The agent gets a wallet

The Signal

On September 10, Amazon Ads and OpenAI opened a pilot that lets select U.S. advertisers extend campaigns already running in Amazon DSP into the ChatGPT app, placing sponsored text and image units beneath ChatGPT responses. It runs as a managed service — Amazon Ads teams help build campaigns and supply 'context hints' to match ads to topics and conversations, while OpenAI retains control of placement — with inventory available on both a cost-per-click and CPM basis. Delta Vacations is the first named brand, and it is the first time buyers in Amazon's DSP can reach a surface Amazon does not own.

Why It Matters

ChatGPT reached roughly 900 million weekly users with essentially no ad monetization; this deal hands its ramp to the ad platform with the deepest commerce data in the market, and reporting pegs OpenAI's ad ambitions at roughly $2.5 billion in revenue. For advertisers, retail-media-adjacent budgets are now one click from conversational AI inventory — no new buying platform required — and it signals how AI-search monetization will likely be intermediated: through existing DSPs with commerce signals, not standalone AI ad platforms.

The Opportunity

Brands already active in Amazon DSP can test AI-assistant placements this quarter with existing plumbing, and early inventory in a new channel is historically underpriced before auctions thicken. The questions that decide whether this becomes a real line item: what commerce signals Amazon exposes for targeting, and whether reporting bridges ChatGPT exposure to purchase outcomes. If closed-loop measurement lands on conversational inventory, the search budget split between Google and AI assistants gets renegotiated in 2027 planning.

The Signal

Three announcements in three days put payment rails under AI shopping agents. Meta launched Muse on September 8 — a personal AI agent for U.S. adults that can make purchases on the user's behalf, checking out through Link by Stripe at more than one million businesses using single-use virtual cards, in tiers from free to $20 and $100 a month. On September 9, Mastercard introduced Agent Connect — one integration through which merchants, AI agents, platforms, and payment providers can transact, with Global Payments, Samsung, and Trip.com at launch — plus an expanded Agent Suite for Merchants and a commerce-agent blueprint built with Anthropic. And at Goldman Sachs' Communacopia conference on September 8, Visa CEO Ryan McInerney named the constraint: 'We are seeing adoption for shopping, but not yet for autonomous payments.' Visa's survey data says roughly three-quarters of consumers don't trust agents to pay autonomously — but 61% would if Visa were involved.

Why It Matters

The agentic funnel is splitting: discovery and cart-building are moving to AI agents faster than payment authorization, and the card networks are positioning themselves — not the AI platforms — as the trust broker that closes the gap. For merchants, the integration question is now concrete: Muse-style agents arrive carrying Stripe rails, Mastercard is selling a single pipe to every agent surface, and whoever owns the agent-to-merchant connection will influence which products agents see and surface.

The Opportunity

Merchants should get catalogs, pricing, and availability agent-legible now and evaluate the one-pipe integrations — the cost of hedging across protocols just dropped, and being invisible to agents is the new out-of-stock. Brands should watch what converts through Link-powered agent checkout, because purchase protections and instant checkout remove the friction that made agent commerce theoretical. The tell for the next two quarters: whether merchants adopt network-brokered connections faster than AI platforms can build direct ones.

The Signal

Kroger's September 11 second-quarter report was a study in contrast: identical sales without fuel rose just 0.2%, and the company cut its full-year identical-sales guidance to 0.2%–0.8% (from 1%–2%) — while holding its adjusted EPS guidance, on the strength of adjusted e-commerce sales up roughly 20% and Kroger Precision Marketing profit up 24%. Adjusted EPS came in at $1.09 on operating profit of $971 million. The same week, Walgreens committed to a second attempt at in-store media: digital ad and education screens from Looma across 1,200 stores starting in October, reaching the full footprint by early 2027.

Why It Matters

This is the cleanest data point yet that alternative profit pools are carrying grocery: a retailer holding its profit outlook while cutting its sales outlook is telling you where the margin now comes from. And Walgreens re-entering in-store media — after the Cooler Screens failure — shows RMN investment continuing even at retailers under strain, this time with shelf-adjacent placement and full-funnel measurement as the pitch.

The Opportunity

Retailers leaning on ad profit to hold guidance need advertiser dollars more than ever — that is negotiating leverage for brands heading into 2027 joint business planning, and it predicts continued inventory expansion and measurement investment from KPM regardless of comp softness. In-store screens at Walgreens scale add a measurable reach layer for shopper budgets in health and beauty. Watch whether Kroger starts disclosing more ad-business detail as the gap between merchandising and media performance widens — the disclosure itself becomes the investor story.

The Signal

Deloitte's holiday forecast, released September 9, projects November–January retail sales up 4.0%–4.8% to $1.70–$1.71 trillion — accelerating from last season — with e-commerce growing 7.5%–8.4% to roughly $316–$319 billion, underpinned by disposable-income growth of 4.5%–5.2%. Salesforce's parallel predictions describe a 'K-shaped' season with lower- and middle-income shoppers trading down, and project 20% of holiday e-commerce traffic will originate from AI agents. Then Friday delivered the other half of the picture: the University of Michigan's preliminary September sentiment index fell to 47.8 against 51.0 expected — the weakest since May's record low, with year-ahead inflation expectations jumping to 4.6% — and the August CPI showed headline inflation holding at 3.4% on a gas surge while grocery prices went flat for a second straight month, up just 2.2% year over year.

Why It Matters

Both halves are true at once: income growth funds an accelerating topline, while sentiment near record lows dictates how it gets spent — deals, brand-switching, and trading across retailers at every income level, per Deloitte's own survey work. Digital growing roughly twice as fast as total retail is the cleanest justification yet for shifting incremental holiday budget into retail media, and the AI-agent traffic projection means product data has to be legible to intermediaries that read structured feeds, not display ads.

The Opportunity

Plan to a bifurcated consumer: resilient high-income spend and aggressively value-driven everyone else, which means price transparency, value guarantees, and promo-led conversion tactics over full-price brand plays. Early deal events will capture anxious budgets before wallets thin. And if even a fraction of Salesforce's 20% agent-traffic projection materializes, feed quality, pricing accuracy, and availability data become holiday-critical merchandising work — this is the first season where being unreadable to an agent costs real revenue.

The Signal

Bloomberg reported September 11 that food companies including Conagra, Campbell's, and Clorox have announced or signaled price increases as tariffs and input costs squeeze margins: Conagra told retailers it will raise prices across Healthy Choice, Banquet, Marie Callender's, Blue Bonnet, and Birds Eye lines effective late September, citing sustained increases in input costs including packaging and ingredients, while McCormick expects pricing to bolster its second half. Campbell's, presenting at Barclays' consumer conference during the week, detailed the math behind the increases it announced with its fiscal Q4 report: 4%–5% on roughly 60% of the portfolio, against an assumed 1.5x price elasticity — planning, explicitly, to trade volume for margin — with costs expected up 5%–6% including double-digit logistics inflation. General Mills, at the same conference September 8, reaffirmed its outlook instead; its Q1 report lands September 23.

Why It Matters

The industry is repricing into Q4 at the exact moment official grocery inflation went to zero — store prices were flat for a second straight month in the August CPI — which means widening gaps versus private label just as holiday planning locks and consumers are primed to switch. Retailers will arbitrate who gets to pass through cost, and a 1.5x elasticity assumption is a public admission that volume will leave: the defense is promo intensity, which lands in trade budgets and retail media.

The Opportunity

Frozen and center-store become conquesting territory as list prices diverge: brands holding price gain a rare shelf-price advantage they can amplify with sponsored placement, and private label gets its best comparison backdrop in years. Watch unit volumes in the weeks after Conagra's late-September increases hit — if elasticity runs hotter than the models, the promo spiral deepens and Q4 trade budgets get rewritten mid-season. General Mills' September 23 print is the next elasticity data drop, and the first test of whether holding price is the smarter trade.