What is a proactive AI assistant?
A proactive AI assistant finds your work, so you do not start from an empty prompt. We explain the three kinds, and who decides what the assistant does.
By @miguelaeh_ ·
The short answer
A proactive AI assistant is an AI assistant that brings you something before you ask for it. A chat assistant, like ChatGPT, starts from an empty prompt, so you have to think of the work and type it with its context. A proactive assistant reads the places where your work arrives, like your email, your chats, your meetings, etc., and it starts from what it reads there.
The word “proactive” can sound like an assistant that acts without you, but that is a different thing. A good proactive assistant finds the work and shows it to you, and it does a task only after you start it.
A reactive assistant starts from your prompt, and a proactive assistant starts from your apps
| Reactive assistant | Proactive assistant | |
|---|---|---|
| Where the work starts | An empty prompt | Your email, chats, meetings and apps |
| Who thinks of the work | You | The assistant finds it, and you choose |
| Where the context comes from | You paste it or type it | The assistant already read it |
| An example | “Write a reply to this email: …” | “Marta is waiting for the refund date. Here is a reply, ready to send.” |
Products use the word “proactive” for three different things
Products use the word for different things, so two assistants that are both called proactive can leave you with very different amounts of work. For example, imagine that a customer called Marta writes to ask for the date of her refund. One kind of assistant tells you that Marta wrote. A second kind answers her only if you set up a routine for refund emails first. A third kind shows you a task for Marta’s email, with a plan. Before you choose a product, check which kind it is.
- It tells you. This kind sends you a morning briefing, an inbox summary, or a nudge, and you still do the work yourself. Poke and the Daily Brief in Gemini Spark are examples.
- It runs what you set up. You write routines, triggers or checklists, and the assistant runs them on a schedule or when something happens. It does only the work that you described first. Lindy and OpenClaw are examples.
- It finds the work and has it ready. This kind reads your apps and finds what is yours: what people ask you, what you promised, and what comes up about your work. Each of these things reaches you as a task with a plan. Sol does this for Gmail. Tecla does it across your work apps.
The first question to ask is who decides what the assistant does
We think this is the first question to ask about a proactive assistant, because an assistant that sends emails or books things when you did not ask is a problem. Some early products had this problem (TechCrunch).
In Tecla, a task starts when you click Run on it, and every task shows its plan before it starts, so you know what Tecla will do. A message or an email in your name waits for your approval before it goes out. In the example, the reply to Marta stays a draft until you approve it. While a task runs, you see what it did, what it is doing and what comes next, and you can stop it. Autopilot, which starts new tasks for you, is off until you turn it on.
Check these five things before you choose a proactive assistant
- Does it find the work, or only tell you about it? A briefing is useful, but you still have to do the work that it lists.
- Where does it look? Some assistants read only email. But work also arrives in chats, meetings, code and tools, so an assistant that reads only email finds only a part of your work.
- Who decides what runs? Look for a plan that you can read before a task starts, because the plan tells you what the assistant will do.
- How do you see the status of each thing? A list of tasks shows the status of each task. One long chat thread does not show it, because the messages about all your requests are mixed together.
- What happens with your data? Check if the company uses your data to train AI models.
We compare the products themselves in the best proactive AI assistants in 2026.
Questions
- What does proactive mean for an AI assistant?
- It means that the assistant brings you something before you ask: a briefing, a task it found, or a draft. It does not have to mean that it acts without you. In Tecla, it finds your tasks. You click Run on a task, and Tecla does the work.
- Is ChatGPT a proactive AI assistant?
- Not in its main form. In a ChatGPT conversation you write a prompt and it answers. OpenAI’s dots, which comes with ChatGPT Pro, looks through your connected apps for ways to help, but it works mostly toward goals you give it. See Tecla vs dots.
- Are proactive AI assistants safe?
- It depends on who decides what runs. Choose one that shows its plan before it acts, that lets you stop it, and that does not use your data to train AI models.
- What is an example of a proactive AI assistant?
- Tecla is one. It reads your email, chats, meetings and work apps, and finds your tasks with the context to do them. You click Run on a task, and Tecla does the work. Others are Sol, Lindy and Poke.
- What is the difference between an AI agent and a proactive AI assistant?
- An AI agent is software that does work in steps, with tools. Most agents start when you give them a goal. A proactive assistant also finds the work. Tecla is both: it finds your tasks, and its agents do the work after you click Run on a task.
