By handling repetitive work, AI creates opportunities to scale with fewer people even in labor-intensive businesses such as purchasing services. Yet Moloco CEO Ikkjin Ahn stresses that even in the fiercely competitive AI era, growth begins with a good product. What matters is a virtuous cycle: introduce that product to customers who need it, measure results, and reinvest revenue in the product and marketing.

1. Beyond SaaS: Neo Firms That Finish the Job

As generative AI spreads, people increasingly accomplish tasks once handled by separate software by talking to AI or building the functionality they need themselves. SaaS, which provides business software over the internet, may not disappear immediately, but selling features that make work easier no longer offers the differentiation it once did.

At KV Insightful Day 2026, hosted by Kakao Ventures, the concept of a Neo Firm was introduced as the next stage of this change. SaaS provides tools for doing work; a neo firm uses AI to handle the customer's work from beginning to end and deliver the result. Instead of renting out a fishing rod, it catches the fish and brings it to you.

The difference from a conventional agency lies in how the work is organized. A typical agency must hire more employees as customers and workloads increase. A neo firm designs its organization and workflows from the outset so that AI handles a substantial share of the core work. This is a more fundamental change than adding automation features to existing operations one at a time.

2. Modepick Changed the Cost Structure of Purchasing Services

The overseas purchasing service Modepick makes the concept easy to understand. Primarily operating in Japan, it buys products that are difficult to find in Korea, cheaper locally, or limited editions, then handles everything from inspection to international shipping, customs clearance, and customer inquiries.

Purchasing services have traditionally been difficult to scale. Products must be found and listed, then purchased locally; their materials and uses determine customs eligibility and shipping requirements. Shipping delays, exchanges, returns, and customer inquiries add further work, so more orders usually mean more operations staff and higher costs.

Modepick redesigned this around AI. Excluding development and marketing staff, its purchasing, customs, and logistics operations employ one person in Korea and one in Japan—two in total. According to the company, cumulative sales exceed KRW 13 billion, expected revenue this year is approximately KRW 3 billion, and each operations employee handles about KRW 400 million worth of goods per month.

Product listing, administration, customs, and repetitive customer inquiries are automated as far as possible, allowing employees to focus on local purchasing and hands-on inspection and packing. The company says handling 100 routine inquiries requires less than ten minutes of human checking and interaction. People remain more deeply involved in exceptions requiring physical verification, such as exchanges and returns.

3. Accumulated Decision Records Matter More Than AI Alone

When everyone can use AI, Modepick's differentiator is ground-truth data accumulated through real operations. Customs decisions depend on a product's materials, intended use, and certification, making accurate handling difficult for AI relying solely on internet information.

Modepick records which products actually cleared customs, where AI made mistakes and how people corrected them, and which situations required exceptions. It connects those records to orders, inquiries, and sales data, organizing everything from the outset so AI agents can retrieve inventory, order, and customs information whenever needed.

The point is not simply having lots of data. As the company builds a structure that applies experience from completing work and correcting errors to subsequent tasks, similar situations require less repeated human intervention.

It also keeps reducing AI costs. Costs vary by model and the amount of information supplied. Through improvements by its internal development team, the company says it reduced product-listing costs from about KRW 700 per item when outsourced to freelancers to KRW 4–15.

This suggests opportunities even in fields previously considered incapable of becoming large businesses. Redesigning products, operations, and costs around AI can allow existing work to scale or enable services previously too expensive to offer. The article proposes asking the following when looking at a new AI service:

"Which cost that we used to take for granted has this service reduced, and what has that made newly possible?"

4. The Starting Point for Growth, According to Moloco's CEO

The author heard Moloco CEO Ikkjin Ahn deliver a keynote at MOLOCON 26, then interviewed him and CTV partnerships head Hyejung Lee privately for about 30 minutes afterward. Ahn shared figures illustrating competition in the AI era:

"In the first half of this year, app launches on Apple's App Store more than doubled year over year, but app downloads grew by only 2%."

AI has made apps easier to build, but customer time and attention have not increased at the same rate. Asked what matters for a new service to grow quickly, Ahn emphasized the product before advertising.

"On average, exposure is becoming harder to obtain and most new apps struggle, but some still grow rapidly. Even in the AI era, the key is the product."
— Ikkjin Ahn, CEO of Moloco

His growth sequence is build a good product, help more customers discover it, and reinvest steady revenue in the product and marketing. Sustainable growth requires connecting product quality, customer acquisition, and reinvestment.

5. Zeta Demonstrates the Impact of Product Improvements

The Japanese expansion of Scatter Lab's AI character-chat service Zeta supports the argument that growth starts with the product. After roughly six months of stalled growth following launch, Scatter Lab stopped marketing and other development work to concentrate on improving its AI model.

Although users received no separate announcement, usage time, return-visit metrics, and revenue all doubled the day after the new model launched, according to the company. Because users create stories through conversations with characters, improving the model directly improved the stories they experienced.

A better product also made promotion more effective. Users who arrived through ads shared enjoyable moments on social media, bringing in others who saw those posts. Scatter Lab said the increase in customers arriving without ads improved paid advertising efficiency, enabling more aggressive investment in Japanese marketing. At the time of the presentation, Japanese revenue was twice Korean revenue.

The diagnostic sequence matters. If existing users keep returning but few new users arrive, consider how to get the word out. Conversely, if people install the app but rarely return, examine where they lose interest before increasing advertising spend.

6. Specialized Advertising Technology Expands Customer Reach

Once you have a good product, the question becomes where to find customers who need it. Advertising inventory now extends beyond TV, YouTube, and Instagram to shopping-site search results, product pages, and streaming services. Moloco calls advertising spaces scattered across apps and services outside major platforms such as Meta and Google the "open internet."

AI selects ads people may find relevant based on signals such as products they viewed, then incorporates actual purchases into subsequent decisions. Two people searching for the same sunscreen, for example, may see different ads if one prioritizes price and the other seeks a particular feature.

Why would a platform use a specialist's technology instead of building such important advertising capabilities itself? Ahn cited Netflix's decision to use Amazon's cloud service, AWS, despite Amazon operating a competing video service.

"Amazon already had a video service when Netflix decided to use AWS. People asked why it would use someone else's infrastructure for a core service. Netflix made a bold decision, and we all know what happened afterward."

"People used to make kimchi at home, too. As markets mature, specialized services develop. In advertising, there are areas where we can provide value globally."
— Ikkjin Ahn, CEO of Moloco

The harder task is not creating advertising inventory but deciding which ad to show at every moment. A system must predict purchase likelihood, bid for inventory in real time, and feed actual outcomes back into its decisions. Building a system that handles this across multiple services simultaneously requires substantial technology and investment.

Ahn described Moloco as a partner working alongside a platform's engineering team. Platforms can set their own rules—for example, how much customer data to share externally and how often an individual sees the same ad—while obtaining operating technology from specialists. Making it easier for platforms to enter advertising also creates more places for brands to reach customers.

7. TV Advertising Can Start Small and Be Measured

TV advertising was traditionally seen as a medium for established growth companies such as Kurly, Musinsa, and Ohouse to build scale by introducing themselves to a mass audience at once. With CTV, or internet-connected television, advertisers can target interested customers as they do online.

Hyejung Lee explained that because US viewers move between streaming apps, advertisers can seek people interested in a product across several apps instead of advertising in just one.

"Many advertisers actually run CTV campaigns continuously. Efficiency compares reasonably well with existing channels, and they can reach users those channels do not."
— Hyejung Lee, Head of CTV Partnerships at Moloco

Small brands can now test TV advertising with videos they have already produced, then adjust budgets based on results. However, when someone sees an ad on TV and later purchases on a phone, attributing the purchase to the ad is difficult. App installations can be tracked with separate tools, but purchases on the web are harder to attribute to specific ads.

This is where incrementality matters. Showing ads to only one of two comparable groups and comparing their purchase rates allows advertisers to estimate the additional purchases caused by advertising. Dividing ad spend by that additional purchase count reveals the cost of generating one more purchase, informing whether to expand the budget.

AI has changed how companies find customers and operate advertising, but the fundamentals of growth remain: build a good product customers need, help the right people discover it, measure actual results, and reinvest. Those are the conditions for sustainable growth.