MGS 2026 Recap: CRM Message Automation at Kolon FnC

Are you spending so much time setting up the same recurring messages every week that you never get to plan new campaigns? The more repetitive setup piles up, the less room there is to design messages around the customer journey.

The daglo marketing team attended Modern Growth Stack (MGS) 2026, hosted by AB180 at the Westin Seoul Parnas from July 21 to 24, 2026. This recap covers "AI That Changes How an Organization Works: Introducing and Scaling a Message Automation Program," presented by Yeonhan Kim of Kolon FnC, the fashion company behind the Kolon Mall online store. We transcribed it with daglo and focused on what CRM practitioners can apply.

CRM message automation means letting AI-connected tools handle repetitive, rule-based message setup, such as promoting sales events. Kolon Mall automated the setup of roughly 200 recurring campaigns over seven weeks, while writing one rule into every document: "AI never sends messages."

Why CRM message automation was needed: 200 recurring campaigns in seven weeks

Kolon Mall is an online platform with about 20 in-house brands and more than 500 domestic and international brands. With so many sales events, it sent around 200 recurring campaigns to about 60 segments over seven weeks, and setting up those messages was the biggest bottleneck.

The sheer volume of recurring messages kept the team from building more detailed behavior-based journeys in Braze Canvas, and setup was where the work piled up.

• What setup involves: choosing the segment, frequency, schedule, and channel, and checking every sale link.
• Workload: about 24 campaigns a week at roughly 30 minutes each. Experienced marketers get it down to 10 minutes, while new hires take one to two hours at first.

The rules themselves are already mature. Where mature rules meet repetitive manual work is the sweet spot for automation.

AI works best on work you already do: opening trust in stages

Kim's view is that AI is strongest when attached to repetitive work you already do, not when you ask it to break new ground, because AI can't know on its own what new thing to do. So the team started with tasks that follow clear, repeatable rules.

To ease the fear of AI making mistakes, the team set up a framework where people judge and approve while AI only executes, and followed a sequence of small validation (MVP), standardization, then rollout. Using MCP, a standard way for AI to connect to outside tools, they linked Braze, Google Sheets, Notion, internal data, and link tools so AI could run the whole process, then opened permissions in four stages.

1. Read: At first, AI only reads. Starting with write access makes mistakes too hard to undo.
2. Diagnose and detect: AI uses what it reads to spot and flag problems.
3. Turn findings into rules: Verified checks become rules that block problems automatically.
4. Approved writes: Only at the end does AI get write access, with a person approving each change.

Once one workflow was deep and standardized, the team handed it to neighboring teams.

Inside the message automation program: a 7-step pipeline and safeguards

Kolon Mall's program runs in seven steps from collecting requests to setting up campaigns in Braze. AI reads the schedule, builds links, and runs review rules, but a person always signs off before anything is set up.

1. CRM marketers collect requests and set the schedule.
2. AI reads the schedule sheet automatically.
3. It generates a token that gives each message a unique value.
4. The token is used to create a unique link automatically.
5. The link is entered into Notion.
6. Review rules run automatically in Notion to catch problems.
7. Once a person confirms everything is fine, the campaign is set up in Braze.

Kim said building the standard format took more time than anything else. Here are the key design choices.

The safeguards are just as explicit. The first sentence the team agreed on was "AI never sends messages," and it is written into every working document. AI still has no write access and runs tests only after approval. Problems trigger an immediate block or alert, and if an incident happens, an automatic routine shuts the system off until it is fixed.

Teaching AI your domain: documents and memory

AI doesn't know your business, so how much information, documentation, and rules you give it decides how good it is. Kolon FnC taught AI through a fixed routine, from explaining the domain to recording what it learned.

The routine goes like this: explain the domain, let AI work it out, have it explain the cause when something is wrong, turn correct answers into rules, and record them in memory. The team had AI write about 24 Markdown documents plus a master index, and fed it a 25-page CRM guide written by a team member, then had AI check real data through MCP before proposing optimized automation rules.

• Going deep: Break work into stages and finish one at a time. Without a shared data standard, AI reads different data each time, draws the wrong conclusions, and errors in live campaigns are very hard to roll back.

A practical checklist for starting CRM message automation

The session left three takeaways. AI is strongest on work you already do. The information, documents, and rules you feed it decide its skill. And automation that keeps human judgment and accountability is the kind that spreads.

• Find candidates: List your team's weekly repetitive tasks by count and time spent, then pick those where mature rules meet manual work.
• Standards first: Define your key format and data standards before automating.
• Open access in stages: Read, diagnose, turn findings into rules, then approved writes.
• Write prohibitions into documents: Put the things AI must never do into every working document, not just a prompt.
• One source of truth: Let everyone see campaigns and send status in the same place.

For recordings where it matters who said what, like multi-speaker sessions or Q&A, daglo's Speaker ID is a useful feature. It remembers the voices of the people in your recordings and fills in their real names instead of "Speaker 1" from the next recording on, and daglo is the first service in Korea to offer it to every user. Learn more about Speaker ID

For the other sessions we attended, see our full MGS 2026 recap. Event details are on the official MGS 2026 site.

#MGS2026 #ModernGrowthStack #KolonFnC #CRMAutomation #CRMMarketing #MarketingAutomation #AIWorkflow #daglo

👉 Transcribe your next session