Language Proficiency for Your AI Persona
Define your AI persona’s language fluency levels and charming imperfections for authentic conversations.
Overview
We’re excited to unveil a brand-new feature in My Persona: native-speaker accuracy and intentional imperfection. From now on, every AI persona you create can declare exactly which languages and how well it speaks. Users can enter proficiencies however they like, for example:
Swedish: Mother tongue, English: Fluent, French: Understand at C1, speak at B1, Turkish: Basic
Authenticity Through “Perfectly Imperfect” Speech
Your persona’s language skills are visible to both humans and AI assistants, and they’re used to generate more genuine, relatable dialogue. If your persona isn’t proficient in the language of a user interacting with it:
- It may switch to CEFR A2‑style replies: short, simple sentences, playful typos and mis‑conjugations, occasional grammar hiccups.
- You’ll still understand the meaning, just with the endearing quirks of a real learner.
Why degrade performance on purpose? Because perfect AI often feels inhuman and boring. By embracing a bit of “wobbly” language, your conversations become more vibrant, more personal, and more fun.
Hotel AI Assistants Stay Polished
Rest assured, our in‑Hotel Assistants continue to deliver flawless language mastery. This new flexibility applies only to My Persona agents.
Try It Today!
Ready to see it in action? Ask any existing AI persona on inHotel (e.g., Jan's) which languages it speaks, pick one it isn’t proficient in and watch it stumble charmingly through your questions!
Note: Training LLMs to make deliberate mistakes is surprisingly difficult. While consistency in error generation varies across models, the most advanced models deliver imperfect responses more often than those optimized for conversations. For now, we’ve selected a model that balances speed with occasional imperfections, so you won’t encounter errors in every interaction. It will only improve from here.
Give it a spin and let us know how those adorable errors brighten your day! 😊
Frequently Asked Questions
How do I configure my AI with language skills and fluency levels?
Your inHotel AI persona’s language settings now let you enter each language along with your self‑assessed proficiency (e.g., “Urdu: Mother tongue; Hindi: Fluent; English: B2”). Your persona will then reply at the corresponding CEFR levels.
How do I make my personal AI agent speak imperfectly to reflect my foreign‑language skills?
To make your AI persona speak with endearing imperfections, set its proficiency for that language below C1 (for example, “Spanish: B1” or “Thai: A2”), or simply leave it out of your inHotel My Persona Languages list. Once you’ve done that, your agent will automatically switch to shorter, simpler sentences and sprinkle in intentional errors (typos, mis‑conjugations, missing articles) just like a real learner.
Why would I want my AI persona to make intentional mistakes?
Intentional mistakes make your AI persona feel more human and relatable. No one speaks perfectly all the time, especially when learning a new one. By embracing a few typos, grammatical slips, and quirky word choices, conversations become more engaging and authentic rather than robotic. This “perfectly imperfect” style sparks empathy, builds trust, and adds a playful element to your interactions.
How does the persona decide when to switch between languages?
The inHotel AI persona mirrors the language of the user’s last message: if that language is listed in your My Persona settings, it replies at the configured CEFR level; if not, it automatically falls back to A2‑style “learner” mode in the language you used.
What mistakes will my AI persona make when not proficient in the user’s language?
When your inHotel AI persona isn’t proficient in the user’s language, it will use shorter, simpler sentences and introduce intentional errors such as typos, incorrect verb conjugations, missing articles, odd word choices, and occasional word‑order mix‑ups. These “learner‑style” mistakes mimic the quirks of someone still mastering a language while keeping the meaning clear.
