What is a personal AI agent

Everyone claims to build 'AI agents'. Few know what it really means. An operational definition, with examples.

What a personal AI agent is NOT

The 5 characteristics of a true personal AI agent

1. Persistent memory

It knows who you are. It remembers what you said yesterday, last week, 3 months ago. Its memory is structured (not just "the last N conversations") and survives every session reset.

2. A defined persona

It has a tone, a point of view, thinking frameworks. It's not the default "polite assistant" from OpenAI/Anthropic. When it answers, there's a recognizable mindset behind it.

3. Proactivity

It messages you. It doesn't wait for you to open the app. Briefings, reminders, check-ins, alerts: the system stays disciplined for you.

4. Multi-modal input

It accepts text, voice, images, photos. You think the way you think, it understands and structures. No UI forcing you into a particular way of entering data.

5. Framework-driven

It doesn't improvise. It uses specific mental models (GTD, ICE, the DRIP Matrix, etc.) to classify and suggest. The quality of its output is a function of the quality of its frameworks.

Otto as an example

Otto embodies all 5 characteristics: it lives on Telegram, has a 3-layer memory (long/mid/short term), has Chris La Porta's persona baked in, messages you proactively, accepts text+voice+photo, and reasons with David Allen's GTD framework.

A personal AI agent isn't a tool you use. It's a presence that works with you.

Ready to try Otto?

15 days free. Up to 200 messages. No card.

Create your Otto →