Dots, Bots, Pets & Agents: A Field Guide to AI’s New Zoo

AI is acquiring a whole new cast of characters. Dots. Bots. Pets. Agents. They sound like the start of a beautifully chaotic children’s book, and in a way they are: different interfaces, different levels of autonomy, different places for your work to live.

The useful distinction is less about the cute name and more about three questions: What can I hand over? Does it keep working when I leave? And how am I charged?

A chat assistant can help you think, draft, research and create while you are in the conversation. An agent can plan steps and use tools towards a goal. An automation starts work when a schedule or event tells it to. A pet is usually an interface companion, not a new capability. The names can be playful. The budget and the responsibility still need to be clear.

The tiny field guide

Chat assistant. The AI you work with conversationally. It can help with brainstorming, writing, analysis and creation. Connected tools can widen what it can access or do, but you are still steering the exchange.

Agent. AI that can plan steps and use tools towards a defined goal. It might research, create files, navigate an app or carry out a multi-step task. How independently it operates, what it can access and what needs approval vary enormously by product and account.

Dot. OpenAI describes Dot as an always-on agent on eligible ChatGPT Pro tiers. In practical terms, it is designed for ongoing work, memory and a cloud computer that can continue while your laptop is closed. It is not simply a cute label for every ChatGPT conversation.

Grok Bot. xAI positions Grok Bot access within paid SuperGrok plans. Think of it as a more agentic, work-across-tools layer rather than a synonym for ordinary chat. Check the current plan page for your region and the relevant usage allowance before designing a workflow around it.

Pet. An optional animated companion in an interface. A pet can make the experience warmer, easier to enter or more fun to watch, but choosing one does not automatically give an account persistent work, more permissions or a new job to delegate. Pip is our own reminder that a welcoming guide and a capable system can sit happily together.

Automation. Work started by a time, trigger or event. “Every Monday, review this inbox” or “When this form arrives, create a draft.” An automation can call an agent, but it is not automatically an agent itself.

Do not let the language hide the handover

Before you buy anything, name the actual handover. Is the tool helping you make a better first draft? Is it being asked to gather public research, create a file, use an account or keep a task moving while you are offline?

The further a tool moves from a conversational draft towards an ongoing action, the more useful it is to get specific about access, approval, privacy, error handling and the cost of additional usage. A charming pet does not remove those questions. A powerful agent makes them more important.

The money has three layers

A subscription usually pays for access and an included allowance. Usage can introduce credits, extra capacity or API charges. Running costs such as hosting, cloud computers and connected software can sit alongside both.

The table below is a price map, not a ranking. Prices were checked on 30 September 2026 from the linked provider pages. USD and Australian-dollar plans are shown in their published currencies, without an exchange-rate conversion. Tax, app-store billing, introductory offers, plan eligibility, regional availability and usage limits can all change the final number. The budget snapshots are subscription arithmetic, not equivalent capability.

Use one real assignment to compare tools

If you are trying to understand whether an agentic layer is worth it, do not compare feature lists alone. Give shortlisted tools the same contained assignment: research ten suitable sponsors, verify publicly available contacts, create a shortlist, and draft tailored outreach for your review.

Then look at four things: completion quality, your intervention time, total cost and the amount of human judgment still needed before anything moves. That gives you an answer you can use.

The best question is not “Which AI animal should I collect next?” It is: What can I responsibly hand over, and how much attention does it still need from me? That is the field guide worth keeping.