The Operating System for Commerce Media: Jeffrey Cohen on Amazon Ads’ Rise, Retail Media’s Measurement Reckoning, and the Shift from Dashboards to Agents

<p>Retail media was built on a powerful promise: connect advertising directly to transactions and finally show marketers what their media dollars produced. But as the category matures, a harder question is replacing simple attribution: <strong>Did the advertising actually cause the sale?</strong></p><p>In this episode of <em>Signal &amp; Noise</em>, Rio Longacre and Brett House sit down with <strong>Jeffrey Cohen, Chief Business Development Officer at Skai and former Principal Evangelist at Amazon Ads</strong>, for a wide-ranging conversation about the evolution of commerce media—and what happens as intelligent agents begin taking over work once performed through dashboards, spreadsheets, rules engines, and manual campaign management.</p><p>Jeff had a front-row seat during Amazon Ads’ extraordinary rise. He explains what Amazon got right, from the simplicity of “search, find, buy” to its ability to connect media exposure with actual commerce. But from outside Amazon, he now sees the larger challenge: consumers don’t live inside walled gardens. TikTok can drive an Amazon purchase. Prime Video can influence a Walmart sale. In-store activity affects digital behavior, and digital media influences what happens on the shelf. </p><p>That creates a measurement reckoning. ROAS and last-touch attribution can tell marketers what received credit for a transaction, but not necessarily what <strong>caused</strong> it. The conversation digs into incrementality, cross-channel effects, independent measurement, and why historically siloed media, shopper marketing, trade promotion, and commerce teams increasingly need to operate together.</p><p>Then the discussion moves from <strong>dashboards to agents</strong>.</p><p>Jeff explains how Skai is approaching this transition through Celeste, Skai Studio, MCP connectivity, and agentic workflows. Instead of asking an AI assistant for an insight, marketers can build systems capable of analyzing signals, finding audiences, recommending budgets, creating campaigns, monitoring performance, and escalating decisions to humans for approval.</p><p>The goal, Jeff argues, shouldn't be another marketing <strong>black box</strong>, but a <strong>glass box</strong>: automation where marketers can understand the logic behind decisions, establish guardrails, audit actions, and retain control over critical choices—especially where budgets are concerned.</p><p>One particularly striking finding: Jeff says <strong>45% of the questions asked through