Prime Big Deal Days 2026: Rerun Your Playbook in 3 AI Check-ins

Reading time: 5 minutes
‍

Check Amazon ads three times on a sale day: early morning for traffic, afternoon for pace and ACOS, evening for winners. Free AI skills keep your playbook running smoothly.

Prime Big Deal Days runs October 6–7, and your team already knows what to do: which clients protect margin, which ones buy share, which numbers to ignore early in the day. The hard part is doing it across every account, every few hours, for two days. Write that playbook into a free AI skill, a saved set of instructions your assistant follows, and it reruns at three check-ins (early morning, afternoon, evening) while you approve every change. For an agency: run Prep on every client tonight, about 5 minutes each plus a report pull from the console, then run the check-ins on your top accounts only.

What do you check at each check-in?

A simple rhythm: three check-ins on a sale day, each with one job.

**Illustration · modelled on a Day-of Pivot report ·** Without an MCP, you pick the rows and get a bulksheet; nothing changes until you upload it. What an afternoon check-in returns: two winners to fund, one unused budget to move, small steps only.

Early morning: live and getting traffic? ASINs and campaigns live, sale budgets in place, bids high enough to win clicks, impressions and clicks coming in. Leave ROAS alone: too few orders have arrived.
Afternoon: on pace and on goal? Budget pace against plan, ACOS against goal, sales against plan. Move budget to winners from campaigns that can't spend theirs. Leave keywords under 20 clicks alone.
Evening: what's working, and what's next? Back the winners, make the final push for the next peak, shift budget quickly. Don't judge the last three hours yet.

Try this prompt: “It's the afternoon of Prime Big Deal Days. Here is my hourly Campaign report for account [X]. Show budget pace, ACOS against our [30]% goal and sales against plan, and suggest moving up to $[200] a day from campaigns that can't spend theirs to winners that will run out before the evening peak.”

Why does the obvious read go wrong on deal day?

Deal day changes three things at once: clicks cost more, orders arrive hours after the click, and deal prices lower order value.

“94% of budget spent, raise everything.” Only winners under your ACOS limit need more.
“Zero orders since this morning, cut bids.” Under 20 clicks, it's too early to judge.
“ROAS is down, something's wrong.” Often clicks cost more while total sales are well above normal.

You know your clients better than AI. Why use it at all?

A skill lets your AI assistant run your judgement instead of replacing it. On its own, an assistant gives generic advice. A skill is where you write your playbook down once, and the assistant reruns it for every client, at every check-in.

Numbers to ignore: “Don't judge a keyword under 20 clicks or the last 3 hours.”
Each client's strategy: “[Client A] protects margin: ACOS cap [25]%. [Client B] buys share: up to [45]%.”
Brand rules: “Never cut brand terms.”
Approval lines: “Budget moves over $[500] a day go to the account lead.”

Try this prompt: “Here is our deal-day SOP [attach]. Add it to the Tentpole skills as our house rules: keep our thresholds, client strategies and approval rules exactly as written, and ask me about anything the SOP leaves open.”

What do the free Tentpole skills do?

The free Tentpole skills are a ready-made event playbook: four AI skills that turn your Amazon Ads console reports into a plan and an upload-ready bulksheet. Xnurta runs an AI platform for Amazon ads; these skills work without it. Skills are a format Anthropic introduced for saved instructions; the download is four instruction files plus setup steps for ChatGPT, Gemini and Claude. You add them once and reuse them for every event.

Skill	When	What you get
Prep	Before the event	Budgets, bids and boosts sized to your budget, plus a rollback file
Day-of Pivot	Each check-in	Small steps judged on settled hours and at least 20 clicks
Cool-down	The morning after	What to keep, and how to unwind event raises
Summary	About two weeks after	What the event really did, plus a Black Friday playbook

**Real screen · money blurred ·** A real Prep run: one sentence, Quick (about 5 minutes) or Full (about 15), then the verdict. Here the budget was spread too thin, and the plan said so.

Prep plans on event cost per click (2× today's by default, or your own last event) and lists the rows that need a quick yes or no, so you approve all or pick the ones you want.

Where do your client reports go?

The skills run inside the AI assistant your agency already uses, so your reports go to that assistant, not to Xnurta. Use the plan your company has approved for client work, usually Claude Team or Enterprise, or ChatGPT Business or Enterprise: Anthropic and OpenAI don't train on those by default. No approved plan yet? Start with an internal brand, or a report with brand names removed. The skills never delete anything, and never plan a change above your ACOS limit or daily total.

How does it work across 15 client accounts?

For 15 clients: about 75 minutes of Prep the night before, plus pulling one or two console reports per client, then check-ins on your top accounts only. Split the clients across your ad managers. Each Prep run leaves its own bulksheet and rollback file; for the rest of your clients, the Prep plan carries the day. Each client's strategy goes into your copy of the skill, so every ad manager runs the event the same way.

Start today: Run the free Prep skill on one client. Nothing changes until you upload the bulksheet.