MODE: Standalone Retail AI Automation situation: Small gift shop using AI for product descriptions and customer messages, with unclear tool/data practices and no output review; seeking clearer rules over four months. 🔍 Auto-detected: Priority axes A6 Human Oversight, A9 Data Privacy, A10 Accuracy Checks, A1 Use Case Focus, A5 Pilot Scope, target gaps A1 Use Case Focus, A6 Human Oversight, A9 Data Privacy, A10 Accuracy Checks (based on: tools were chosen for popularity, outputs go unchecked, data handling is unclear, and usage needs clearer rules) — override anything above by specifying it manually in your brief. TOP-3 DIRECTIONS: 1️⃣ A1 Use Case Focus + A6 Human Oversight + A9 Data Privacy + A10 Accuracy Checks + A5 Pilot Scope + A2 Process Clarity (k=6) — Clear rules with controlled review and testing 🟢 BEST-FIT — READY-TO-USE PROMPT BELOW Target Set covered: Yes; Compatibility: Compatible — A1 fits the small scope and customer contact can remain supervised Core: A6 Human Oversight + A10 Accuracy Checks; Weak link: A5 Pilot Scope + A2 Process Clarity 2️⃣ A1 Use Case Focus + A6 Human Oversight + A9 Data Privacy + A10 Accuracy Checks + A4 Task Fit + A7 Staff Skills (k=6) — Practical task boundaries with trained human review Target Set covered: Yes; Compatibility: Compatible — A4 suits routine shop tasks while A7 supports manageable staff adoption 3️⃣ A1 Use Case Focus + A6 Human Oversight + A9 Data Privacy + A10 Accuracy Checks + A3 Data Readiness + A11 Tool Integration (k=6) — Stronger data handling and workflow consistency Target Set covered: Yes; Compatibility: Partial — A3 and A11 favor broader setup, needing adaptation for one small shop ──────────────────── 🟢 READY-TO-USE PROMPT (based on 1️⃣): Give me a structured discussion aid for improving AI and automation use in a small gift shop with one owner and one helper. The current use includes AI-assisted product descriptions and customer messages. The main situation is that tools were selected because they were popular, outputs are not routinely checked, and it is unclear where business or customer data goes. The goal is to establish clearer rules for AI use over the next four months. Do not assume that any particular software, model, platform, supplier, or brand should be retained or adopted. Start by restating the situation in neutral terms and separate known facts from assumptions. Treat the four-month period as the working decision window. Mark every assumption explicitly. Never invent figures, volumes, costs, performance results, customer numbers, or technical capabilities. Do not choose an option, supplier, product, platform, or brand for me. Do not set prices or make customer promises. Do not forecast savings, sales, accuracy, or any other business outcome. Treat the result as a discussion aid rather than professional advice. A1 Use Case Focus: Work with me to define the specific retail tasks for which AI is actually useful, rather than starting from tools. Identify the purposes of product-description drafting and customer-message assistance, then separate suitable tasks from tasks that should remain fully human. For each task, define the intended outcome, who remains responsible, what information the assistant may use, and what would make the task unsuitable for AI assistance. Run this readiness check together with me and record unresolved questions rather than filling gaps with guesses. A6 Human Oversight: Establish a review routine for AI-generated product descriptions and customer messages. Define which outputs require human review before use, what the reviewer checks, when an output must be rewritten or rejected, and who is accountable for the final customer-facing result. Consider a lightweight review process appropriate for one owner and one helper. The routine must not assume that AI output is correct merely because it sounds plausible. Run the readiness check with me and identify where human review is currently missing. A9 Data Privacy: Help me map what kinds of business and customer information are entered into AI-assisted workflows and what information should not be entered without appropriate review. Identify questions about storage, retention, access, sharing, deletion, and handling of customer messages. Do not make legal, privacy-law, tax, or regulatory conclusions. Instead, identify questions that should be taken to a qualified legal or data-protection adviser. Run this readiness check with me using categories of information rather than unnecessary personal details. A10 Accuracy Checks: Design a simple way to detect, record, correct, and revisit errors in AI-generated descriptions and customer replies. Define examples of factual, tone, product, pricing, availability, and customer-service errors that should be checked where relevant, without assuming which errors actually occur. Explain how a small shop could review a limited sample during the pilot and decide whether the checking process itself needs adjustment. Do not assign an accuracy rate or predict how accurate the system will be. Run this readiness check with me and leave unknowns explicitly marked. A5 Pilot Scope: Compare a small controlled pilot with a broader immediate rule change. The pilot should limit the number of tasks and workflows being tested, define what human review is mandatory, and establish a clear start and revisit point. Do not predict savings, sales, accuracy, or workload improvements. A pilot may be considered ready only when task boundaries, oversight, data handling questions, and error checks are sufficiently understood. A2 Process Clarity: Turn the agreed rules into short, repeatable workflow steps that the owner and helper can follow consistently. Document who performs each step, what gets reviewed, what happens when information is uncertain, and when work stops for human judgment. Avoid creating procedures that depend on undocumented knowledge or assumptions about how a particular tool works. Use these stages: Stage 1 — Map current AI-assisted tasks and identify intended uses. Go if each proposed use has a clear task purpose and responsible human owner. No-go if the purpose or responsibility remains unclear. Stage 2 — Establish review, data-handling, and error-checking rules. Go if the rules can be followed consistently and unresolved professional questions have been identified. No-go if customer-facing outputs would remain unchecked or data handling is materially unclear. Stage 3 — Run a limited pilot with defined tasks and documented review. Go if the pilot boundaries and stop conditions are understood. No-go if the pilot would expand usage before the controls are ready. Stage 4 — Revisit the evidence and decide whether to continue, change, pause, or stop the approach. Do not make that decision for me. For an IT or security adviser, ask what data enters each workflow, where it may be stored or transmitted, what access controls exist, what retention or deletion settings matter, and what integration or account-security issues require attention. For a legal or data-protection adviser, ask which customer or business information requires special handling, what disclosures or permissions may be relevant, what retention or deletion practices should be reviewed, and what obligations apply to the intended use. Do not answer these questions yourself. For an automation specialist, ask how the proposed workflows could be documented, where human review should occur, how errors could be logged, what existing systems need to connect, and what should remain manual. List risks including unchecked customer-facing output, inappropriate data sharing, unclear responsibility, inaccurate product information, inconsistent staff practices, excessive workflow complexity, and expanding the pilot before controls are ready. For each material risk, identify a revisit point or a stop condition. If a professional adviser identifies a material concern, pause the affected workflow until it has been addressed. For the first 30 days, or until the four-month decision date if it comes sooner, create a practical next-steps list covering task mapping, review rules, data-flow questions, error logging, pilot boundaries, staff agreement, adviser questions, and the first revisit checkpoint. Keep the list feasible for one owner and one helper. At the end, provide a concise comparison of the viable approaches, the unresolved questions, the proposed gates, and the evidence still needed. Do not declare a winner. Clearly distinguish facts supplied by me, assumptions, open questions, and decisions that remain mine. Negative prompt / avoid: choosing tools or vendors, unchecked AI output, invented figures, privacy or legal conclusions, forecasts of savings sales or accuracy, customer promises, hidden assumptions, unnecessary personal data, immediate full rollout, and treating fluent AI text as verified fact. ──────────────────── ℹ️ "This tool compares retail AI automation directions and builds a discussion prompt; it does not choose for you, pick tools, give legal, privacy, tax, or financial advice, or predict savings or how a change will turn out." Assumptions made: - No assumptions were needed. ❓ Want a different direction? Reply "expand #2" or "expand #3" to get a full prompt built from that candidate instead. 💡 Want to take this further? All three tools below are built for Claude. - Not sure your brief is ready? Search "Universal Query Auditors" on PromptBase to audit and rewrite your request before comparing directions. - Have a clear brief already and just want one direct prompt? Search "Universal Instruction Generator" for a single-pass generator. - Got a finished prompt and want it reviewed? Search "Universal Instruction Auditors" to check it against a full quality checklist before use. ⚙️ "Built for ChatGPT; other AI models, providers, and versions are untested. Use at your own discretion. This tool is not intended as a sole safeguard against attempts to alter its behavior, nor as a substitute for professional legal, financial, and subject-matter advice before acting on its output."