Wrong Problem
The agentic AI conversation today often focuses on automation. Let the machine handle pricing decisions, manage promotions, optimize shelf resets. AI agents can execute processes 100+ times faster than any person.
But if we automate the current process without fixing the accountability problem underneath it, we haven’t solved anything. We’ve made the same unaccountable decisions faster, at massively greater scale, with no opportunity to catch them before they compound.
Accountability in Retail with AI
Instead of reviewing category performance once each quarter and hoping the plan survived first contact with reality, you have a system that tells you what happened, diagnoses why, surfaces what is likely to happen next and makes a justified recommendation for what to do about it. Every recommendation carries its reasoning. Every action traces back to the strategy and the business outcome you are trying to drive. Accountability stops being a merchant standing up in a Monday morning meeting to read the tea leaves on a judgment call. It becomes continuous, evidence-driven, and built into the way decisions are made every day.
This only works if you design for it from the start, and that is the step many AI solutions gloss over. You need guardrails. You need clear decision roles. You need human oversight at the points that matter. Most importantly, you need a framework that is explicit about when AI makes the call, when a human makes the call, and why.
Where Human Judgment Belongs in an AI-Driven Decision Workflow
AI is genuinely superhuman at processing volume and complexity. No merchant team could analyze billions of signals, identify patterns across thousands of SKUs and optimize across every dimension simultaneously. That work should belong to AI.
But AI does not know what a meaningful business outcome looks like. It does not understand the strategic intent behind a specific promotion. It cannot reason about why you are willing to take a short-term loss on one product to drive traffic to another, or what the brand signal is in a pricing decision that looks like a margin sacrifice on paper. The portfolio logic, the customer relationship you are building, the long-term category position you are protecting: that is human judgment, and it has to stay there.
The retailers who connect those two things deliberately are the ones who win. That means designing systems where AI handles the analytical heavy lifting, humans own the strategic direction and the handoff between them is structured, explicit and traceable — not bolted on afterward as a governance or security exercise, but intentionally designed into the system and workflow itself.
What it Costs to get this Wrong
I talk to organizations who built AI-powered decision systems without building accountability into them. They have systems that generate insights, create recommendations, and in some cases take autonomous actions. But nobody trusts it. There is no framework for who owns the outcome. There is no explanation or audit trail. There is no clear line between what the AI decided and what a merchant approved, and no way to course-correct when the recommendations start drifting from the strategy.
That is an expensive place to end up, and it is entirely avoidable.
The retailers who build accountability in as a design principle from day one end up with something different: faster decisions, yes, but more importantly, decisions that teams will actually act on. Decisions with a clear line back to strategy that a merchant can defend, a CFO can audit, and a board can trust.
AI changes the speed and scale of retail decision-making. Accountability determines whether that speed creates value or amplifies the same unaccountable decisions we’ve always had, just faster. The retailers who treat accountability as a core design requirement, not an afterthought, will be compounding value while everyone else is still debugging.
John Lin is Senior Vice President, Solution Architecture – Retail at SymphonyAI, leading the application of predictive, generative, and agentic AI across merchandising, store operations, and supply chain. With over two decades at the intersection of retail, AI, and hyperscale data platforms, he built and scaled a market-leading retail analytics business through its $500M acquisition. His team drives profit, growth, and operational performance for global retailers and CPGs through applied AI and advanced data platforms.
