Hyundai Card ran two blind writing tests pitting a generative AI model against a PR professional with a journalism background, as the South Korean financial services group works out where machine-generated text fits into its communications workflow. Staff judged the entries without knowing which was which, and the human writer won both rounds - though the margin was narrower than some expected.
The contests are part of a broader internal AI transformation programme that aims to move beyond ad hoc use of generative AI toward repeatable processes for press releases, video scripts, and other communications output.
How the two contests played out
The first exercise asked both sides to write a press release for Orange, one of Hyundai Card's premium credit cards. Employees evaluated the drafts on strategic narrative, message structure and logical flow. The human version came out ahead, but staff feedback showed the distinction was not always obvious. One employee said, "Both had their own strengths and weaknesses, making it difficult to definitely declare a winner," while another added, "The more you read, the harder it became to tell which was written by AI."
Views on the same draft diverged. One evaluator said the human text felt authentic because "the theme is clear and the benefits are specific," while a colleague reading the same version said "the way the benefits are listed reads distinctly like AI."
The second contest focused on a pitch release for international media about Alphabet Card, Hyundai Card's newly launched independent brand. Again, the PR professional won. Evaluators said the human version "naturally incorporates Hyundai Card's voice and characteristic vocabulary," while the AI text drew praise for being "concise and easy to read, free of superfluous fluff."
Finding the right division of labour
The goal was never to crown a winner. Hyundai Card wanted to identify which tasks suit each side best. The company has since applied that thinking to daily work. For a recent in-house promotional video, AI produced an initial narrative structure, then human staff added context and refined the messaging before completing the final version.
"The two contests were never just about deciding who could write better; they were an exercise in validating the unique workplace competencies of both humans and AI," a Hyundai Card representative said. "AI significantly cut down the time required for research and drafting while enabling us to achieve a solid baseline of completeness in a short turnaround. When we integrate that with the context gathered through in-house reporting, Hyundai Card's distinct corporate vocabulary, and human judgment regarding core messaging, the overall quality of the content is substantially elevated."
The company plans to extend the model to domestic and international press releases and YouTube scripts. This reflects a pattern seen across corporate AI projects: firms often find more value in narrowing the scope of use than in trying to automate an entire function at once. For PR professionals looking to build these skills systematically, an AI Learning Path for Public Relations Specialists can provide structured guidance on integrating generative tools into communications workflows.
Training leaders alongside staff
The PR project sits within a wider push to train employees on large language models. Rather than rolling out a mandatory course for everyone at once, Hyundai Card prioritised organisational leaders and employees who wanted early hands-on access. Training covered news monitoring, executive briefing notes, summarising research papers, turning large volumes of material into mind maps and slide presentations, and using AI tools with Excel to automate routine work.
Vice Chairman and CEO Ted Chung attended the sessions alongside employees. "Leaders probably won't be doing vibe coding themselves in their day-to-day work, but executives must grasp the fundamentals to speak the same language as working-level teams," Chung said. He described seeing coding tasks that once took two days completed automatically in under a minute as "nothing short of extraordinary."
The next round of senior leader training will focus on operational risks linked to large language models and the governance needed to manage them - including understanding the limits and weaknesses of AI systems, not just their uses. "While responding swiftly to the rapid pace of AI innovation, we will scale our implementation thoughtfully by weighing the specific operational characteristics and risk profiles of each business function," the representative said.
The governance gap in PR
Hyundai Card cited a survey from PR software company Muck Rack showing that 75% of PR professionals worldwide already use generative AI in their work, yet 55% of organisations have not set guidelines for how AI should be used in PR. That gap between adoption and governance has become a central issue for communications teams, particularly in regulated industries where accuracy, tone and internal controls carry added weight.
For financial services firms, the challenge is not only speed or cost. Automated output must match brand language and comply with internal standards - areas where Hyundai Card's contests suggest human judgment still holds an edge.
Why this matters for PR and communications professionals
The Hyundai Card tests point to a practical model: AI handles early-stage drafting and research, while humans own final editorial judgment, brand voice, and strategic messaging. This division plays to the strengths of each. The findings also reinforce that writing quality alone is not the deciding factor - evaluators cared about whether the copy sounded like the company. Communications teams that invest in AI for PR & Communications training now can build workflows that preserve brand identity while cutting turnaround times on routine drafts. The larger risk is not that AI will replace PR writers, but that organisations without clear guidelines will use the tools inconsistently and erode the voice they are trying to protect.
Your membership also unlocks:
