Adlyse AI

The in-app conversational assistant. Press ⌘K (Ctrl+K on Windows) from anywhere in the app to open the chat drawer, or click the “Adlyse AI” button in the sidebar.

What it can do#

Adlyse AI has the same tools exposed by the MCP servers, plus a few extras that only make sense in the in-app context (e.g. visual creative analysis with pixels injected into the LLM’s vision context). Typical asks:

• “Summarize last week’s performance for my top account.”
• “Which campaigns should I pause to cut waste?” → follows up with a preview diff and a confirm button.
• “Generate the monthly performance report for these three accounts.”
• “Walk me through what changed since Tuesday.”

Skills and guidelines#

The assistant is shaped by:

• Guidelines — always-on brand / platform rules that are injected into every turn’s system prompt.
• Skills — task-specific playbooks that get activated mid-conversation when the task calls for them.

This means the same question can produce different answers in different orgs — the assistant’s tone, what it emphasizes, and which rules it cites are all driven by the org’s own configuration.

Writes are always previewed#

Whenever the assistant proposes a change (pause, budget, targeting), the UI shows you the before/after diff before asking you to approve. On approval, the change is applied and logged to the decision log. You can revoke approval at any step.

MCP equivalence#

Everything the in-app assistant can do, an external AI client can do through the MCP servers — the assistant and external callers share the same tools, scoped to the same organization by the same API key mechanism.

The assistant’s advantage is that it already has your session context (the account you’re looking at, the period you selected, the entity you last clicked) — so “pause this campaign” with a campaign row highlighted works without you having to re-specify.