The Assistant is a chat inside WallyPilot where you ask questions about your Walmart advertising the way you would ask a colleague: "Which keywords wasted the most spend last month?" "Show spend by match type for the last 30 days." "Cut bids on keywords under 2x ROAS."
Every answer comes from your own synced campaign data — the same numbers you see on your dashboard, not generic advice. When you ask it to change something, it drafts the change for you to review. It never touches your account on its own.
The most important thing to understand about the Assistant is what it cannot do. It cannot edit a bid, move a budget, pause a campaign, or block a search term. The only thing it can do is draft a change and show it to you.
Each draft shows a before/after for every keyword or campaign it touches. Nothing reaches Walmart until you click Approve — and larger changes ask you to type a confirmation first, so a stray click cannot apply one.
After a change applies, you have a 24-hour window to undo it and restore the previous values. Drafts themselves also expire after 2 hours, so you can never approve a diff describing numbers that have since moved.
Even an approved change is bounded. The caps below are enforced in code, not by the AI — a draft outside them is refused with the reason spelled out.
The Assistant is at its best on six kinds of question. Copy these and adjust the numbers to your account.
Find where money went without producing orders.
"Which keywords wasted the most spend last month?" · "Show me search terms with clicks but no orders in the last 30 days"
Slice performance any way you can say out loud, and get it charted.
"Show spend by match type for the last 30 days" · "Chart ROAS over time for my top three campaigns"
Turn a rule of thumb into a reviewed draft in one message.
"Cut bids on keywords under 2x ROAS" · "Lower bids 20% on keywords with ACOS over 50%"
Give room to what is working; take it from what is not.
"Raise the daily budget on my best-performing campaign by 50%" · "Which campaigns keep hitting their daily budget cap?"
Queue blocks for the terms taking money without converting. These land in the same reviewed queue as the Negative Keywords tool, with the same guardrails — a term that ever produced an order is never blocked.
"Queue negatives for search terms with 10+ clicks and zero orders"
When something looks empty or stale, ask why before assuming the worst. The Assistant checks each dataset's freshness and the nightly sync record, and tells "you spent nothing" apart from "this report stopped syncing".
"Why does my keyword data look stale?" · "When did my account last sync?"
Walmart's reporting takes roughly two days to settle, so the Assistant treats "two days ago" as the freshest reliable day. This is a Walmart-side lag, not a sync problem — asking about yesterday will get you an honest "too early to say" rather than half-settled numbers.
Walmart's APIs carry no cost data — no cost of goods, referral fees, or fulfillment costs — so no tool reading them can truthfully tell you whether a sale made money. The Assistant talks in ROAS and ACOS and lets you apply your own margins, instead of dressing ad revenue up as profit.
Once you enter unit and fulfillment costs on the Product costs page, the Assistant can report estimated contribution after ad spend — its carefully-named substitute for profit. For items where you have entered no costs, contribution shows as unknown rather than being quietly treated as zero, so a "contribution above zero" question never sweeps in items it knows nothing about.
Each plan includes a number of Assistant messages per day. The counter resets daily, and hitting the limit never blocks anything else in WallyPilot.
The reviewed queue the Assistant's negative-keyword drafts feed into
Read guide →Standing rules that run nightly — for the changes you want made without asking
Read guide →