The AI Summer Autopilot: Six Prompts to Keep a Business Running While You're Offline

TLDR: The AI Break, written by Rui Sousa and Luis Sousa, lays out a "Summer Autopilot" built from six chained prompts. Each prompt feeds the next, starting with a Coverage Auditor that maps everything depending on you and ending with a Re-Entry Brief that compresses your first day back into one page. The pitch is 90 minutes of setup instead of two days of panic-batching the night before your flight.

Summary: The premise is one every solo operator and small team recognizes. June arrives, someone says they cannot really disconnect, and two weeks off turns into a quiet pipeline, dark content channels, and an inbox that reads like a horror movie. The authors frame the usual fix, panic-batching everything the night before a flight, as the actual problem. Their answer is to treat vacation prep as a system rather than a sprint.

The system is six prompts, each one consuming the output of the one before it. The Coverage Auditor maps everything that depends on you. The Content Batcher pre-builds a month of posts and emails. The Inbox Shield writes autoresponders that triage instead of just bouncing people. The Lead Keeper holds the pipeline warm on a schedule. The Delegation Pack turns tasks into handoff briefs with decision rules attached. The Re-Entry Brief takes the chaos of your return and reduces it to a single page. The chaining is the interesting part. You are not firing off isolated prompts, you are running a small pipeline where the categorization done in step one drives the work in every later step.

The first prompt is the one I would actually pay attention to. The Coverage Auditor casts the model as a Chief Operating Officer preparing the business for your absence. You give it your dates, your availability level (fully offline, thirty minutes a day, or emergencies only), and a raw brain dump of every task, client, deal, and recurring commitment. The model then extracts every open loop and assigns each one exactly one of four strategies: automate, batch ahead, delegate, or pause. It flags the three highest-risk items, the ones that could cost revenue or a relationship, and it asks up to five clarifying questions before producing a Coverage Map table with item, category, strategy, risk level, and a definition of what "handled" actually looks like. The authors add a sharp tip: feed in your calendar and your recurring invoices, because the things that burn you are never the things you remembered.

The second prompt, the Content Batcher, takes the items tagged "batch ahead" and turns them into a publishing calendar plus drafts for the full window, leaning on evergreen topics so nothing feels stale if the news moves. From there the newsletter goes behind a paywall, so the remaining four prompts are described by purpose but not shown in full. That is worth naming plainly. The free portion gives you the architecture and two working prompts, and the rest is a paid upsell along with a course, Chrome extensions, and a SaaS discount vault.

What the authors avoid thinking about is reliability. "Smart autoresponders that protect deals instead of killing them" is a nice line, but an LLM-written autoresponder firing unsupervised against live client email is exactly the kind of unattended automation that goes wrong while you are unreachable at the beach. There is no mention of guardrails, no human-in-the-loop checkpoint, no fallback when a draft is wrong. The strongest idea here is not the AI at all. It is the four-way triage of every commitment into automate, batch, delegate, or pause. That is a forcing function that works with a spreadsheet and a clear head. The model just speeds up the sorting.

Key takeaways:

The system is six prompts chained together, where the Coverage Map from prompt one feeds categorization into every later step.
The Coverage Auditor sorts every commitment into automate, batch ahead, delegate, or pause, then flags the three items that could actually cost money or relationships.
Including your calendar and recurring invoices in the brain dump catches the obligations you would otherwise forget.
The Content Batcher favors evergreen topics so pre-written posts do not go stale during a multi-week absence.
The free email covers the framework and first two prompts, the remaining four sit behind a paid subscription.

Why do I care: This is a business and operations story more than an engineering one, but the chaining pattern is the part worth stealing. A pipeline where one prompt produces structured output (a categorized table) that becomes the typed input to the next prompt is exactly how I would build a real agent workflow, and seeing it spelled out for a non-technical audience is a useful reminder that "agentic" usually just means good handoffs between steps. The reliability gap is the real lesson for anyone building this for production: turning AI loose on live client email without a human checkpoint or a kill switch is how you come back from vacation to a bigger mess than you left. Borrow the four-way triage, keep a person on the approve button for anything that touches a customer.

Tutorial: Build an AI System That Keeps Working While You Take a Break