AI Copilot
Copilot
The Copilot is an in-thread assistant for your support team. It researches a conversation, drafts replies in context, and can act on connected tools with your approval. It works alongside an agent inside a thread rather than replying to customers directly.
Where to find it
Settings > Features > AI > Copilot.
The page opens with a short capability list, then either the configuration controls or an upgrade prompt, depending on your plan.
Controls on this page
Capability list
At the top, the page states what the Copilot does:
Summarizes and drafts replies in context.
Grounded in your help center and past threads.
Acts on Stripe and your database with approval.
This list is informational and has no controls.
Custom prompt
Shown only when your plan includes the Copilot. This is a textarea where you provide instructions and context the Copilot uses when it researches and drafts. It saves on blur. See the AI custom prompt article for how the field shapes behavior, including the "Suggest from past replies" helper that appears once you have enough answered threads.
PostgreSQL
Shown only when your plan includes the Copilot. A row linking to Settings > Integrations > Postgres. Connecting a database lets the Copilot answer with read-only queries against your data.
Upgrade to enable Copilot
Shown when your plan does not include the Copilot. It states that the Copilot is available on the Scale plan and offers an "Explore plans" button that opens Settings > Plans. While this prompt is showing, the custom prompt and PostgreSQL controls are hidden.
How to use the Copilot
Confirm your workspace is on the Scale plan or higher. If not, open Settings > Features > AI > Copilot and use "Explore plans" to upgrade.
Open a thread in your inbox and invoke the Copilot from within the conversation to have it research and draft a reply.
Review the draft it produces and send or edit it. Actions on connected tools like Stripe or your database run only after you approve them.
To ground it further, add a Custom prompt on this settings page and connect PostgreSQL for read-only data lookups.
Viewing sources
The Copilot grounds its drafts in your help center and past threads, but the inbox does not show agents which sources it used for a given draft. Source citations currently render only on the end-user side, in the widget, when the Productlane Agent answers a customer directly.
There is no agent-side view of Copilot's sources in the inbox today. This is a known gap and a feature request is tracked internally, but there is no workaround or delivery timeline yet.
Notes
Plan gating: the Copilot requires the Scale plan and up. On lower plans, the page shows the upgrade prompt and the configuration controls stay hidden.
The Copilot is grounded in your help center and past threads by default. Connecting PostgreSQL adds read-only database queries as a source.
When it points to a specific button, tab, or field from a docs article, it can show a cropped screenshot from that article's image alongside its answer instead of describing the spot in words alone.
Tool actions (for example, Stripe operations or database reads) require your approval before they run. Each one is logged in the thread's activity timeline with the arguments sent to the tool, not the raw response it returned. See "Inspecting tool calls" in the Productlane Agent article for what these logs contain and how to troubleshoot an answer that used one.
The custom prompt field saves automatically on blur.
Source visibility: agents see the Copilot's draft in the inbox but not a list of the sources behind it. Customers see cited sources only when the Productlane Agent answers them in the widget.
Was this helpful?