Unilever automates everything except the relationship.

By Jeremiah Rosen, Founder & CEO, Sundae Collective

We're obsessed with creator infrastructure for a reason. One-off investments don't compound. The right infrastructure does — better results over time, for less effort, not more. Meta just gave us more of it: Ads Manager now talks to Claude, ChatGPT, and Perplexity, and there's a new Creator Marketing Hub that puts creator discovery and paid activation in the same workflow. The advantage in either one doesn't go to whoever installs it fastest. It goes to whoever didn't spend the last three years storing their creator data in a spreadsheet named "FINAL_v2."

Neither tool works without a live operator behind the switchboard. An AI assistant with nothing organized to query is just a very confident guesser. Regular readers know where I land on this: AI everywhere except the source content. Automate the routing. Never the work itself (see story #2, below).

Picture the version of this that's already possible on your team. Someone running partnership ads asks a connected assistant which creator content is fatiguing this week, gets a ranked list with the underlying performance data, and ships a budget shift before their first meeting, coffee still hot. No exported spreadsheet, no waiting on a platform rep, no context-switching between three tools to answer one question. But that answer only exists because someone organized the performance data, the creator history, and the content tagging well enough for an assistant to find it in one query instead of ten. The assistant does the lookup. The person still decides what matters. The assistant doesn't get an opinion. Yet.

Now picture the earlier stage of the same job. A campaign brief comes in for a product launch. Instead of a separate discovery pass through Creator Marketplace — about as fun as assembling an IKEA wall unit without the diagram — someone pulls up the Hub, filters existing creator content by product tag, and has a shortlist of assets already tied to real performance data before the strategy meeting starts.

Ask a well-organized system a sharp question and you get a ranked list in seconds. Ask a disorganized one and you get the same mess you had before, just narrated back to you faster and with more confidence than it's earned. Garbage in, garbage out was true in 1957. We've just added a chat window. The switchboard doesn't replace the operator. It rewards whoever trained the operator well.

That's the same principle Unilever is running at a completely different scale with its 300,000-creator network (see story #1, below) — proof this isn't just a tidy metaphor, it's a real budget decision. Automate the retrieval, keep the judgment human, and make sure what's underneath is worth retrieving from. Meta just handed the switchboard for that system to a team of five, no enterprise budget required.

The real head start this quarter isn't installing Meta's connectors first. It's making sure the data underneath is worth asking. Get that right once, and every AI layer built on top of it — this one and the next one — starts paying off immediately.

— Jeremiah Rosen, Sundae Collective

Automate the busywork, never the trust — that's the line running through this week's stories, and not everyone's on the right side of it.

Digiday, July 10, 2026

Unilever has grown its creator network from 10,000 to 300,000 and is now routing 50% of its ad budget through social, up from 30%. To manage that scale, the company is automating creator discovery, vetting, brand safety checks, and administrative workflows — including a tool that scans social video for people already talking positively about its products and surfaces them as potential partners. What stays human is the relationship itself: deciding who fits the brand, who can build trust, and who gets creative latitude. As Leandro Barreto, CMO of Unilever's beauty and wellbeing group, put it: technology is there "to augment the human choices."

Sundae take: This is the org-chart version of the theme connecting these stories — software should absorb retrieval and routine, not judgment. The first process worth automating is whichever one is least differentiated, not most glamorous. Unilever kept the differentiated part, deciding who's actually right for the brand, for humans, and that's the split worth copying regardless of your network's size.

Read on Digiday →

The Wall Street Journal, July 15, 2026

AI-generated product demos, built on synthetic avatars or digital duplicates of real creators, are spreading fast across TikTok Shop, where US sales are projected to climb 48% to $23.4 billion this year. TikTok's own tools let sellers spin up shoppable AI clips in minutes for a few dollars, and platform policy permits fully AI content as long as it carries a disclosure tag. Affiliate creators say the clips are siphoning off ad spend and sales that used to go to real product demonstrations. SharkNinja has responded by banning AI content outright and stripping commissions from any affiliate who uses it.

Sundae take: SharkNinja's ban is the disclosure-label test nobody ran: a label solves a legal problem, not a trust problem, and TikTok Shop's affiliate economy is proving the difference in real time. Brands that write an explicit AI-content policy for their affiliate programs this week get to set the standard everyone else ends up copying.

Read on WSJ →

TikTok Newsroom, July 10, 2026

TikTok has now labeled more than 3 billion videos as AI-generated, using a mix of Content Credentials, creator labeling tools, and invisible watermarking. This week the platform joined the C2PA Steering Committee alongside Adobe, Google, Microsoft, and OpenAI, giving it a direct hand in shaping the content-provenance standard rather than just adopting it. TikTok is also testing tighter detection for AI-spam accounts in politics, finance, and health content, and rolling out new AI literacy resources for its community.

Sundae take: Provenance infrastructure is becoming table stakes, not a nice-to-have. Knowing exactly where a platform draws the line on AI content isn't just risk management — it's what lets a brand move fast and confidently on new formats instead of hedging on everything that comes out with an "AI" label attached.

Read on TikTok Newsroom →

Associated Press (via ABC News), July 13, 2026

A European Commission-commissioned expert panel has recommended banning social media access for children under 13 unless a parent or teacher is present, and requiring platforms serving 13-to-18-year-olds to build in guardrails against compulsive use, like limits on infinite scrolling. Commission President Ursula von der Leyen cited data showing European teens average four to six hours a day on social media, with close to 60% reporting emotional or psychological struggles tied to their time online. Member states remain split on the exact age line, with proposals ranging from 13 to 16. Von der Leyen is expected to turn the recommendations into a formal law proposal in September.

Sundae take: September is the date that matters here, not whatever implementation date eventually follows — that's when this becomes an actual law proposal instead of a panel recommendation. Brands with EU-facing creator programs who understand where the guardrails are heading can design campaigns that work with the new rules instead of around them, which is a real positioning advantage over competitors still planning as if platform access stays unlimited.

Read on ABC News →

Curated by Isla Novak, Sundae Collective.

Working on something in creator marketing and want a second set of eyes? Get in touch.