AI Agent

The AI agent works like a marketer inside the FlareLane console. Open it with the AI button at the bottom right and it looks at the screen you are on, runs the lookups, and builds segments, messages, and journey automations up to a draft. Creating and editing take effect only once you approve them. Review the work, then approve.

For example, try asking for these.

• Summarize the last 7 days of sends by channel and point out anything that stands out.
• Find out why the email click rate dropped this month compared with last month.
• Build a segment for users who added to cart and did not purchase within three days.
• Design an onboarding journey for users who just made their first purchase, and create it as a draft.
• Find two or three data-backed ways to win back dormant users.
• Write two versions of the push copy for this week.
• Build a funnel report from add to cart through purchase on an analytics board.
• Plan next week’s send schedule around the holidays.
• Change the segment conditions on this screen to people who purchased in the last 30 days.

Lookups and analysis run on their own; creating and editing need your approval. Nothing changes until you approve it.

• Once approved it is saved and you land on the screen where it was created. New things are saved inactive, so turning them on is yours.
• One-off and scheduled sends come into the composer as a draft. Review it and press send to send it.
• An automation that is already running, or an in-app message currently being shown, has to be paused before its content can be edited.

Involved work is carried out through a separate plan document. Turn on plan mode under the composer and the agent asks what to settle first, then writes the plan. Even with it off, a request with several steps is organized into one.

• A plan carries the goal, how success will be verified, the audience, the channels, the schedule, and the steps to run.
• Approving it runs the steps in order and stops on the final result screen. The composer is locked while it runs.
• Approval also schedules the plan’s verification as a to-do.
• If it is heading the wrong way, discard it and ask again.

To-dos are where the work the agent will carry out later is collected. When a request needs time to pass before it can be judged, such as a before-and-after comparison, a Keep button appears under the answer. Press it and the item goes on the list.

• Once the time comes, it shows up under the clock icon at the top right of the chat window, on the new chat screen, and on the calendar.
• Run opens a new conversation that checks with the data as of then. The current conversation does not carry over, only the instruction that was kept.
• Discard removes it once there is nothing left to check.

AI Settings at the top right of the chat window is where you set what the agent refers to and how it answers.

• The Project tab holds your service description, business goal, and the guidelines it has to follow. Proposals line up with that goal, and the settings apply to everyone on the team.
• The Personal tab sets the answer tone (Professional, Friendly, Challenger) and your own instructions. Those apply only to your conversations, even within the same project.

• Sending a message, confirming a schedule, activating or deactivating automations and in-app messages, and deleting are all yours to do in the console.
• A scheduled send takes copy changes only, and its send time cannot be moved.
• Sensitive personal data such as email, phone number, name, and date of birth cannot be looked up. Individual user lists and one person’s history are out too, and data comes back only in aggregate.
• The size of a segment’s audience has to be checked on the segment screen.
• Only the project you have selected is visible, and event data is aggregated for up to the last year.
• Setup tasks such as creating or deleting a project, enabling channels, and changing member permissions, along with requesting Kakao Alimtalk template inspection, are only possible in the console.