1 Walmart's Sparky proved agents can move the basket
Walmart's Q1 earnings call gave the agentic-commerce story its first hard number set. Sparky weekly actives more than doubled in a quarter, units purchased through it quadrupled, and Sparky shoppers spend roughly 35% more per order than non-users. CEO John Furner used the phrase "becoming AI native" and put it next to a 37% jump in global ad revenue. Two days later, every retailer's investor deck has a new comparison to make.
The signal isn't that an AI chatbot works — it's that Walmart is now reporting agent KPIs alongside revenue. Once a number gets onto an earnings call, it becomes a target every other CEO has to answer to. Expect Target, Kroger, Best Buy and the Amazon-adjacent world to ship visible "shopping assistant" surfaces by Q3 just to be able to talk about them. Plan your retail media and PDP work for a world where the buyer arrives in a chat, not a search box.
2 Bayer put a gatekeeper agent on the buying agent
At Digiday's Programmatic Marketing Summit this week, marketers from Bayer and Kelly Scott Madison outlined the live guardrails they're putting around AI agents in programmatic: spend caps, data anonymisation requirements, and internal "gatekeeper" agents that check the working agent against brand rules before campaigns move. The fear is concrete — an autonomous system overspending, leaking data, or making opaque optimisation calls — and the answer is policy in code.
This is the most under-covered shift of the agentic moment. The interesting org chart isn't "AI agent reports to media director." It's "gatekeeper AI agent reports to brand safety, working AI agent reports to performance, humans on exceptions." If your team is piloting agentic media buying without those two roles separated, you don't have a pilot — you have a runaway budget waiting to happen.
3 X's Creator Connect made xAI the matchmaker
X launched Creator Connect, an AI-powered service that uses xAI to match brands with creators based on campaign goals, audience overlap, and live trend signals on the platform. X runs outreach, content, and distribution as a managed workflow — closer to an agency engagement than a self-serve marketplace — with brand approval at each step. Early tests included a premium laptop matched with tech creators who also follow F1, and a horror-film studio matched with niche horror creators.
It's a sharp positioning move: X turned its weakest asset (perception of brand safety) into a paid service offering (managed creator workflow). The risk for buyers is structural — the same AI choosing your creator is owned by the platform selling your media. Treat the recommendation list the way a smart media planner treats any house-favored allocation: useful, biased, worth a second source.
4 PubMatic and DoubleVerify shipped agents into the seat
Two of the more useful agentic launches this week were quietly inside existing dashboards. PubMatic's Detailed Reasoning Agent added Chat-to-Data queries inside its programmatic dashboards, with scorecards for how products rank inside Alexa, Sparky, and ChatGPT. DoubleVerify launched content verification for Meta Threads, evaluating brand-adjacent posts at the post level with AI-driven risk flagging.
Both moves matter because they're not new products — they're agentic layers stitched into the workflow you already run. That's where AI gets sticky. "Open a separate AI tool" is friction; "answer my question in the seat I'm already in" is just better software. Vendors that bolt LLMs onto their existing dashboards win the workflow. Vendors that ask you to learn a new one don't.
5 Pattern Pi and Creatify Agent put commerce on the auto-tune
Pattern launched Pattern Intelligence (Pi), an AI-driven commerce platform for pricing, content and visibility across Amazon, Walmart and TikTok Shop. Creatify launched Creatify Agent, a video creative system trained on past performance to generate ads aligned with what's working on TikTok, Meta and YouTube this week. Both productise a job that used to live with an agency.
The pattern across both: AI that ingests platform-winning patterns and writes back to them. The honest tradeoff: you get more competent commerce and creative, faster, and a lot of it will look like everyone else's because everyone's optimising on the same signal. The defence is upstream — distinctive supply, distinctive brand voice, distinctive product. The agent can multiply that, but it cannot invent it.
6 Cannes Lions added an AI Craft category — the prize finally caught up
The Cannes Lions Festival of Creativity, running June 22–26 this year, confirmed an AI Craft subcategory inside the Creative Data Lions, with 39 finalists on the 2026 shortlist across categories including Data-Enhanced Creativity, Real-Time & Dynamic Creative, and AI Craft. The festival's framing: the work that couldn't exist without AI and human craft together.
The award category is the industry's lagging indicator — once Cannes recognises a discipline, it's officially a discipline. AI Craft now sits next to copywriting and direction in creative-team org charts. For agency leaders, the practical move is to hire (or upskill) for the role this year, not the next budget cycle. The work that wins in Cannes 2027 is being briefed right now.
7 Typeform Growth Flow turned forms into a routing agent
Typeform launched Growth Flow, a form-completion system that triggers downstream actions when respondents finish — data straight to sales pipelines, email lists, and support tickets — and uses AI to read sentiment and adjust subsequent questions. The static form, finally, is a workflow.
The detail to notice is the sentiment-branching. A form that asks one question of a happy respondent and a different one of a frustrated one is closer to a conversation than a survey, and it generates training data for the agentic CX system you build next. If your team still treats forms as gates, you're leaving data and empathy on the table.
8 Kovva quietly automated the ops layer
Adtech startup Kovva launched AI agents focused on the repetitive ops work that dominates programmatic: QA, discrepancy investigation, reporting reconciliation, creative-fatigue detection. Less "make my creative," more "stop the team from doing the same Excel job at 11pm."
The agentic launches that change headcount math the fastest aren't on the buy side; they're in the back office. Programmatic ops is full of low-glory, high-leverage tasks where an agent pays for itself in a week. If you run an in-house team, the right question this quarter isn't "what can AI make?" It's "what can it stop the team from making by hand?"
9 Walmart credits Sparky for ad revenue too — the loop tightens
Underneath Walmart's 37% Q1 ad revenue growth is the same agent making the shopping recommendations. Sparky is now both a merchandising surface and an ad surface — sponsored placements show up in chat responses, and the agent's own context becomes a targeting signal for Walmart Connect. The screen, the basket, and the ad are now one product.
This is what retail media looks like when the retailer also owns the agent: closed-loop, but on the retailer's terms. Brands buying Walmart Connect should push hard for visibility into how Sparky's recommendation logic ranks them, and how sponsored content is labelled inside the chat. The closed loop is powerful; the part you can't see is where margin goes to die.
10 Kustomer Architect made "AI transformation" a SKU
Kustomer announced Kustomer Architect, a guided capability inside its AI-native CX platform aimed at CSAT, retention, operational efficiency, and revenue. The vendor stops selling "a chatbot" and starts selling a transformation outcome with metrics attached.
This is what the maturing AI pitch looks like — an outcome line item on a contract instead of "deploy our model." The thing to watch isn't the framework, it's whether the committed metrics are ones the CFO already tracks. AI programs that report on CSAT and retention get renewed. The ones reporting on "agent adoption" don't survive contract review.