Why Is Korean Politics Still Stuck at 'Chatbots'?

Why Is Korean Politics Still Stuck at 'Chatbots'?

2026: Where Party Digital Infrastructure Stands, and What Comes Next

These days, you hear the phrase "adopting AI" a lot in politics. Parties are building chatbots, putting generative AI to work in election campaigns, and experimenting with new digital services that connect candidates and voters.

The direction is right.

But the problem is execution.

The June 3, 2026 local elections were arguably the first nationwide local elections in which generative AI made a full-fledged entry into the day-to-day campaign work of parties and candidate camps. Everything appeared, from chatbots that help voters easily access pledges to an AI campaign manager that maps out a candidate's stumping route.

And yet, the way Korean politics uses AI today is still stuck at the stage of bolting on a chatbot. What actually matters is not the chatbot itself, but the data infrastructure and operating system working behind it.

This piece is not an attempt to criticize any one party. It is a memo meant to point out the structural limits of the digital infrastructure that all of Korean politics shares, and to lay out where the next step lies.

Seen through the eyes of someone who has built products at a startup, this problem is not unfamiliar. If anything, it's familiar. And so the answer, too, is relatively clear.

1. Every Party Handles AI Differently

Right now, each party uses AI in its own way.

The Democratic Party launched a dedicated local-election site, '2026WIN.kr,' offering candidate information and a policy-proposal feature. According to an announcement from the Democratic Party's Gyeonggi Provincial chapter, the site provides a promotional space for each candidate and a space for voters to submit policy proposals.

The party also added a chatbot called 'Reobeuri' to the site, providing AI-based information such as summaries and analyses of candidate pledges and predictions of personality, leanings, and MBTI type. There were limits, but in terms of direction alone, it comes closest to the most comprehensive, platform-style approach.

The People Power Party mainly used AI to streamline its candidate-nomination process. Ahead of the June 3 local elections, it introduced an AI- and data-driven online nomination system and attached an AI chatbot that answers questions about nomination application procedures, eligibility requirements, and required documents. For the nomination review and vetting process, it adopted an AI-based "political credit-rating" model that quantifies and comparatively analyzes applicants' contributions to the party, public activity in their region, and integrity.

The Reform Party has been comparatively aggressive about supporting candidates' on-the-ground work. Through its 'AI Campaign Manager' app, it maps out a candidate's stumping schedule and route, and it proposes campaign strategies that account for foot traffic by area and the candidate's mode of transport. Party leader Lee Jun-seok personally demonstrated the app, and confirmed nominees were able to use it out in the field during actual campaigning.

The Reform Party also uses AI in the realm of pledges. When a candidate enters a pledge, the system reviews things like its alignment with the party platform and policy, its similarity to the central party's pledges, how it compares to the pledges of 2022 local-election winners, and whether it overlaps with other candidates' pledges — then presents an overall score and suggestions for improvement.

Each party's approach is different.

A comprehensive platform, internal operations, candidate support, pledge review. Rather than any one of these being absolutely right or wrong, what's interesting is that each party has a different philosophy and set of priorities when it looks at digital.

2. The Shared Limit Lies Between 'Technology' and 'User Experience'

The approaches differ from party to party, but there is a wall they all run into.

They've adopted the technology, but it doesn't translate into a finished user experience.

A representative case was the error in the Democratic Party's 'Reobeuri' chatbot. Asked who the sitting president was, Reobeuri answered with a former president as though he were the incumbent; the party later explained the cause as the cutoff of the model's training data and its limits in reflecting up-to-date information. The chatbot was subsequently suspended and scrapped.

Technically, this explanation is understandable.

But that's not where the heart of the matter lies.

That an LLM doesn't know the latest information is a well-known limitation by now. That's why, in a real service, you design RAG, system prompts, fine-tuning, guardrails, human operator review, and a data-refresh system together. In particular, a core fact that directly affects a service's credibility — like "who is the sitting president?" — should have been defended with a separate data source or a "guardrail."

In other words, this problem is less a limitation of the LLM itself than a gap in product design and the operating system around it.

Similar problems repeat across politics' digital services. There's a chatbot, but it can't properly answer users' actual questions. There's a website, but it's hard to find where the key information is. Data gets collected, but that data doesn't connect to decisions or follow-up action.

In the end, the problem is one and the same.

Technology, content, data, and user experience are not being designed as a single system.

3. Why This Keeps Happening

This problem is not a simple mistake. There are reasons it is bound to recur structurally.

First, elections are short on time.

Whether it's a presidential, general, or local election, the date is fixed. As a result, digital assets tend to become temporary projects thrown together in a rush to fit the election calendar, rather than infrastructure that accumulates in peacetime.

Second, the dependence on outsourcing is high.

Parties don't have enough PMs, engineers, designers, and content strategists in-house to own a digital product from start to finish. So they hand it off to outside vendors. But a vendor struggles to deeply understand the party's political context, and the party struggles to fully understand the technical details. The completeness of the service falls through this gap.

Third, the post-launch operating system is weak.

A chatbot is not a product you build once and are done with. You have to analyze user questions, collect cases of wrong answers, update prompts and data, and continuously refresh policy and candidate information. And yet, when the election ends, the system stops along with it. At the next election, they build it again from scratch. The assets never accumulate.

Fourth, the decision-making structure is still centered on politics.

A digital product should be designed around user behavior. But in practice, decisions are often made on the basis of political seniority and organizational logic. A deliverable that a political veteran judges as "good enough" can feel awkward and uncomfortable to the actual users in their 20s to 40s.

In the end, today's limits are not the problem of an individual or a particular party.

They are the problem of a structure in which party digital infrastructure still can't operate like a product organization.

4. The Next Step Is Not Chatbots but Infrastructure

For politics' use of AI to reach the next stage, the perspective has to change.

The core is not the chatbot.

The core is the data infrastructure behind the chatbot.

The assets a party should really be handling are member data, candidate data, pledge data, local-issue data, constituent-complaint data, media-exposure data, event data, and relationship data. Only when this data is organized in a consistent structure can a chatbot, an analytics tool, or a decision-making system be built on top of it.

If you build the chatbot first and try to fit the data in afterward, the same problem repeats. Conversely, if the data structure is in place first, the chatbot becomes just one of many interfaces that sit on top of it.

The area politics should pay particular attention to is CRM.

Politics, in the end, is the work of people.

Who met whom, in what context, what promises were exchanged, what follow-up is needed — all of it has to be recorded and connected. If a candidate receives a business card but doesn't know who that person was, what they talked about, or when to reach out again, the organization never accumulates.

A single business card is not just a contact.

It is the starting point of relationship data.

Only when this relationship data builds up does a party become a real organization. It has to break free from relying on a candidate's personal memory and an aide's notebook. The touchpoints of political activity have to be turned into data, and that data has to lead to follow-up action.

5. Content and Technology Have to Be Handled by One Team

Another important thing is the integration of content and technology.

Right now, the content team, the tech team, the policy team, and the organizing team often move separately. So there's a chatbot, but its tone of voice clashes with the party's official messaging. There's video content, but the response data from it doesn't connect to the next messaging strategy. There are policy materials, but they aren't reworked into language users can understand.

Political digital infrastructure is not a mere development project.

It is a comprehensive product in which policy, messaging, data, design, technology, and on-the-ground operations all have to mesh together.

That's why a party needs a PM role that owns the digital product. Not someone who simply places an order for a website, but someone who looks at the user experience, the data flow, the content tone, and the operating system all at once.

A party, too, now has to be a campaign organization and at the same time a media organization, a data organization, and a product organization.

6. Peacetime Infrastructure and Election-Period Operations Have to Be Separated

A structure where you build from scratch at every election is inefficient.

A party needs a backbone system that accumulates in peacetime. Member, candidate, regional, policy, relationship, and content data should be organized and refreshed as a matter of routine. During an election period, you just lay a campaign layer on top of it.

Peacetime infrastructure and election-period operations have to be kept separate.

Only then do the assets remain after the election ends. You don't start again from zero at the next election.

A party's digital capability is not decided by an event page hastily thrown together right before an election.

It is decided by how much data you've accumulated, operated, and connected in peacetime.

7. Overseas Cases Point to the Next Step, Too

Abroad, political digital infrastructure has already established itself as an industry.

In the United States, politics- and campaign-specialized platforms like NationBuilder and NGP VAN support the databases, fundraising, websites, email and text outreach, and compliance work of candidates, parties, civic groups, and nonprofits. NationBuilder provides a people database, fundraising, websites, and email and text tools, while NGP VAN puts political fundraising and compliance software front and center.

In Japan, 'Team Mirai,' a new party founded in May 2025, made AI and digital democracy its core identity. According to Team Mirai's official results, in the 2026 House of Representatives election it won some 3.81 million proportional-representation votes and secured 11 seats.

Team Mirai's official campaign page puts investment in growth industries like AI, robotics, and autonomous driving — and political and administrative reform through digital — front and center, and even connects an AI question feature to its policy pages. It's a case where AI is fused not into a mere promotional tool, but into the party's very identity and way of operating.

Korea is heading in that direction, too.

It's just that, in many cases, it's still stuck at the "we installed a chatbot" stage.

The next step is clear. Not chatbots, but data infrastructure. Not one-off campaigns, but an always-on operating system. Not simple promotional material, but the connection between relationship data and follow-up action.

8. This Is Not Criticism but a Market Opportunity

An important conclusion emerges at this point.

The digital transformation of politics is not something to criticize, but a market to be solved.

That parties depend on outsourcing means, put the other way around, that this is a market open to outside experts. The opportunity belongs to teams that can handle content and technology together, teams that can design relationship data, and teams that can approach this from the perspective of peacetime infrastructure rather than a short-term election project.

Business-card-based CRM, in particular, naturally connects with politics. A politician's work is meeting people, those meetings have to be recorded, and the records have to lead to follow-up action.

A single business card is not a contact, but the starting point of a relationship.

If that relationship lives only in a candidate's personal memory, it never becomes an organizational asset. But once the business card, the conversation context, the region, the interests, the follow-up contact, and the commitments are structured, it becomes a long-term asset of the political organization.

Abroad, politics- and campaign-specialized CRM and organizing platforms have already established themselves as important infrastructure. In Korea, this market has yet to open up in earnest.

Right now is the early phase of that market opening.

9. The Next Step for Political Digital Infrastructure

Let me say it again: this piece is not an attempt to criticize any one party.

The Democratic Party, the People Power Party, and the Reform Party are all attempting AI and digital transformation in their own ways. It's just that Korean politics as a whole is still stuck at the same structural stage.

What's needed now is four things.

Data infrastructure.

Relationship data and CRM.

The integration of content and technology.

A peacetime operating system.

Once these four are in order, the digital operations of the next election can rise to a completely different level.

That change won't happen inside parties alone. It has to be built together with external partners who work alongside parties — product makers, data experts, content directors, and CRM designers.

I'd like to talk with people who are seriously thinking about the next step for political digital infrastructure. Setting aside which camp you're in, the problem to be solved is clear. The tools already exist. What remains is to connect those tools properly to the field that is politics.

Over the next four years, this space could become one of the most interesting markets in Korea.

From the perspective of a product maker.

Currently operating NameGood, a business-card-based CRM intelligence app, in the Korean and Japanese markets.

https://namegood.org

(Development ongoing.)