MGS 2026 Krafton Recap: AI Automation Cuts 60 Hours to 5

You pay for several AI tools, so why is someone on your team still pulling media data by hand every morning? When downloading reports and matching columns eats the whole morning, the decisions and insights that actually matter keep getting pushed back.

AI workflow automation means handing repetitive collection, validation, and reporting to AI and automation tools so people can spend their time on judgment and the next experiment. This post recaps "On Getting My Evenings Back with AI," a PLAY HALL session at Modern Growth Stack (MGS) 2026 presented by Yeowon Jung of Krafton. She cut repetitive user acquisition (UA) work from 60 hours a month to 5 and automated about 95% of it.

The daglo marketing team transcribed the MGS 2026 sessions with daglo and reread them to pull out the parts practitioners can copy right away.

Repetitive work crowds out decisions

The unglamorous truth of UA is repetition: download data from each ad network, align columns, check that numbers reconcile, write the report. As campaigns scale across more networks and countries, that work grows exponentially.

Games add a structural mismatch. Launch dates move often, but headcount and costs are fixed, so every delay leaves people and partner resources idle. UA, with its heavy repetitive load, absorbs a large share of that cost.

Step one is listing the work, not writing code

Before building anything, she laid out every task the team did, about 13 in total, with time spent, owner, and where repetition happened. Then she sorted them by impact and automation difficulty.

Starting with the hardest problem piles up failures faster than wins and drains the organization's appetite for change. So she began with data loading, cleanup, dashboards, and reporting, and pushed campaign setup and media mix down the list, because the first group helped marketing and business teams too.

The matrix also won support. AI work costs tokens and premium subscriptions, and "here's the order I'll change things in" is easier to approve than "let me rebuild everything." She formalized the plan as an official proposal rather than building in secret.

Three build principles and a three-tool team

The result is one product that collects, validates, and analyzes scattered media data and sends a report every morning with no one involved. It ran real campaigns in the first half of the year, built on three principles.

1. Replace the task completely: "Don't try to do your current job better; eliminate it." The scope covered the entire morning routine that once needed five manual touchpoints a day.
2. Give context and decision logic first: Ask AI for "a dashboard" and you get something pretty with no priorities. Skip adjectives like "beautiful" and explain the order in which your team screens data and which metrics matter most.
3. Design for each role: Views for executives and partner teams, a mobile version for a one-minute read on the commute, and natural-language Q&A.

She kept n8n because its step-by-step nodes make errors easy to trace on a busy morning, and it suits routines that run on the same cycle every day.

Practical tips for handing work to AI

Start small, describe the goal and the finish line rather than the method, and have AI plan before it builds anything large.

• Ship small, show early: Pile every feature into one request and everything comes out mediocre. She started with data collection alone, then layered on dashboards, creative analysis, and reporting. For internal tools, the metrics are time saved and how often people use it, not revenue.
• State what, why, and "done when": Clear goals and completion criteria beat long instructions on how.
• Plan first for big jobs: Ask for a plan, give feedback, then execute step by step.
• Query expansion and a devil's advocate: Ending a prompt with a devil's-advocate role stops AI from locking in its first answer. Limit it to questions that could change the conclusion to avoid delays.

The payoff: focus, not leaving early

The team ran a launch campaign without new hires or extra partner support, moved from next-afternoon data to near real time, and cut 60 monthly hours of repetitive work to 5.

Honestly, she said, her evenings got busier. What changed was how she spent them: less time filling spreadsheet gaps to avoid mistakes, more time on growing her game and testing new ideas. She added that this isn't an argument against agencies; it gives teams the option to work more deeply with strong partners while AI handles routine operations.

Takeaways: a checklist to start automating

The lesson is about order more than tools. List the work, win early on easy tasks, and give AI your decision logic first, and automation will last.

• Have you listed every task with time spent, owner, and repetition points?
• Did you sort by impact and difficulty and start with quick wins?
• Have you prioritized automation that helps other teams too?
• Do your AI requests include decision order and completion criteria?

daglo's Speaker ID is handy for recordings where knowing who said what matters, like multi-speaker sessions and Q&A. It remembers a voice once you name it and fills in that real name instead of "Speaker 1" in future transcripts, and daglo is the first in Korea to offer this to every user.

For more sessions, see our full MGS 2026 recap, and find event details on the official MGS site.

#MGS2026 #ModernGrowthStack #AIAutomation #Krafton #UserAcquisition #MarketingAutomation #PromptEngineering #n8n

👉 Transcribe your next session