Ask Lily to inspect live campaigns, explain keyword ROAS, find wasted spend, or prepare a campaign plan. The agent does the account work; you keep the final decision.
Review the last 30 days. Which keywords spent money but produced no paying users?
I found three terms above your spend threshold. I prepared two pauses and one bid reduction for review—nothing has changed yet.
Account context
Live, scoped data
Decision layer
Cost + revenue
Campaign writes
Human approval
01
Account-specific answers
Move beyond generic Apple Ads advice.
A general AI can explain campaign theory. Lily can inspect the account, app, country, campaign, ad group, keyword, and search-term scope you authorized, then answer from the current data.
Compare spend, installs, paying users, and revenue
Trace a performance change to campaigns and search terms
State the date range, account scope, and assumptions
02
Approval-first operation
A recommendation is not a silent campaign change.
Self-serve write tools return a reviewable proposal with the affected objects, current values, proposed values, and rationale. Lily applies it only after confirmation.
Read and analysis workflows can remain read-only
Writes are staged as discrete, scoped actions
Reversible pause behavior is preferred over deletion
03
One agent, several clients
Use Lily in ChatGPT, Claude, Codex, or its own workspace.
The client is the conversation surface; Lily remains the account and tool layer. Apple Ads credentials stay with Lily instead of being pasted into a model prompt.
How it works
A clear path from connection to decision.
01
Connect one app
Authorize the intended Apple Ads account and select the app the agent may analyze.
02
Ask a bounded question
Name the date range, market, objective, and whether the request must remain read-only.
03
Inspect and approve
Review the evidence and exact proposed changes before any self-serve write is applied.
Questions
Before you connect
How is this page different from the Lily Ads homepage?
The homepage introduces the complete Lily Ads product. This page focuses on the self-serve AI Agent: live account analysis, prepared actions, and human approval.
How is the AI Agent different from Autopilot?
The AI Agent works interactively and requires approval for self-serve writes. Autopilot runs recurring actions that are explicitly permitted by a separate program and its guardrails.
Does the AI client receive my Apple Ads credentials?
No. Apple Ads access is authorized to Lily. Compatible AI clients receive scoped Lily tools rather than reusable Apple Ads private keys.
Is RevenueCat or AppsFlyer required?
No. Lily can analyze Apple Ads delivery data without them. A revenue connection adds paying-user, revenue, profit, and ROAS context.