Why custom systems beat generic prompting for microcopy

Most approaches to using AI for content design and UX writing ignore how language models work. Blindly prompting ChatGPT and friends might seem efficient, but these tools only see fragments of what you’re solving and often lead to sloppy output. There’s no understanding of user needs, no guardrails for product voice, and no end to LLM hallucinations.

After building multiple writing systems, Khairul Bahriah has learned that, much like everything else in tech, using these tools effectively for UX needs solid systems and structures, not just better prompts. Static style guides aren’t good enough anymore either. You need comprehensive glossaries, structured rules, and continuous governance.

In this talk, Khairul shares what she built, what she broke, and what she’s learned from exploring and playing around.

In this session, you'll learn:

• How LLMs work: Tokens, context windows, memory, and reasoning.
• When to reach for LLMs: Guidelines for the team.
• Lessons learned the hard way: Real experiments and how they failed.
• Building and governing a system: How living docs, custom spellcheckers, and in-house solutions can work together.