Iris, your AI assistant

The assistant is a conversational AI built into Instasent — it's called Iris. You talk to it in your language, and it has a threefold role: it answers (with your project's real data and questions about how the platform works), it teaches you (it explains how to do something and takes you to the screen), and it prepares (it leaves a change ready for you to confirm with a click). Its golden rule: it proposes, it never executes — nothing that sends, activates, saves, deletes or spends happens without you confirming it.

It's a single assistant: you talk to one face — Iris — and behind it, it coordinates a team of specialists (analytics, audience, help, setup) without you noticing. You open it from the panel and it adapts to the screen you're on, to help you right there or across your whole project.

What it can do

It answers with real data. Ask it about anything in your project and it looks it up for you:

• How is my latest campaign doing?
• Which segment is growing the most?
• How much balance do I have left?
• What's the best time to send?

It explores your audience (contacts, attributes, events and segments), your campaigns and automations and their results, your billing and balance, and your brand and settings. When the question is about how Instasent works, it doesn't answer from memory: it searches these same guides and cites the page so you can read the detail.

It helps you build audience segments and filters. You describe in plain language who you want to reach — "customers who bought last month and haven't opened any email" — and Iris builds the filter, explains what it does and leaves it ready. With one click it takes you to Audience with the filter already set, where you see live how many contacts match and you save it. It can count how many match and refine it with you if the result isn't what you expected.

It teaches you how to do it yourself. If something is done on a screen, it gives you the steps and takes you there with a link, rather than instructions from memory. And it's honest: if it can't do something itself, it tells you clearly and teaches you how; if something isn't an Instasent feature, it says so too — without making it up.

It prepares changes for your confirmation. It can fill in a form or propose a change and leave it ready — and the Save click is always yours.

It proposes, it never executes

Iris's golden rule: it proposes, it never executes. It doesn't send campaigns, activate automations, publish, delete or spend balance. Any action with consequences is confirmed by you with a click — Save, Send, Activate.

This isn't a temporary limitation: it's a design decision. The final control is always human, today and in the future — an assistant that could send messages to thousands of customers on its own would be a risk nobody wants to take.

• When an action has consequences, Iris shows a confirmation card and stops on a "Waiting for your confirmation": nothing happens until you press Confirm (and it never runs anything you don't have permission for).
• Iris also never invents data — if it can't find something, it says so — never promises future features, and never pretends something was done.

What it remembers: project memory

Iris doesn't need you to repeat what your team has already agreed. In every conversation it reads the project's project memory, the agreements, tasks, key dates and notes your team keeps about how the project is run, and takes them into account when it answers and prepares work. It treats them as context, not as orders, and checks anything that matters against your live data.

When something worth keeping comes up ("remember that we never send on Sundays"), Iris proposes the change on a Changes to the project memory card that shows exactly what it adds, changes or removes. Nothing is saved until you confirm it, and the change applies to your whole team. Iris only proposes these changes to people whose role can edit project memory. How boards work, who can edit them and what they never accept is explained in Project memory.

How it handles your data

How much of your audience the assistant actually sees is decided by the same setting that governs everything outside the dashboard: your project's access type. It is explained in one place — Third party data access — and that is the page to read; it isn't repeated here.

What is specific to the assistant is that it tells you which mode it is in, in a line under the message box, with a Settings link beside it that takes you straight to the setting. When the project's access type is anonymized, the assistant keeps counting, segmenting and measuring without being able to identify anyone; when it is blocked, it receives no project data and says so instead of guessing.

Where it appears

Iris is a single assistant you open from the panel, and it adapts to where you are. When you open it on a specific screen, it arrives already focused on it: on a channel page (SMS or RCS) it knows your senders, which countries you can reach and what's under review, and can take you to your senders or walk you through what's left to send. The same Iris helps you in Settings, on your brand profile and across the rest of your project, and you can also open the chat to ask about everything at once.

In the flow editor, Iris works on the flow in front of you: it reads the draft as it is on screen, explains what it does or why a step needs attention, and proposes changes that you confirm before they land on the canvas, unsaved. See Iris in the flow editor.

Availability

The assistant is included in all plans, including the flow assistant in the flow editor. On the Free plan it works with a very limited usage allowance, enough to try it. It may still be unavailable in some situations: an administrator can turn it off for the organization, or it can be temporarily out of service for maintenance.

Want your own agent instead?

The assistant on this page lives inside Instasent. If you'd rather connect your own external agent — Claude, ChatGPT or Gemini — and operate Instasent from there, that's Connected apps.