OnRamp AI on the customer portal

Updated

OnRamp AI is the chat assistant your customers see in the customer portal. It greets them, points them to their next task, answers questions about their project, and can invite teammates, comment on tasks or reassign tasks once the customer confirms.

This feature is available when it's turned on for your organization. If you don't see it, contact your OnRamp team.

When to use it

Turn on OnRamp AI for projects where customers need help finding their next step or understanding what a task asks of them. It works from the project's tasks, assignments and comments, so it's most useful when task names and assignments are kept current.

How it works

Turn it on for a playbook or project

Playbooks and projects each have an Ask OnRamp AI Agent toggle in their settings:

  • In a playbook's settings, it lets projects created from that playbook allow the portal agent in the customer portal.

  • In a project's settings, it lets that project's customer portal display the portal agent.

Super Admins, Creators and Integrators see the toggle. On an archived project, or for someone without permission to update portal settings, the project toggle is disabled.

Defaults:

  • Existing projects and playbooks start with the toggle off.

  • New playbooks start with it on.

  • A new project takes the setting of the playbook it's created from.

A customer sees the chat only when the project's toggle is on and the project's portal is live. They must also be an active customer user on the project. A published Portal Studio portal can also hide the assistant.

The first visit

The chat panel opens by default. On a customer's first visit, the welcome message introduces OnRamp AI by name. When there's a task they can start, it names the task and links to it. When there's no task to point to, it offers suggested prompts.

When the next task isn't ready yet, the message is text only, with no task link. The same kind of message appears on a return visit when the next task isn't ready:

  • If the task is waiting on the onboarding team or a teammate, it tells the customer to leave a comment on the task and @mention them.

  • If the task is owned by another team member, it says the task isn't the customer's next step and tells them to comment and @mention the task's owner.

  • If comment drafting is on, it offers to draft that comment instead.

  • If the reason isn't known, it tells them to check with their onboarding team.

Comment drafting, where OnRamp AI writes a comment for the customer to review and send, is a separate option turned on per organization.

Suggestion chips sit under the message so customers can ask a follow-up without typing. When the next task isn't ready and isn't assigned to the customer, one chip reads Who needs to act first? The second chip depends on the situation:

  • Draft a comment for an update, when comment drafting is on.

  • Leave a comment for an update, when a teammate is holding the task up.

  • Ask my onboarding team, when the onboarding team is holding it up or the reason isn't known.

Return visits

When a customer comes back, the message leads with what matters most:

  • Overdue task: names an overdue task assigned to them and says it's overdue.

  • Newly unblocked task: a task that was waiting and is now ready.

  • New assignment: a task newly assigned to them.

  • New comment: a task with a new comment.

  • Progress: when tasks were completed since their last visit, it notes that progress happened and names the next task.

  • Next task: when nothing else changed, where to pick back up.

The message links to what needs attention, and the suggestion chips change to match.

Chips on task pages

On a task page, the chips are What does this task need from me?, What is the broader context? and Who is the owner of this task? On a task that isn't the customer's next task, they're Why is this blocked?, Who needs to act first? and When can I expect this to clear? The chips update as the customer moves between pages.

Questions

Customers can type questions into the chat panel, such as what's assigned to them, what's overdue, who's on their team, or what a task is asking. The input is unavailable while OnRamp AI is responding.

When OnRamp AI can't answer, it says it doesn't have that information and won't guess. By default, it points the customer to the task's comments and tells them to @mention their project owner. With comment drafting on, it offers to draft a comment to the onboarding team for them to review and send.

Actions and slash commands

The + button in the chat input shows the available actions. Typing / opens the slash command menu, which filters as you type.

Command

What it does

/invite

Adds someone to this project

/comment-task

Drafts a comment for the current task

/reassign

Chooses a new owner for the current task

/comment-task and /reassign need a task open. Without one, the menu shows Open a task to comment on it. or Open a task to reassign it.

With comment drafting on and no task open, OnRamp AI suggests the most likely task for the customer to confirm, then drafts a comment on it. It asks the customer to confirm before it submits anything.

The chat panel

The panel can float or dock to the left or right, and the customer's choice is saved. Docking is available only when the browser window is wide enough, about 960 pixels. In a narrower window, the panel floats. Each response asks Was this response helpful? so customers can give feedback.

Tone and knowledge

To set the agent's tone and what it knows about your company, go to Settings > Aero > Portal agent.

Limits

  • OnRamp AI doesn't surface internal tasks.

  • It doesn't make up answers when it doesn't have the information.

  • It acts only when the customer starts the action, and only after they confirm.

  • The chat panel doesn't load in portal preview mode, so you won't see it when you preview a project's portal.

More in OnRamp Aero