Let AI build your automations

An automation answers comments on your Instagram and Facebook posts for you, around the clock. In Exoconnect you don’t write automations by hand. You say what should happen, in your own words, and Exi or your own AI agent writes it. Exoconnect checks every automation before it can run, the agent tests it, and you decide when it goes live.

What an automation can do

When a new comment arrives on the accounts it watches, an automation can:

• Reply publicly under the comment.
• Send a private message to the person who commented. On Instagram this is text. On Facebook it can have up to three buttons, or be one image, video, audio clip or document.
• Match by words, like “price” or “link”, or by meaning, like “people asking where to buy” or “someone with a problem with their order”. Matching by meaning is on every paid plan, from Starter.
• Look at the post the comment is on: its caption, its hashtags, or tags you gave it. So one automation can answer only on your giveaway posts, or only on certain Reels.
• Wait first, from a minute to 12 hours, and stay quiet if someone on your team already answered.
• Send a form link made for that one person, so their answers land on their contact.

Each comment gets at most one public reply and one private message. An automation answers comments that arrive after it is turned on, within 24 hours of each comment.

For the common cases, such as sending a link to everyone who comments a word, Recipes do it without an agent; see the recipes cookbook. Ask an agent when you want something a recipe can’t express, such as different answers for different questions, or rules about which posts to answer on.

Ask for it

Open Assistant, or your own agent, and describe the result you want. For example:

• “When someone asks about price on any Reel, reply ‘Sent you a DM!’ and message them our price list.”
• “On posts with #giveaway, reply to every entry with a thank-you. Don’t answer people who only tag friends.”
• “If someone complains about an order, don’t reply publicly. Message them to say we’re sorry and will help, and only if nobody on the team answered within an hour.”
• “Answer questions about sizes with our size guide, and questions about shipping with our delivery times.”

You don’t need to name a format or a setting. Say who should get what, and where.

What the agent does

Exi works in this order, and you can ask your own agent to do the same.

1. Looks first. It reads the comments on your recent posts to see how often the pattern really happens, and checks whether an automation or recipe already does this job. If one does, it changes that one rather than adding another.
2. Writes the words in your voice. It drafts the replies to match how you already answer people, and shows them to you.
3. Builds the automation. Exoconnect then checks it before saving it (see below). If a check fails, nothing is saved, and the agent fixes it.
4. Tests it. It tries the automation on real comments from your accounts, or on made-up ones: comments that should get an answer, ones that shouldn’t, and ones in between. A test sends nothing. It only shows what would have been sent.
5. Asks you to turn it on. In the default Auto mode, turning an automation on waits for your approval, because from then on it replies to people as you. You approve it under Drafts in the Assistant’s sidebar.
6. Checks back. After it has run for a while, ask the agent how it went. It reads what the automation sent on each comment and what failed, and fixes what went wrong.

The checks Exoconnect runs

Every automation passes these before it can be turned on, whoever wrote it:

• It only uses your accounts. Every account it names must be a connected Instagram account or Facebook Page in this workspace.
• It’s well formed. Exoconnect checks the automation for mistakes before it runs anything.
• It survives sample comments. Exoconnect runs it on comments made from its own words: every keyword it looks for, in upper and lower case, and, where the automation cares, posts with and without their tags and comments already answered or not. It must not crash, and every reply must be one Instagram or Facebook would accept.
• Its meaning questions mean what they should. An automation that matches by meaning carries example comments with the answers it expects. Each is asked of the real AI model; one wrong answer stops the automation from being saved.

The checks prove that the automation works and sends nothing Meta would refuse. They can’t know what you meant, which is why the agent shows you its replies and its test results before you turn it on.

Find it, watch it, turn it off

Automations an agent built appear under Recipes, next to the ones you made yourself. Opening one says This one is written in code: it was made by an agent, so the recipe builder doesn’t edit it. From there you can:

• Turn off an automation that’s on, or Turn on one that’s off.
• See its runs: each comment it picked up, what it did (replied, sent a message, no reply needed, failed), on which account, and when.
• Delete it, once it’s off.

You can also turn any automation off straight from the Recipes list. Every reply an automation sends is in the workspace’s history, under Settings, then Activity.

To change an automation, ask the agent. It builds and checks a new version, and the old one keeps answering until you approve turning the new one on.

Who it acts as

Turning an automation on hands it your permission to reply. It keeps answering after you sign out. If you later lose permission to reply on that account, or leave the workspace, it stops acting. Every run is recorded under your name.

Plans and costs

• Automations are unlimited on every plan. Your plan counts runs: a run counts only when the automation sends something. Free includes 2,000 runs a month, Starter 25,000 and Pro 100,000.
• Matching by meaning is on every paid plan, from Starter: a small AI call reads each comment, and the examples are checked each time the automation is saved.
• Who writes it: Exi is included from Starter. An outside agent can build automations on any plan, Free included.

See every limit on the pricing page. To learn more about matching by meaning, read Answer comments by meaning.