# Safety, Limits & Data Handling

> Understand what Kommon Poll AI can and can’t do, and how to use it safely.

> Understand what Kommon Poll AI can and can’t do, and how to use it safely.

Kommon Poll AI is designed to be helpful and safe, but it does have limitations.

## 11.3.1 Data Scope

- AI answers are based on data available in your Kommon Poll workspace and the filters you apply.
- It does **not** see conversations outside what Kommon Poll has collected.
- For privacy, it should not be used to rediscover or store personal information beyond your organisation’s policies.

Implementation details—like storage and model behaviour—may be described in your separate **Privacy** or **Security** documentation.

## 11.3.2 Known Limitations

- **Summarisation can miss edge cases** – AI may focus on dominant patterns and miss very rare issues.
- **Nuance in sentiment or sarcasm** – some posts may be misinterpreted, especially with heavy sarcasm or mixed language.
- **Not a replacement for human judgment** – use AI as an assistant, not an oracle. Final decisions should still involve human review.

## 11.3.3 Best Practices for Using Kommon Poll AI

- Verify key claims with charts and mentions, especially for high-stakes decisions.
- Iterate your questions – treat AI as a teammate you can refine and correct.
- Avoid sensitive or confidential prompts that go beyond normal analytics use.
- Use AI mainly for:

  - Faster summarisation.
  - Better storytelling.
  - Idea generation and prioritisation.