Ads that find buyers and agents that close the sale, on apps 3.64 billion people use every day.

Billions of people, a few right buyers

Meta's business rests on a simple trade. People use Facebook, Instagram, WhatsApp, Messenger and Threads without paying, and businesses pay to reach them. The company's own principles say it plainly: its business model is ads so its services can be free. In September 2026 an average of 3.64 billion people used at least one of Facebook, Instagram, Messenger or WhatsApp each day, and advertising produced $196.2 billion of the $201.0 billion Meta reported for 2025. The hard part sits with the buyer. A skincare brand or a mobile game studio wants the few thousand people likely to buy, not everyone. Meta's annual report says its marketers make no long-term commitments, spend only part of their budgets with Meta and move money when they doubt the return.

AI is accelerating our core business today, powering our next generation of products, and opening the door to entirely new enterprise opportunities.

Mark Zuckerberg · Founder and CEO · Second quarter 2026 results release · 2026-07-29

The belief behind Meta's recent choices is that its models pick audiences, placements and creative better than people setting filters by hand. Advantage+, which Meta says it introduced about three years before April 2025, makes those choices automatically. By the company's own research, campaigns using it returned $4.52 in revenue for each dollar spent, 22% more than standard campaigns, and by April 2025 most Meta advertisers used at least one Advantage+ product. In the second quarter of 2026 the end-to-end Advantage+ solutions passed a $75 billion annual revenue run rate, a measure that annualizes the current revenue flowing through them.

How an ad finds its buyer

Take an illustrative case: a Korean skincare brand selling into the United States. The brand uploads its catalog and sets a goal, such as purchases. For each person scrolling a feed, Meta's Andromeda engine narrows tens of millions of candidate ads to a few thousand, and larger ranking models such as GEM predict which one is most likely to lead to a sale. Meta reported in July 2026 that improved user understanding models, combined with GEM, lifted ad clicks on Facebook by 8.3% and conversions by 15.7%. Generative tools make more image and video variations from the brand's own assets, and more than 9 million small businesses use at least one of them. When a shopper taps a click-to-message ad, a chat opens on WhatsApp, Messenger or Instagram, and a Business Agent can answer, recommend and close the sale.

What still needs a person

Two jobs in that loop still need people. The first is the relationship. Someone has to understand a large client's business, budget and goals well enough to plan where Meta fits, which is how the Seoul Client Partner posting describes its work with top advertisers. The second is proof. A brand keeps spending only if it believes the ads caused the sales. Meta's Marketing Science team designs conversion lift and brand lift tests for that reason, and Meta said in January 2026 that its incremental attribution feature, which optimizes for conversions that would not have happened otherwise, raised incremental conversions 24% over standard attribution.

Meta is moving the hard parts of advertising into its own models: targeting, creative and now the conversation after the click. That shifts the value of people toward explaining results to a client and testing them honestly, and away from setting filters.

All three Seoul roles sit where Meta's systems meet a client: the commercial relationship, the measurement, and the technical design of messaging and agents. Expect to be asked how you would show that Meta, and not something else, moved a client's numbers.

Three products behind the Seoul roles

Advantage+ is the product most advertisers touch. An advertiser chooses a goal, such as sales, app installs or leads, and the end-to-end campaign automates audience, budget and placement together, while single-step tools automate one part. Meta says its sales campaigns lowered cost per acquisition by 9% on average and app campaigns by 7%, and that lead campaigns cut cost per qualified lead by 10%. Underneath sit the ranking systems described in the first chapter and an Opportunity Score that suggests changes. Advertisers buy it directly, through agencies such as the 26 partners in Meta Korea's 2026 Agency First program, or through resellers, and it competes for the same performance budgets as other ad platforms.

Measurement is what keeps the budget. A lift study splits people into an exposed group and a control group and credits the difference in sales or recall to the ads. The Seoul Marketing Science posting describes designing conversion lift, brand lift and attribution tests with clients and agencies, working with third party vendors that run studies, and building learning agendas, a plan of what a client will test and act on. Incremental attribution brings the same idea into delivery itself, and Meta says it reached a multi-billion dollar annual run rate seven months after launch. For whole media plans, Meta Marketing Science publishes Robyn, an open source marketing mix model that can be calibrated with lift results.

Business messaging is the newer line of business. Click-to-message ads open a chat instead of a website, and Meta said US revenue from them grew more than 50% in the fourth quarter of 2025. WhatsApp's paid messaging, where businesses pay to send messages through the Business Platform, passed a $2 billion annual run rate in the same quarter. On June 3, 2026 Meta expanded Business Agent to businesses of all sizes. It answers in the customer's language, recommends from the catalog, books appointments and closes sales, and by July more than one million businesses were using business agents every week. Large companies build on the Business Agent Platform, which connects to hundreds of systems such as Shopify, Zendesk and Shopee, often through messaging partners.

A Korean brand's agent in Italy

The clearest named case for the Seoul messaging role involves a Korean brand abroad. Kia Italia, part of the South Korean automaker Kia Corporation, wanted a faster path from ad to sale than dealership visits and call center follow-ups. With Meta and the messaging partner Infobip, it ran ads that click to WhatsApp, where an AI agent collected contact details in conversation, answered questions on models, prices and financing, and passed ready buyers to the nearest dealer. In four weeks the campaign produced 144 qualified leads, 116 quote requests and five direct car sales, with a 40x return on ad spend and 53% lower cost per car sold than Kia Italia's usual approach. The page labels the results self-reported.

Two other named cases show the same pattern. Movida, a Brazilian car rental company, rebuilt its WhatsApp bookings on the Business Agent Platform; replies came in under 15 seconds instead of 30 to 60, and in week four of a one-month test its conversion rate reached 14.9% against a previous best of 9.7%. Underneat, an Indian apparel brand that used to set up each campaign by hand, moved to Advantage+ sales campaigns and reported a 13% incremental lift in purchases.

Reading the case numbers

Each case measures something different: return on ad spend, conversion rate, incremental purchases. All are published by Meta or WhatsApp Business, and the Movida test appears with a different headline figure on the earnings call, a 44% rise in daily bookings. Treat them as evidence of what the product can do in one deployment, not as a benchmark.

In an interview, retell one of these cases with its measure and its limit, then say how you would test the same idea with a Korean client, for example a brand selling abroad through WhatsApp.