Every agentic merchandising conversation I walk into starts in the wrong place. The CFO wants the software cost. The CDO wants the roadmap. Everyone circles the investment like it is the risk.
It is not. The status quo is the risk.
McKinsey put a hard number on the doubt this year. In a new McKinsey survey, 71 percent of merchants say that AI merchandising tools have had limited to no effect on their business so far. Read that in a boardroom and the instinct is to wait. That instinct is the mistake.
That statistic does not measure the technology. It measures deployment discipline. Most of those 71 percent bought a tool, bolted it onto dirty product data, and called it a strategy. The retailers pulling ahead did the unglamorous work first. So let me show you how I actually run the math.
The setup: one real retailer
Take a Tier 2 specialty retailer. Roughly $800M in annual revenue. Around 40,000 active SKUs. Eight buyers. Assortment, pricing, and promotion decisions still run on spreadsheets, vendor emails, and merchant intuition.
McKinsey frames the upside well. Imagine you have 20 analysts for every single buyer. This is the new reality that agentic AI systems, autonomous and goal-driven systems that can plan, act, and learn, are bringing to retail merchandising. Good vision. But a vision does not survive a budget cycle. A CODN model does.
CODN, the Cost of Doing Nothing, flips the entire framing. Instead of asking what the program costs, you calculate what the current run rate costs you every quarter you leave it untouched. You put the status quo on the P&L, where it has been hiding all along.
The worked math: three leaks
I model three leaks first because they are the easiest to defend to a skeptical CFO.
Leak one: excess markdown. This retailer takes markdowns on roughly 30 percent of units. Assume a conservative 1.5 point margin recovery from better-timed assortment and price actions. On $800M, that is real money before you touch anything exotic. Call it $8M to $10M annually left on the table.
Leak two: stockout-driven lost sales. Manual reorder cycles mean the best sellers go dark for days. Even a modest 0.5 percent revenue recovery from tighter inventory sensing is $4M. That is demand that walked to a competitor and did not come back.
Leak three: buyer hours. Eight buyers spending half their week building reports instead of negotiating vendors and shaping assortment. That is not a soft cost. That is your highest-leverage talent doing work an agent does at 3am.
Add the defensible portion of those three and CODN clears $14M a year. The agentic program, done right, costs a fraction of that. Suddenly the investment decision is not close. It was never close. The spreadsheet just hid the losing side.
This is why the 71 percent number is a gift, not a warning. The market is discounting the whole category because most players executed badly. That is your window.
Why most deployments still fail
The technology is ready. In 2026, always-on agentic systems mark a paradigm break: AI agents now autonomously detect trends, competitive moves, pricing anomalies, inventory issues, and customer sentiment continuously. These systems can reprice SKUs, re-balance inventory, or fine-tune promotions dynamically, without waiting for human batch cycles or manual intervention.
So why the graveyard of pilots? Data. Every time. A merchandising agent is only as good as the product information underneath it. If silhouettes, materials, and size ranges are inconsistent, the agent inherits your mess and amplifies it at machine speed.
The retailers getting this right are auditing attributes before they deploy. On April 14, 2026, David’s Bridal joined Shopify’s Agentic Storefronts for ChatGPT and Microsoft Copilot, while also auditing product attributes such as silhouette, neckline, fabric, and size range to make its assortment easier to find across AI-shopping experiences. That is the pattern. Fix the foundation, then unleash the agents.
And the deadline is closing faster than most planning cycles assume. By 2030, analysts project that 25 percent of global e-commerce sales will be enabled by AI agents, and 55 percent of digital consumers will already begin product research using large language model platforms. Your assortment either shows up in that flow or it does not exist.
Here is the forward-deployed truth. The $14M is not a projection you chase. It is a bill you are already paying, quietly, every quarter, in markdowns and stockouts and buyers stuck in spreadsheets. CODN just makes it visible.
The next four quarters will separate the retailers who priced their status quo from the ones who kept pretending it was free. Build the CODN model this week. The decision will make itself.