{"id":8390,"date":"2026-09-07T08:50:07","date_gmt":"2026-09-07T08:50:07","guid":{"rendered":"https:\/\/launchlemonade.app\/?p=8390"},"modified":"2026-09-07T08:50:26","modified_gmt":"2026-09-07T08:50:26","slug":"how-to-write-ai-agent-instructions-real-examples-now","status":"publish","type":"post","link":"https:\/\/launchlemonade.app\/blog\/how-to-write-ai-agent-instructions-real-examples-now\/","title":{"rendered":"How to Write AI Agent Instructions With Real Examples"},"content":{"rendered":"<h1 class=\"text-2xl font-bold mt-4 mb-2\">How to Write AI Agent Instructions With Real Examples<\/h1>\n<section id=\"quick-answer\">\n<h3 class=\"text-lg font-semibold mt-3 mb-1\">Quick Answer<\/h3>\n<p class=\"my-2\">To write effective AI agent instructions, define one job, the available context, the process, and the limits.<br \/>\nTell the agent what a good result looks like.<br \/>\nAdd examples for tasks where consistency matters.<br \/>\nThen test the instructions using realistic inputs and edge cases.<\/p>\n<\/section>\n<h3 class=\"text-lg font-semibold mt-3 mb-1\">AI Summary<\/h3>\n<p class=\"my-2\">Strong AI agent instructions read like a clear operational brief. They specify the agent\u2019s purpose, information sources, decision rules, output format, and escalation path. The best instructions do not try to predict every possible situation. Instead, they set useful defaults and state when the agent must pause, ask, or hand work to a person.<\/p>\n<section id=\"ai-summary\">\n<h3 class=\"text-lg font-semibold mt-3 mb-1\">What This Guide Covers<\/h3>\n<ul class=\"list-disc list-outside my-2 space-y-1 pl-6\">\n<li class=\"pl-2\">The difference between a prompt and an agent instruction set<\/li>\n<li class=\"pl-2\">A six-part framework for reliable agent behaviour<\/li>\n<li class=\"pl-2\">Real instruction examples for research, client communication, and reporting<\/li>\n<li class=\"pl-2\">Common mistakes that create vague or risky outputs<\/li>\n<li class=\"pl-2\">A testing method for improving an agent over time<\/li>\n<\/ul>\n<\/section>\n<h2 class=\"text-xl font-bold mt-3 mb-2\">What Makes AI Agent Instructions Work?<\/h2>\n<p class=\"my-2\">Effective instructions reduce ambiguity before the agent starts working. They give it a clear destination, helpful constraints, and a dependable way to handle uncertainty.<\/p>\n<p class=\"my-2\">A chat prompt may be one request. Agent instructions guide repeated work across many requests, documents, tools, and decisions. That makes them closer to a job description and process manual than a one-off question.<\/p>\n<p class=\"my-2\"><a class=\"text-blue-600 dark:text-blue-400 underline hover:no-underline font-medium\" href=\"https:\/\/www.anthropic.com\/engineering\/building-effective-agents\" target=\"_blank\" rel=\"noopener noreferrer\">Anthropic\u2019s guidance on effective agents<\/a>\u00a0makes a useful distinction: use the simplest pattern that can complete the work reliably. Not every task needs an autonomous agent. Some tasks work better as a structured workflow with defined steps.<\/p>\n<div class=\"my-2 overflow-x-auto max-w-full\">\n<div style=\"background-color: #111827; border: 1px solid #374151; border-radius: 12px; overflow-x: auto; max-width: 100%; margin: 16px 0;\">\n<table style=\"width: 100%; border-collapse: collapse; font-size: 14px;\">\n<thead>\n<tr style=\"background-color: rgba(255, 255, 255, 0.08); border-bottom: 2px solid #4B5563;\">\n<th style=\"padding: 14px 16px; text-align: left; font-weight: bold; color: #ffffff; border-right: 1px solid #374151;\">Instruction Element<\/th>\n<th style=\"padding: 14px 16px; text-align: left; font-weight: bold; color: #ffffff; border-right: 1px solid #374151;\">What It Answers<\/th>\n<th style=\"padding: 14px 16px; text-align: left; font-weight: bold; color: #ffffff;\">Example<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr style=\"border-bottom: 1px solid #1F2937; background-color: rgba(255, 255, 255, 0.02);\">\n<td style=\"padding: 12px 16px; color: #ffffff; font-weight: 500; border-right: 1px solid #1F2937;\">Role<\/td>\n<td style=\"padding: 12px 16px; color: #d1d5db; border-right: 1px solid #1F2937;\">Who is the agent?<\/td>\n<td style=\"padding: 12px 16px; color: #d1d5db;\">\u201cYou are a research assistant for a financial advisory firm.\u201d<\/td>\n<\/tr>\n<tr style=\"border-bottom: 1px solid #1F2937;\">\n<td style=\"padding: 12px 16px; color: #ffffff; font-weight: 500; border-right: 1px solid #1F2937;\">Goal<\/td>\n<td style=\"padding: 12px 16px; color: #d1d5db; border-right: 1px solid #1F2937;\">What must it achieve?<\/td>\n<td style=\"padding: 12px 16px; color: #d1d5db;\">\u201cCreate a source-backed briefing for the adviser.\u201d<\/td>\n<\/tr>\n<tr style=\"border-bottom: 1px solid #1F2937; background-color: rgba(255, 255, 255, 0.02);\">\n<td style=\"padding: 12px 16px; color: #ffffff; font-weight: 500; border-right: 1px solid #1F2937;\">Context<\/td>\n<td style=\"padding: 12px 16px; color: #d1d5db; border-right: 1px solid #1F2937;\">What can it use?<\/td>\n<td style=\"padding: 12px 16px; color: #d1d5db;\">\u201cUse the attached client notes and approved research sources.\u201d<\/td>\n<\/tr>\n<tr style=\"border-bottom: 1px solid #1F2937;\">\n<td style=\"padding: 12px 16px; color: #ffffff; font-weight: 500; border-right: 1px solid #1F2937;\">Workflow<\/td>\n<td style=\"padding: 12px 16px; color: #d1d5db; border-right: 1px solid #1F2937;\">How should it work?<\/td>\n<td style=\"padding: 12px 16px; color: #d1d5db;\">\u201cExtract facts, identify gaps, then draft a concise summary.\u201d<\/td>\n<\/tr>\n<tr style=\"border-bottom: 1px solid #1F2937; background-color: rgba(255, 255, 255, 0.02);\">\n<td style=\"padding: 12px 16px; color: #ffffff; font-weight: 500; border-right: 1px solid #1F2937;\">Boundaries<\/td>\n<td style=\"padding: 12px 16px; color: #d1d5db; border-right: 1px solid #1F2937;\">What must it avoid?<\/td>\n<td style=\"padding: 12px 16px; color: #d1d5db;\">\u201cDo not make suitability recommendations or send client emails.\u201d<\/td>\n<\/tr>\n<tr style=\"border-bottom: 1px solid #1F2937;\">\n<td style=\"padding: 12px 16px; color: #ffffff; font-weight: 500; border-right: 1px solid #1F2937;\">Output<\/td>\n<td style=\"padding: 12px 16px; color: #d1d5db; border-right: 1px solid #1F2937;\">What does success look like?<\/td>\n<td style=\"padding: 12px 16px; color: #d1d5db;\">\u201cUse headings, source links, risks, and open questions.\u201d<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<\/div>\n<p class=\"my-2\">Clear instructions also improve review. A manager can see what the agent was intended to do, compare that with its output, and update one rule at a time.<\/p>\n<h2 class=\"text-xl font-bold mt-3 mb-2\">How Do You Write AI Agent Instructions Step by Step?<\/h2>\n<p class=\"my-2\">The best way to write AI agent instructions is to move from job definition to controlled execution. Start simple, then add detail only where real work requires it.<\/p>\n<h3 class=\"text-lg font-semibold mt-3 mb-1\">1. Define One Specific Job<\/h3>\n<p class=\"my-2\">Avoid giving one agent every responsibility in a department. A broad instruction such as \u201chelp with client service\u201d invites inconsistent decisions.<\/p>\n<p class=\"my-2\">Instead, define a narrow, observable outcome:<\/p>\n<blockquote class=\"border-l-4 border-muted-foreground\/30 pl-4 my-2 italic\">\n<p class=\"my-2\">Create a first draft of a meeting brief using the client\u2019s latest notes, portfolio documents, and agenda.<\/p>\n<\/blockquote>\n<p class=\"my-2\">This makes testing simpler. It also clarifies which information the agent needs.<\/p>\n<p class=\"my-2\"><a class=\"text-blue-600 dark:text-blue-400 underline hover:no-underline font-medium\" href=\"https:\/\/platform.openai.com\/docs\/guides\/agents\" target=\"_blank\" rel=\"noopener noreferrer\">OpenAI\u2019s developer guidance<\/a>\u00a0recommends clearly defining agent responsibilities and the tools available for a task. The same principle applies whether your team is coding an agent or configuring one through a no-code platform.<\/p>\n<h3 class=\"text-lg font-semibold mt-3 mb-1\">2. State What the Agent Can Use<\/h3>\n<p class=\"my-2\">An agent should not guess which data is relevant. Name the approved sources, expected inputs, and any source hierarchy.<\/p>\n<p class=\"my-2\">For example:<\/p>\n<blockquote class=\"border-l-4 border-muted-foreground\/30 pl-4 my-2 italic\">\n<p class=\"my-2\">Use the meeting transcript, the client profile, and the approved planning template. If sources conflict, flag the conflict instead of deciding which source is correct.<\/p>\n<\/blockquote>\n<p class=\"my-2\">This simple rule prevents a common failure: the agent blending old and new information without telling the reviewer.<\/p>\n<h3 class=\"text-lg font-semibold mt-3 mb-1\">3. Describe the Working Process<\/h3>\n<p class=\"my-2\">You do not need to prescribe every thought. However, you should specify the checkpoints that matter.<\/p>\n<p class=\"my-2\">A practical workflow might be:<\/p>\n<ol class=\"list-decimal list-outside my-2 space-y-1 pl-6\">\n<li class=\"pl-2\">Read the input documents.<\/li>\n<li class=\"pl-2\">Extract relevant facts.<\/li>\n<li class=\"pl-2\">Identify missing information.<\/li>\n<li class=\"pl-2\">Produce the required draft.<\/li>\n<li class=\"pl-2\">Flag assumptions and unresolved issues.<\/li>\n<\/ol>\n<p class=\"my-2\"><a class=\"text-blue-600 dark:text-blue-400 underline hover:no-underline font-medium\" href=\"https:\/\/ai.google.dev\/gemini-api\/docs\/prompting-strategies\" target=\"_blank\" rel=\"noopener noreferrer\">Google\u2019s prompt design guidance<\/a>\u00a0similarly stresses clear task instructions, formatting requirements, and examples when you need more dependable outputs.<\/p>\n<h3 class=\"text-lg font-semibold mt-3 mb-1\">4. Add Boundaries and Escalation Rules<\/h3>\n<p class=\"my-2\">Boundaries are not negative instructions added at the end. They are part of the job.<\/p>\n<p class=\"my-2\">Include rules such as:<\/p>\n<ul class=\"list-disc list-outside my-2 space-y-1 pl-6\">\n<li class=\"pl-2\">Do not invent facts, citations, or client details.<\/li>\n<li class=\"pl-2\">Do not take irreversible actions without approval.<\/li>\n<li class=\"pl-2\">Ask a clarifying question if required information is missing.<\/li>\n<li class=\"pl-2\">Escalate legal, regulatory, medical, or financial advice requests.<\/li>\n<li class=\"pl-2\">State uncertainty clearly when evidence is incomplete.<\/li>\n<\/ul>\n<p class=\"my-2\">These rules help the agent choose a safe next step rather than producing a confident but unreliable answer.<\/p>\n<h3 class=\"text-lg font-semibold mt-3 mb-1\">5. Specify the Output Format<\/h3>\n<p class=\"my-2\">\u201cWrite a good report\u201d is subjective. A structured result is easier to review and reuse.<\/p>\n<p class=\"my-2\">For a client meeting brief, request:<\/p>\n<ul class=\"list-disc list-outside my-2 space-y-1 pl-6\">\n<li class=\"pl-2\">A five-bullet summary<\/li>\n<li class=\"pl-2\">Key decisions from the previous meeting<\/li>\n<li class=\"pl-2\">Open questions for the adviser<\/li>\n<li class=\"pl-2\">Relevant risks or missing documents<\/li>\n<li class=\"pl-2\">A draft agenda<\/li>\n<\/ul>\n<p class=\"my-2\"><a class=\"text-blue-600 dark:text-blue-400 underline hover:no-underline font-medium\" href=\"https:\/\/learn.microsoft.com\/en-us\/azure\/ai-foundry\/openai\/concepts\/prompt-engineering\" target=\"_blank\" rel=\"noopener noreferrer\">Microsoft\u2019s guidance on prompt engineering<\/a>\u00a0supports this approach. Clear structure reduces interpretation and gives users more predictable outputs.<\/p>\n<h3 class=\"text-lg font-semibold mt-3 mb-1\">6. Provide Examples, Then Test Them<\/h3>\n<p class=\"my-2\">Examples show the level of detail, tone, and judgement you expect. They are especially valuable when the agent must classify, summarise, or draft in a consistent house style.<\/p>\n<p class=\"my-2\">The goal is not to add dozens of examples. Use two or three representative cases, including one edge case.<\/p>\n<h2 class=\"text-xl font-bold mt-3 mb-2\">What Is a Reusable AI Agent Instruction Template?<\/h2>\n<p class=\"my-2\">A reusable template gives teams a consistent starting point. It also makes agents easier to review, maintain, and share.<\/p>\n<div class=\"relative my-2\">\n<pre class=\"bg-muted p-3 pr-12 rounded-lg whitespace-pre-wrap break-words overflow-wrap-anywhere\"><code class=\"block bg-muted p-3 rounded-lg my-2 text-sm font-mono whitespace-pre-wrap break-words overflow-wrap-anywhere\">ROLE\r\nYou are [agent role] for [team or business].\r\n\r\nPRIMARY GOAL\r\nYour job is to [specific outcome].\r\n\r\nWHEN TO ACT\r\nStart when [trigger or user request].\r\n\r\nAPPROVED INPUTS\r\nUse only [documents, systems, tools, or user-provided information].\r\n\r\nPROCESS\r\n1. [Step one]\r\n2. [Step two]\r\n3. [Step three]\r\n\r\nDECISION RULES\r\n- If [condition], then [action].\r\n- If information is missing, [clarify or flag it].\r\n- If sources conflict, [state the conflict].\r\n\r\nBOUNDARIES\r\n- Do not [prohibited action].\r\n- Do not invent information.\r\n- Do not take [sensitive action] without human approval.\r\n\r\nOUTPUT FORMAT\r\nProvide:\r\n1. [Section or field]\r\n2. [Section or field]\r\n3. [Section or field]\r\n\r\nQUALITY CHECK\r\nBefore responding, confirm that you followed the output format and flagged uncertainty.\r\n<\/code><\/pre>\n<p>&nbsp;<\/p>\n<\/div>\n<p class=\"my-2\">This template works because it separates purpose from process. The agent can understand its job before it encounters detailed rules.<\/p>\n<h2 class=\"text-xl font-bold mt-3 mb-2\">How to Write AI Agent Instructions With Real Examples<\/h2>\n<p class=\"my-2\">The following examples show how a vague prompt becomes a usable instruction set.<\/p>\n<h3 class=\"text-lg font-semibold mt-3 mb-1\">Example 1: Research Briefing Agent<\/h3>\n<p class=\"my-2\"><strong class=\"font-bold\">Weak instruction<\/strong><\/p>\n<blockquote class=\"border-l-4 border-muted-foreground\/30 pl-4 my-2 italic\">\n<p class=\"my-2\">Research this company and tell me what matters.<\/p>\n<\/blockquote>\n<p class=\"my-2\">The agent has no audience, source standard, or expected output. It may return a long summary without practical next steps.<\/p>\n<p class=\"my-2\"><strong class=\"font-bold\">Improved instruction<\/strong><\/p>\n<div class=\"relative my-2\">\n<pre class=\"bg-muted p-3 pr-12 rounded-lg whitespace-pre-wrap break-words overflow-wrap-anywhere\"><code class=\"block bg-muted p-3 rounded-lg my-2 text-sm font-mono whitespace-pre-wrap break-words overflow-wrap-anywhere\">You are a business research assistant.\r\n\r\nCreate a concise company briefing for a consultant preparing for a first meeting.\r\n\r\nUse only the company website, official filings, and reputable primary sources.\r\nDo not treat marketing claims as verified facts.\r\n\r\nFollow this process:\r\n1. Identify the company\u2019s products, customers, leadership, and recent developments.\r\n2. Separate verified facts from reasonable inferences.\r\n3. Flag missing information and source conflicts.\r\n\r\nOutput:\r\n- Company overview, maximum 100 words\r\n- Three relevant recent developments\r\n- Likely business priorities\r\n- Five meeting questions\r\n- Sources as direct links\r\n\r\nIf a source is unavailable, say so. Do not fill gaps with guesses.\r\n<\/code><\/pre>\n<p>&nbsp;<\/p>\n<\/div>\n<p class=\"my-2\">This agent has a role, an audience, a source standard, a process, and a usable output.<\/p>\n<p class=\"my-2\">For teams using connected tools, tool access needs equally clear rules.\u00a0<a class=\"text-blue-600 dark:text-blue-400 underline hover:no-underline font-medium\" href=\"https:\/\/www.langchain.com\/blog\/tool-calling-with-langchain\" target=\"_blank\" rel=\"noopener noreferrer\">LangChain\u2019s overview of tool calling<\/a>\u00a0explains why tool use should be explicit and structured rather than assumed.<\/p>\n<h3 class=\"text-lg font-semibold mt-3 mb-1\">Example 2: Client Email Drafting Agent<\/h3>\n<p class=\"my-2\"><strong class=\"font-bold\">Weak instruction<\/strong><\/p>\n<blockquote class=\"border-l-4 border-muted-foreground\/30 pl-4 my-2 italic\">\n<p class=\"my-2\">Reply professionally to clients.<\/p>\n<\/blockquote>\n<p class=\"my-2\">This fails because \u201cprofessionally\u201d means different things to different people. It also offers no rule for sensitive requests.<\/p>\n<p class=\"my-2\"><strong class=\"font-bold\">Improved instruction<\/strong><\/p>\n<div class=\"relative my-2\">\n<pre class=\"bg-muted p-3 pr-12 rounded-lg whitespace-pre-wrap break-words overflow-wrap-anywhere\"><code class=\"block bg-muted p-3 rounded-lg my-2 text-sm font-mono whitespace-pre-wrap break-words overflow-wrap-anywhere\">You are a client communications drafting assistant.\r\n\r\nDraft clear, warm email replies for a professional services firm.\r\nYour drafts are always reviewed by a team member before sending.\r\n\r\nUse the client\u2019s name, the message context, and the firm\u2019s approved tone guidance.\r\nKeep emails under 180 words unless the user asks for more detail.\r\n\r\nDo:\r\n- Confirm the client\u2019s request.\r\n- State the next step and owner.\r\n- Ask one clear question if information is missing.\r\n\r\nDo not:\r\n- Give legal, investment, tax, or compliance advice.\r\n- Promise a deadline that is not stated in the source material.\r\n- Send, schedule, or finalise communications.\r\n\r\nOutput only the draft email and a short \u201cReview Notes\u201d section.\r\n<\/code><\/pre>\n<p>&nbsp;<\/p>\n<\/div>\n<p class=\"my-2\">The key improvement is not the wording. It is the operating boundary. The agent drafts. A person approves.<\/p>\n<p class=\"my-2\">On\u00a0<a class=\"text-blue-600 dark:text-blue-400 underline hover:no-underline font-medium\" href=\"https:\/\/launchlemonade.app\/platform\/teams\" target=\"_blank\" rel=\"noopener noreferrer\">LaunchLemonade\u2019s Teams platform<\/a>, admins can define which agent actions need human review before they run. That is useful for sensitive tasks, such as client emails or final reports.<\/p>\n<h3 class=\"text-lg font-semibold mt-3 mb-1\">Example 3: Monthly Reporting Agent<\/h3>\n<p class=\"my-2\"><strong class=\"font-bold\">Weak instruction<\/strong><\/p>\n<blockquote class=\"border-l-4 border-muted-foreground\/30 pl-4 my-2 italic\">\n<p class=\"my-2\">Create our monthly report.<\/p>\n<\/blockquote>\n<p class=\"my-2\">The agent needs to know which numbers matter, how to treat gaps, and who will read the result.<\/p>\n<p class=\"my-2\"><strong class=\"font-bold\">Improved instruction<\/strong><\/p>\n<div class=\"relative my-2\">\n<pre class=\"bg-muted p-3 pr-12 rounded-lg whitespace-pre-wrap break-words overflow-wrap-anywhere\"><code class=\"block bg-muted p-3 rounded-lg my-2 text-sm font-mono whitespace-pre-wrap break-words overflow-wrap-anywhere\">You are an operations reporting assistant.\r\n\r\nPrepare a monthly performance report for the leadership team.\r\n\r\nUse the approved spreadsheet and meeting notes only.\r\nCheck that all figures include a reporting period and unit.\r\n\r\nProcess:\r\n1. Summarise the five agreed performance metrics.\r\n2. Compare each metric with the previous month.\r\n3. Identify material changes.\r\n4. List unresolved data-quality issues separately.\r\n\r\nOutput:\r\n- Executive summary, maximum 150 words\r\n- Metrics table\r\n- Key changes and likely causes\r\n- Risks and decisions needed\r\n- Data gaps\r\n\r\nDo not calculate figures from incomplete data.\r\nIf a value is missing, write \u201cData unavailable\u201d and identify the owner.\r\n<\/code><\/pre>\n<p>&nbsp;<\/p>\n<\/div>\n<p class=\"my-2\">This agent creates a decision-ready report while protecting against silent assumptions.<\/p>\n<h2 class=\"text-xl font-bold mt-3 mb-2\">Why Do Agent Instructions Need Guardrails?<\/h2>\n<p class=\"my-2\">Guardrails tell an agent when not to continue. They are essential when the work affects clients, compliance, data, or external systems.<\/p>\n<p class=\"my-2\">A useful guardrail is specific and actionable. \u201cBe careful with data\u201d is not actionable. \u201cDo not use client data outside the approved workspace\u201d is clearer.<\/p>\n<div class=\"my-2 overflow-x-auto max-w-full\">\n<div style=\"background-color: #111827; border: 1px solid #374151; border-radius: 12px; overflow-x: auto; max-width: 100%; margin: 16px 0;\">\n<table style=\"width: 100%; border-collapse: collapse; font-size: 14px;\">\n<thead>\n<tr style=\"background-color: rgba(255, 255, 255, 0.08); border-bottom: 2px solid #4B5563;\">\n<th style=\"padding: 14px 16px; text-align: left; font-weight: bold; color: #ffffff; border-right: 1px solid #374151;\">Risk Area<\/th>\n<th style=\"padding: 14px 16px; text-align: left; font-weight: bold; color: #ffffff; border-right: 1px solid #374151;\">Weak Guardrail<\/th>\n<th style=\"padding: 14px 16px; text-align: left; font-weight: bold; color: #ffffff;\">Strong Guardrail<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr style=\"border-bottom: 1px solid #1F2937; background-color: rgba(255, 255, 255, 0.02);\">\n<td style=\"padding: 12px 16px; color: #ffffff; font-weight: 500; border-right: 1px solid #1F2937;\">Missing data<\/td>\n<td style=\"padding: 12px 16px; color: #d1d5db; border-right: 1px solid #1F2937;\">\u201cUse your judgement\u201d<\/td>\n<td style=\"padding: 12px 16px; color: #d1d5db;\">\u201cAsk for the missing document before drafting conclusions.\u201d<\/td>\n<\/tr>\n<tr style=\"border-bottom: 1px solid #1F2937;\">\n<td style=\"padding: 12px 16px; color: #ffffff; font-weight: 500; border-right: 1px solid #1F2937;\">Sensitive action<\/td>\n<td style=\"padding: 12px 16px; color: #d1d5db; border-right: 1px solid #1F2937;\">\u201cBe cautious\u201d<\/td>\n<td style=\"padding: 12px 16px; color: #d1d5db;\">\u201cPrepare a draft only. Human approval is required before sending.\u201d<\/td>\n<\/tr>\n<tr style=\"border-bottom: 1px solid #1F2937; background-color: rgba(255, 255, 255, 0.02);\">\n<td style=\"padding: 12px 16px; color: #ffffff; font-weight: 500; border-right: 1px solid #1F2937;\">Source quality<\/td>\n<td style=\"padding: 12px 16px; color: #d1d5db; border-right: 1px solid #1F2937;\">\u201cUse reliable sources\u201d<\/td>\n<td style=\"padding: 12px 16px; color: #d1d5db;\">\u201cUse primary sources first. Label unverified claims as unverified.\u201d<\/td>\n<\/tr>\n<tr style=\"border-bottom: 1px solid #1F2937;\">\n<td style=\"padding: 12px 16px; color: #ffffff; font-weight: 500; border-right: 1px solid #1F2937;\">Advice risk<\/td>\n<td style=\"padding: 12px 16px; color: #d1d5db; border-right: 1px solid #1F2937;\">\u201cDo not give advice\u201d<\/td>\n<td style=\"padding: 12px 16px; color: #d1d5db;\">\u201cProvide general information only and escalate personalised recommendations.\u201d<\/td>\n<\/tr>\n<tr style=\"border-bottom: 1px solid #1F2937; background-color: rgba(255, 255, 255, 0.02);\">\n<td style=\"padding: 12px 16px; color: #ffffff; font-weight: 500; border-right: 1px solid #1F2937;\">Tone<\/td>\n<td style=\"padding: 12px 16px; color: #d1d5db; border-right: 1px solid #1F2937;\">\u201cBe professional\u201d<\/td>\n<td style=\"padding: 12px 16px; color: #d1d5db;\">\u201cUse plain English, short paragraphs, and no promotional language.\u201d<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<\/div>\n<p class=\"my-2\"><a class=\"text-blue-600 dark:text-blue-400 underline hover:no-underline font-medium\" href=\"https:\/\/www.nist.gov\/itl\/ai-risk-management-framework\" target=\"_blank\" rel=\"noopener noreferrer\">The NIST AI Risk Management Framework<\/a>\u00a0is useful background for teams formalising controls. Its core message is practical: manage risk throughout the AI lifecycle, not only after an incident.<\/p>\n<p class=\"my-2\">LaunchLemonade is built for regulated small and medium businesses that need governance around AI agents. Teams can customise ready-made agents with their templates, tone, source documents, and workflows, or build their own through the no-code builder.\u00a0<a class=\"text-blue-600 dark:text-blue-400 underline hover:no-underline font-medium\" href=\"https:\/\/launchlemonade.app\/platform\/teams\" target=\"_blank\" rel=\"noopener noreferrer\">Explore the Teams platform<\/a>\u00a0to see how role-based access controls, audit trails, PII detection, and approval workflows can support governed agent use.<\/p>\n<h2 class=\"text-xl font-bold mt-3 mb-2\">What Mistakes Make Agent Instructions Fail?<\/h2>\n<p class=\"my-2\">Most instruction failures come from ambiguity, not from a lack of clever prompting.<\/p>\n<h3 class=\"text-lg font-semibold mt-3 mb-1\">Trying to Solve Everything With One Agent<\/h3>\n<p class=\"my-2\">A single agent that researches, drafts, approves, sends, and reports creates too many competing rules. Split the work into separate agents or a defined workflow.<\/p>\n<p class=\"my-2\">This aligns with\u00a0<a class=\"text-blue-600 dark:text-blue-400 underline hover:no-underline font-medium\" href=\"https:\/\/www.anthropic.com\/engineering\/building-effective-agents\" target=\"_blank\" rel=\"noopener noreferrer\">Anthropic\u2019s workflow and agent guidance<\/a>: begin with the simplest reliable structure, then add complexity only when the task needs it.<\/p>\n<h3 class=\"text-lg font-semibold mt-3 mb-1\">Hiding Critical Rules in Long Paragraphs<\/h3>\n<p class=\"my-2\">Long narrative instructions are easy to write and difficult to maintain. Use headings, lists, and decision rules so a reviewer can identify the important constraints quickly.<\/p>\n<h3 class=\"text-lg font-semibold mt-3 mb-1\">Leaving \u201cGood\u201d Undefined<\/h3>\n<p class=\"my-2\">Words such as accurate, concise, professional, and useful need evidence. Define the source standard, word limit, headings, tone rules, and expected audience.<\/p>\n<h3 class=\"text-lg font-semibold mt-3 mb-1\">Forgetting Failure Behaviour<\/h3>\n<p class=\"my-2\">Your agent needs a response for incomplete, conflicting, or sensitive inputs. \u201cAsk\u201d, \u201cflag\u201d, \u201cstop\u201d, and \u201cescalate\u201d are often more valuable than more detail about the happy path.<\/p>\n<h3 class=\"text-lg font-semibold mt-3 mb-1\">Never Testing With Real Inputs<\/h3>\n<p class=\"my-2\">A polished example can hide failures. Test instructions against routine cases, difficult cases, missing data, and conflicting directions.<\/p>\n<p class=\"my-2\"><a class=\"text-blue-600 dark:text-blue-400 underline hover:no-underline font-medium\" href=\"https:\/\/platform.openai.com\/docs\/guides\/evals\" target=\"_blank\" rel=\"noopener noreferrer\">OpenAI\u2019s evaluation guidance<\/a>\u00a0is helpful here. Reliable AI work improves through defined test cases and repeatable evaluation, not one successful demo.<\/p>\n<h2 class=\"text-xl font-bold mt-3 mb-2\">How Should You Test AI Agent Instructions?<\/h2>\n<p class=\"my-2\">Test the agent as you would test a new team process. Use examples drawn from real work, remove sensitive information where needed, and decide what \u201cpass\u201d means before reviewing outputs.<\/p>\n<h3 class=\"text-lg font-semibold mt-3 mb-1\">Build a Small Test Set<\/h3>\n<p class=\"my-2\">Start with 10 to 20 realistic tasks:<\/p>\n<ul class=\"list-disc list-outside my-2 space-y-1 pl-6\">\n<li class=\"pl-2\">A standard request<\/li>\n<li class=\"pl-2\">A request with missing information<\/li>\n<li class=\"pl-2\">Conflicting source documents<\/li>\n<li class=\"pl-2\">An ambiguous request<\/li>\n<li class=\"pl-2\">A sensitive request that requires escalation<\/li>\n<li class=\"pl-2\">A request outside the agent\u2019s scope<\/li>\n<\/ul>\n<h3 class=\"text-lg font-semibold mt-3 mb-1\">Score the Results<\/h3>\n<div class=\"my-2 overflow-x-auto max-w-full\">\n<div style=\"background-color: #111827; border: 1px solid #374151; border-radius: 12px; overflow-x: auto; max-width: 100%; margin: 16px 0;\">\n<table style=\"width: 100%; border-collapse: collapse; font-size: 14px;\">\n<thead>\n<tr style=\"background-color: rgba(255, 255, 255, 0.08); border-bottom: 2px solid #4B5563;\">\n<th style=\"padding: 14px 16px; text-align: left; font-weight: bold; color: #ffffff; border-right: 1px solid #374151;\">Test Question<\/th>\n<th style=\"padding: 14px 16px; text-align: left; font-weight: bold; color: #ffffff;\">Pass Standard<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr style=\"border-bottom: 1px solid #1F2937; background-color: rgba(255, 255, 255, 0.02);\">\n<td style=\"padding: 12px 16px; color: #ffffff; font-weight: 500; border-right: 1px solid #1F2937;\">Did the agent complete the assigned task?<\/td>\n<td style=\"padding: 12px 16px; color: #d1d5db;\">The result addresses the requested outcome.<\/td>\n<\/tr>\n<tr style=\"border-bottom: 1px solid #1F2937;\">\n<td style=\"padding: 12px 16px; color: #ffffff; font-weight: 500; border-right: 1px solid #1F2937;\">Did it use the approved inputs?<\/td>\n<td style=\"padding: 12px 16px; color: #f87171;\">No unsupported claims or irrelevant sources appear.<\/td>\n<\/tr>\n<tr style=\"border-bottom: 1px solid #1F2937; background-color: rgba(255, 255, 255, 0.02);\">\n<td style=\"padding: 12px 16px; color: #ffffff; font-weight: 500; border-right: 1px solid #1F2937;\">Did it follow the output format?<\/td>\n<td style=\"padding: 12px 16px; color: #d1d5db;\">All required headings and fields are present.<\/td>\n<\/tr>\n<tr style=\"border-bottom: 1px solid #1F2937;\">\n<td style=\"padding: 12px 16px; color: #ffffff; font-weight: 500; border-right: 1px solid #1F2937;\">Did it handle uncertainty correctly?<\/td>\n<td style=\"padding: 12px 16px; color: #d1d5db;\">Gaps and conflicts are clearly flagged.<\/td>\n<\/tr>\n<tr style=\"border-bottom: 1px solid #1F2937; background-color: rgba(255, 255, 255, 0.02);\">\n<td style=\"padding: 12px 16px; color: #ffffff; font-weight: 500; border-right: 1px solid #1F2937;\">Did it respect boundaries?<\/td>\n<td style=\"padding: 12px 16px; color: #d1d5db;\">It did not take prohibited or unapproved actions.<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<\/div>\n<p class=\"my-2\"><a class=\"text-blue-600 dark:text-blue-400 underline hover:no-underline font-medium\" href=\"https:\/\/ai.google.dev\/gemini-api\/docs\/evaluation\" target=\"_blank\" rel=\"noopener noreferrer\">Google\u2019s evaluation guidance<\/a>\u00a0recommends using systematic evaluation rather than relying on ad hoc impressions. That is particularly important when multiple people will depend on the agent\u2019s output.<\/p>\n<p class=\"my-2\">Track failures by category. For example, \u201cmissed source conflict\u201d is more useful than \u201cbad answer.\u201d Update the smallest rule likely to resolve that category, then rerun the same test set.<\/p>\n<h2 class=\"text-xl font-bold mt-3 mb-2\">How Should Teams Govern Agent Instructions?<\/h2>\n<p class=\"my-2\">Teams should treat agent instructions as controlled operational assets. Someone should own each agent, approve material changes, and review performance after workflows or source documents change.<\/p>\n<p class=\"my-2\">LaunchLemonade makes this easier for non-technical teams. Its no-code builder lets domain experts create and update agents without engineering support. You can also create structured, multi-step workflows with decision points, tool calls, output formatting, schedules, or event triggers.<\/p>\n<p class=\"my-2\">For consultants and professional-services experts who want to package specialist workflows,\u00a0<a class=\"text-blue-600 dark:text-blue-400 underline hover:no-underline font-medium\" href=\"https:\/\/launchlemonade.app\/platform\/builders\" target=\"_blank\" rel=\"noopener noreferrer\">LaunchLemonade\u2019s Builders platform<\/a>\u00a0is designed for creating no-code agents around a clear client problem.<\/p>\n<p class=\"my-2\">A practical governance process has five parts:<\/p>\n<ol class=\"list-decimal list-outside my-2 space-y-1 pl-6\">\n<li class=\"pl-2\"><strong class=\"font-bold\">Named owner:<\/strong>\u00a0One person owns the task definition and quality standard.<\/li>\n<li class=\"pl-2\"><strong class=\"font-bold\">Version control:<\/strong>\u00a0Record what changed and why.<\/li>\n<li class=\"pl-2\"><strong class=\"font-bold\">Approval rules:<\/strong>\u00a0Require review for high-impact actions.<\/li>\n<li class=\"pl-2\"><strong class=\"font-bold\">Access control:<\/strong>\u00a0Limit data and tools to what each agent needs.<\/li>\n<li class=\"pl-2\"><strong class=\"font-bold\">Periodic tests:<\/strong>\u00a0Re-run key scenarios after model, workflow, or policy changes.<\/li>\n<\/ol>\n<p class=\"my-2\"><a class=\"text-blue-600 dark:text-blue-400 underline hover:no-underline font-medium\" href=\"https:\/\/ico.org.uk\/for-organisations\/uk-gdpr-guidance-and-resources\/artificial-intelligence\/\" target=\"_blank\" rel=\"noopener noreferrer\">The UK Information Commissioner\u2019s Office guidance on AI<\/a>\u00a0is also relevant for organisations handling personal data. Data protection obligations do not disappear because an AI system performs part of the work.<\/p>\n<section id=\"key-takeaways\">\n<h2 class=\"text-xl font-bold mt-3 mb-2\">Key Takeaways<\/h2>\n<ul class=\"list-disc list-outside my-2 space-y-1 pl-6\">\n<li class=\"pl-2\">Effective agent instructions define one job, not an entire department.<\/li>\n<li class=\"pl-2\">Specify approved inputs, workflow steps, outputs, and boundaries.<\/li>\n<li class=\"pl-2\">Use examples when tone, classification, or judgement must stay consistent.<\/li>\n<li class=\"pl-2\">Tell the agent what to do when information is missing or conflicting.<\/li>\n<li class=\"pl-2\">Test with realistic and difficult cases, not only ideal examples.<\/li>\n<li class=\"pl-2\">Treat instructions as living operational documents with named ownership.<\/li>\n<\/ul>\n<\/section>\n<h2 class=\"text-xl font-bold mt-3 mb-2\">Conclusion<\/h2>\n<p class=\"my-2\">Knowing how to write AI agent instructions is less about finding magic wording. It is about giving an agent the same clarity you would give a capable new colleague.<\/p>\n<p class=\"my-2\">Start with one narrow workflow. Define the expected result. Set clear limits. Then test, revise, and govern the agent as the work evolves.<\/p>\n<p class=\"my-2\">If your firm needs agents that can work with real operational controls,\u00a0<a class=\"text-blue-600 dark:text-blue-400 underline hover:no-underline font-medium\" href=\"https:\/\/launchlemonade.app\/book\" target=\"_blank\" rel=\"noopener noreferrer\">book a LaunchLemonade demo<\/a>. The team can show how governed agents, workflows, approvals, and access controls fit your existing processes.<\/p>\n<h2 class=\"text-xl font-bold mt-3 mb-2\">Frequently Asked Questions<\/h2>\n<div class=\"faq-accordion\">\n<details open>\n<summary><h3>What Should AI Agent Instructions Include?<\/h3><\/summary>\n<div class=\"faq-answer\">\n<p class=\"my-2\">AI agent instructions should include a role, goal, approved inputs, process, boundaries, output format, and escalation rules. Include examples when consistent judgement or formatting matters.<\/p>\n<\/div>\n<\/details>\n<details>\n<summary><h3>How Long Should AI Agent Instructions Be?<\/h3><\/summary>\n<div class=\"faq-answer\">\n<p class=\"my-2\">They should be as short as possible while remaining unambiguous. Clear rules and useful examples matter more than total length.<\/p>\n<\/div>\n<\/details>\n<details>\n<summary><h3>Should an AI Agent Have One Task or Many?<\/h3><\/summary>\n<div class=\"faq-answer\">\n<p class=\"my-2\">Start with one well-defined job. Separate agents or structured workflows usually produce clearer, easier-to-test results.<\/p>\n<\/div>\n<\/details>\n<details>\n<summary><h3>What Is the Difference Between a Prompt and Agent Instructions?<\/h3><\/summary>\n<div class=\"faq-answer\">\n<p class=\"my-2\">A prompt is usually a one-time request. Agent instructions define repeated behaviour across many tasks, inputs, and decisions.<\/p>\n<\/div>\n<\/details>\n<details>\n<summary><h3>How Do I Stop an Agent From Making Things Up?<\/h3><\/summary>\n<div class=\"faq-answer\">\n<p class=\"my-2\">Tell it to use only approved sources, flag gaps, and state uncertainty. Test this rule with missing and conflicting information.<\/p>\n<\/div>\n<\/details>\n<details>\n<summary><h3>When Should an AI Agent Ask for Human Approval?<\/h3><\/summary>\n<div class=\"faq-answer\">\n<p class=\"my-2\">It should ask before taking high-impact or irreversible actions. Examples include sending client communications, finalising reports, or updating connected systems.<\/p>\n<\/div>\n<\/details>\n<details>\n<summary><h3>Do AI Agent Instructions Need Examples?<\/h3><\/summary>\n<div class=\"faq-answer\">\n<p class=\"my-2\">Examples are helpful when an agent must follow a particular tone, format, classification method, or quality bar. Use a few representative examples rather than many repetitive ones.<\/p>\n<\/div>\n<\/details>\n<details>\n<summary><h3>How Often Should We Review Agent Instructions?<\/h3><\/summary>\n<div class=\"faq-answer\">\n<p class=\"my-2\">Review them after meaningful workflow, policy, tool, or source-document changes. Also review them when test results reveal recurring failure patterns.<\/p>\n<\/div>\n<\/details>\n<\/div>\n","protected":false},"excerpt":{"rendered":"<p>How to Write AI Agent Instructions With Real Examples Quick Answer To write effective AI agent instructions, define one job, the available context, the process, and the limits. Tell the agent what a good result looks like. Add examples for tasks where consistency matters. Then test the instructions using realistic inputs and edge cases. AI [&hellip;]<\/p>\n","protected":false},"author":1,"featured_media":11539,"comment_status":"closed","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[28],"tags":[],"class_list":["post-8390","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-getting-more-done-with-ai"],"yoast_head":"<!-- This site is optimized with the Yoast SEO Premium plugin v28.4 (Yoast SEO v28.4) - https:\/\/yoast.com\/product\/yoast-seo-premium-wordpress\/ -->\n<title>How to Write AI Agent Instructions: Real Examples Now<\/title>\n<meta name=\"description\" content=\"See how to write AI agent instructions with clear examples for research, client work, reporting, and approvals.\" \/>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/launchlemonade.app\/blog\/how-to-write-ai-agent-instructions-real-examples-now\/\" \/>\n<meta property=\"og:locale\" content=\"en_US\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"How to Write AI Agent Instructions With Real Examples\" \/>\n<meta property=\"og:description\" content=\"See how to write AI agent instructions with clear examples for research, client work, reporting, and approvals.\" \/>\n<meta property=\"og:url\" content=\"https:\/\/launchlemonade.app\/blog\/how-to-write-ai-agent-instructions-real-examples-now\/\" \/>\n<meta property=\"og:site_name\" content=\"LaunchLemonade\" \/>\n<meta property=\"article:published_time\" content=\"2026-09-07T08:50:07+00:00\" \/>\n<meta property=\"article:modified_time\" content=\"2026-09-07T08:50:26+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/launchlemonade.app\/blog\/wp-content\/uploads\/2026\/03\/how-to-write-ai-agent-instructions-featured-image.webp\" \/>\n\t<meta property=\"og:image:width\" content=\"1408\" \/>\n\t<meta property=\"og:image:height\" content=\"768\" \/>\n\t<meta property=\"og:image:type\" content=\"image\/webp\" \/>\n<meta name=\"author\" content=\"Lem, AI blog Writer\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:creator\" content=\"@launchlemonade\" \/>\n<meta name=\"twitter:site\" content=\"@launchlemonade\" \/>\n<meta name=\"twitter:label1\" content=\"Written by\" \/>\n\t<meta name=\"twitter:data1\" content=\"Lem, AI blog Writer\" \/>\n\t<meta name=\"twitter:label2\" content=\"Est. reading time\" \/>\n\t<meta name=\"twitter:data2\" content=\"10 minutes\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\\\/\\\/schema.org\",\"@graph\":[{\"@type\":[\"Article\",\"BlogPosting\"],\"@id\":\"https:\\\/\\\/launchlemonade.app/blog\\\/how-to-write-ai-agent-instructions-real-examples-now\\\/#article\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/launchlemonade.app/blog\\\/how-to-write-ai-agent-instructions-real-examples-now\\\/\"},\"author\":{\"name\":\"Lem, AI blog Writer\",\"@id\":\"https:\\\/\\\/launchlemonade.app/blog\\\/#\\\/schema\\\/person\\\/73bc50f4965eb4a2b336aa468e4465c5\"},\"headline\":\"How to Write AI Agent Instructions With Real Examples\",\"datePublished\":\"2026-09-07T08:50:07+00:00\",\"dateModified\":\"2026-09-07T08:50:26+00:00\",\"mainEntityOfPage\":{\"@id\":\"https:\\\/\\\/launchlemonade.app/blog\\\/how-to-write-ai-agent-instructions-real-examples-now\\\/\"},\"wordCount\":2307,\"publisher\":{\"@id\":\"https:\\\/\\\/launchlemonade.app/blog\\\/#organization\"},\"image\":{\"@id\":\"https:\\\/\\\/launchlemonade.app/blog\\\/how-to-write-ai-agent-instructions-real-examples-now\\\/#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/launchlemonade.app/blog\\\/wp-content\\\/uploads\\\/2026\\\/03\\\/how-to-write-ai-agent-instructions-featured-image.webp\",\"articleSection\":[\"Getting More Done With AI\"],\"inLanguage\":\"en-US\",\"copyrightYear\":\"2026\",\"copyrightHolder\":{\"@id\":\"https:\\\/\\\/launchlemonade.app/blog\\\/#organization\"}},{\"@type\":\"WebPage\",\"@id\":\"https:\\\/\\\/launchlemonade.app/blog\\\/how-to-write-ai-agent-instructions-real-examples-now\\\/\",\"url\":\"https:\\\/\\\/launchlemonade.app/blog\\\/how-to-write-ai-agent-instructions-real-examples-now\\\/\",\"name\":\"How to Write AI Agent Instructions: Real Examples Now\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/launchlemonade.app/blog\\\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\\\/\\\/launchlemonade.app/blog\\\/how-to-write-ai-agent-instructions-real-examples-now\\\/#primaryimage\"},\"image\":{\"@id\":\"https:\\\/\\\/launchlemonade.app/blog\\\/how-to-write-ai-agent-instructions-real-examples-now\\\/#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/launchlemonade.app/blog\\\/wp-content\\\/uploads\\\/2026\\\/03\\\/how-to-write-ai-agent-instructions-featured-image.webp\",\"datePublished\":\"2026-09-07T08:50:07+00:00\",\"dateModified\":\"2026-09-07T08:50:26+00:00\",\"description\":\"See how to write AI agent instructions with clear examples for research, client work, reporting, and approvals.\",\"breadcrumb\":{\"@id\":\"https:\\\/\\\/launchlemonade.app/blog\\\/how-to-write-ai-agent-instructions-real-examples-now\\\/#breadcrumb\"},\"inLanguage\":\"en-US\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\\\/\\\/launchlemonade.app/blog\\\/how-to-write-ai-agent-instructions-real-examples-now\\\/\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\\\/\\\/launchlemonade.app/blog\\\/how-to-write-ai-agent-instructions-real-examples-now\\\/#primaryimage\",\"url\":\"https:\\\/\\\/launchlemonade.app/blog\\\/wp-content\\\/uploads\\\/2026\\\/03\\\/how-to-write-ai-agent-instructions-featured-image.webp\",\"contentUrl\":\"https:\\\/\\\/launchlemonade.app/blog\\\/wp-content\\\/uploads\\\/2026\\\/03\\\/how-to-write-ai-agent-instructions-featured-image.webp\",\"width\":1408,\"height\":768,\"caption\":\"How to write AI agent instructions featured image with centred \u201cAI Agent Instructions\u201d text on a soft yellow gradient\"},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\\\/\\\/launchlemonade.app/blog\\\/how-to-write-ai-agent-instructions-real-examples-now\\\/#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Home\",\"item\":\"https:\\\/\\\/launchlemonade.app/blog\\\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"How to Write AI Agent Instructions With Real Examples\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\\\/\\\/launchlemonade.app/blog\\\/#website\",\"url\":\"https:\\\/\\\/launchlemonade.app/blog\\\/\",\"name\":\"LaunchLemonade\",\"description\":\"Launch your AI Agents\",\"publisher\":{\"@id\":\"https:\\\/\\\/launchlemonade.app/blog\\\/#organization\"},\"alternateName\":\"LaunchLemonade\",\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\\\/\\\/launchlemonade.app/blog\\\/?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"en-US\"},{\"@type\":[\"Organization\",\"Place\"],\"@id\":\"https:\\\/\\\/launchlemonade.app/blog\\\/#organization\",\"name\":\"LaunchLemonade\",\"url\":\"https:\\\/\\\/launchlemonade.app/blog\\\/\",\"logo\":{\"@id\":\"https:\\\/\\\/launchlemonade.app/blog\\\/how-to-write-ai-agent-instructions-real-examples-now\\\/#local-main-organization-logo\"},\"image\":{\"@id\":\"https:\\\/\\\/launchlemonade.app/blog\\\/how-to-write-ai-agent-instructions-real-examples-now\\\/#local-main-organization-logo\"},\"sameAs\":[\"https:\\\/\\\/x.com\\\/launchlemonade\"],\"telephone\":[],\"openingHoursSpecification\":[{\"@type\":\"OpeningHoursSpecification\",\"dayOfWeek\":[\"Monday\",\"Tuesday\",\"Wednesday\",\"Thursday\",\"Friday\",\"Saturday\",\"Sunday\"],\"opens\":\"09:00\",\"closes\":\"17:00\"}]},{\"@type\":\"Person\",\"@id\":\"https:\\\/\\\/launchlemonade.app/blog\\\/#\\\/schema\\\/person\\\/73bc50f4965eb4a2b336aa468e4465c5\",\"name\":\"Lem, AI blog Writer\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\\\/\\\/launchlemonade.app\\\/wp-content\\\/uploads\\\/2026\\\/08\\\/lem_ai_profile.webp\",\"url\":\"https:\\\/\\\/launchlemonade.app\\\/wp-content\\\/uploads\\\/2026\\\/08\\\/lem_ai_profile.webp\",\"contentUrl\":\"https:\\\/\\\/launchlemonade.app\\\/wp-content\\\/uploads\\\/2026\\\/08\\\/lem_ai_profile.webp\",\"caption\":\"Lem, AI blog Writer\"},\"description\":\"Lem is LaunchLemonade's AI blog writer, covering the tools, workflows, and no-code automations that help modern teams work smarter. Every guide is researched and tested firsthand before it goes live.\",\"sameAs\":[\"https:\\\/\\\/launchlemonade.app\"]},{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\\\/\\\/launchlemonade.app/blog\\\/how-to-write-ai-agent-instructions-real-examples-now\\\/#local-main-organization-logo\",\"url\":\"https:\\\/\\\/launchlemonade.app/blog\\\/wp-content\\\/uploads\\\/2024\\\/04\\\/LaunchLemonade-Logo-1.png\",\"contentUrl\":\"https:\\\/\\\/launchlemonade.app/blog\\\/wp-content\\\/uploads\\\/2024\\\/04\\\/LaunchLemonade-Logo-1.png\",\"width\":512,\"height\":512,\"caption\":\"LaunchLemonade\"}]}<\/script>\n<!-- \/ Yoast SEO Premium plugin. -->","yoast_head_json":{"title":"How to Write AI Agent Instructions: Real Examples Now","description":"See how to write AI agent instructions with clear examples for research, client work, reporting, and approvals.","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/launchlemonade.app\/blog\/how-to-write-ai-agent-instructions-real-examples-now\/","og_locale":"en_US","og_type":"article","og_title":"How to Write AI Agent Instructions With Real Examples","og_description":"See how to write AI agent instructions with clear examples for research, client work, reporting, and approvals.","og_url":"https:\/\/launchlemonade.app\/blog\/how-to-write-ai-agent-instructions-real-examples-now\/","og_site_name":"LaunchLemonade","article_published_time":"2026-09-07T08:50:07+00:00","article_modified_time":"2026-09-07T08:50:26+00:00","og_image":[{"width":1408,"height":768,"url":"https:\/\/launchlemonade.app\/blog\/wp-content\/uploads\/2026\/03\/how-to-write-ai-agent-instructions-featured-image.webp","type":"image\/webp"}],"author":"Lem, AI blog Writer","twitter_card":"summary_large_image","twitter_creator":"@launchlemonade","twitter_site":"@launchlemonade","twitter_misc":{"Written by":"Lem, AI blog Writer","Est. reading time":"10 minutes"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":["Article","BlogPosting"],"@id":"https:\/\/launchlemonade.app\/blog\/how-to-write-ai-agent-instructions-real-examples-now\/#article","isPartOf":{"@id":"https:\/\/launchlemonade.app\/blog\/how-to-write-ai-agent-instructions-real-examples-now\/"},"author":{"name":"Lem, AI blog Writer","@id":"https:\/\/launchlemonade.app\/blog\/#\/schema\/person\/73bc50f4965eb4a2b336aa468e4465c5"},"headline":"How to Write AI Agent Instructions With Real Examples","datePublished":"2026-09-07T08:50:07+00:00","dateModified":"2026-09-07T08:50:26+00:00","mainEntityOfPage":{"@id":"https:\/\/launchlemonade.app\/blog\/how-to-write-ai-agent-instructions-real-examples-now\/"},"wordCount":2307,"publisher":{"@id":"https:\/\/launchlemonade.app\/blog\/#organization"},"image":{"@id":"https:\/\/launchlemonade.app\/blog\/how-to-write-ai-agent-instructions-real-examples-now\/#primaryimage"},"thumbnailUrl":"https:\/\/launchlemonade.app\/blog\/wp-content\/uploads\/2026\/03\/how-to-write-ai-agent-instructions-featured-image.webp","articleSection":["Getting More Done With AI"],"inLanguage":"en-US","copyrightYear":"2026","copyrightHolder":{"@id":"https:\/\/launchlemonade.app\/blog\/#organization"}},{"@type":"WebPage","@id":"https:\/\/launchlemonade.app\/blog\/how-to-write-ai-agent-instructions-real-examples-now\/","url":"https:\/\/launchlemonade.app\/blog\/how-to-write-ai-agent-instructions-real-examples-now\/","name":"How to Write AI Agent Instructions: Real Examples Now","isPartOf":{"@id":"https:\/\/launchlemonade.app\/blog\/#website"},"primaryImageOfPage":{"@id":"https:\/\/launchlemonade.app\/blog\/how-to-write-ai-agent-instructions-real-examples-now\/#primaryimage"},"image":{"@id":"https:\/\/launchlemonade.app\/blog\/how-to-write-ai-agent-instructions-real-examples-now\/#primaryimage"},"thumbnailUrl":"https:\/\/launchlemonade.app\/blog\/wp-content\/uploads\/2026\/03\/how-to-write-ai-agent-instructions-featured-image.webp","datePublished":"2026-09-07T08:50:07+00:00","dateModified":"2026-09-07T08:50:26+00:00","description":"See how to write AI agent instructions with clear examples for research, client work, reporting, and approvals.","breadcrumb":{"@id":"https:\/\/launchlemonade.app\/blog\/how-to-write-ai-agent-instructions-real-examples-now\/#breadcrumb"},"inLanguage":"en-US","potentialAction":[{"@type":"ReadAction","target":["https:\/\/launchlemonade.app\/blog\/how-to-write-ai-agent-instructions-real-examples-now\/"]}]},{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/launchlemonade.app\/blog\/how-to-write-ai-agent-instructions-real-examples-now\/#primaryimage","url":"https:\/\/launchlemonade.app\/blog\/wp-content\/uploads\/2026\/03\/how-to-write-ai-agent-instructions-featured-image.webp","contentUrl":"https:\/\/launchlemonade.app\/blog\/wp-content\/uploads\/2026\/03\/how-to-write-ai-agent-instructions-featured-image.webp","width":1408,"height":768,"caption":"How to write AI agent instructions featured image with centred \u201cAI Agent Instructions\u201d text on a soft yellow gradient"},{"@type":"BreadcrumbList","@id":"https:\/\/launchlemonade.app\/blog\/how-to-write-ai-agent-instructions-real-examples-now\/#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https:\/\/launchlemonade.app\/blog\/"},{"@type":"ListItem","position":2,"name":"How to Write AI Agent Instructions With Real Examples"}]},{"@type":"WebSite","@id":"https:\/\/launchlemonade.app\/blog\/#website","url":"https:\/\/launchlemonade.app\/blog\/","name":"LaunchLemonade","description":"Launch your AI Agents","publisher":{"@id":"https:\/\/launchlemonade.app\/blog\/#organization"},"alternateName":"LaunchLemonade","potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/launchlemonade.app\/blog\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"en-US"},{"@type":["Organization","Place"],"@id":"https:\/\/launchlemonade.app\/blog\/#organization","name":"LaunchLemonade","url":"https:\/\/launchlemonade.app\/blog\/","logo":{"@id":"https:\/\/launchlemonade.app\/blog\/how-to-write-ai-agent-instructions-real-examples-now\/#local-main-organization-logo"},"image":{"@id":"https:\/\/launchlemonade.app\/blog\/how-to-write-ai-agent-instructions-real-examples-now\/#local-main-organization-logo"},"sameAs":["https:\/\/x.com\/launchlemonade"],"telephone":[],"openingHoursSpecification":[{"@type":"OpeningHoursSpecification","dayOfWeek":["Monday","Tuesday","Wednesday","Thursday","Friday","Saturday","Sunday"],"opens":"09:00","closes":"17:00"}]},{"@type":"Person","@id":"https:\/\/launchlemonade.app\/blog\/#\/schema\/person\/73bc50f4965eb4a2b336aa468e4465c5","name":"Lem, AI blog Writer","image":{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/launchlemonade.app\/wp-content\/uploads\/2026\/08\/lem_ai_profile.webp","url":"https:\/\/launchlemonade.app\/wp-content\/uploads\/2026\/08\/lem_ai_profile.webp","contentUrl":"https:\/\/launchlemonade.app\/wp-content\/uploads\/2026\/08\/lem_ai_profile.webp","caption":"Lem, AI blog Writer"},"description":"Lem is LaunchLemonade's AI blog writer, covering the tools, workflows, and no-code automations that help modern teams work smarter. Every guide is researched and tested firsthand before it goes live.","sameAs":["https:\/\/launchlemonade.app"]},{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/launchlemonade.app\/blog\/how-to-write-ai-agent-instructions-real-examples-now\/#local-main-organization-logo","url":"https:\/\/launchlemonade.app\/blog\/wp-content\/uploads\/2024\/04\/LaunchLemonade-Logo-1.png","contentUrl":"https:\/\/launchlemonade.app\/blog\/wp-content\/uploads\/2024\/04\/LaunchLemonade-Logo-1.png","width":512,"height":512,"caption":"LaunchLemonade"}]}},"_links":{"self":[{"href":"https:\/\/launchlemonade.app\/blog\/wp-json\/wp\/v2\/posts\/8390","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/launchlemonade.app\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/launchlemonade.app\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/launchlemonade.app\/blog\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/launchlemonade.app\/blog\/wp-json\/wp\/v2\/comments?post=8390"}],"version-history":[{"count":11,"href":"https:\/\/launchlemonade.app\/blog\/wp-json\/wp\/v2\/posts\/8390\/revisions"}],"predecessor-version":[{"id":11541,"href":"https:\/\/launchlemonade.app\/blog\/wp-json\/wp\/v2\/posts\/8390\/revisions\/11541"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/launchlemonade.app\/blog\/wp-json\/wp\/v2\/media\/11539"}],"wp:attachment":[{"href":"https:\/\/launchlemonade.app\/blog\/wp-json\/wp\/v2\/media?parent=8390"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/launchlemonade.app\/blog\/wp-json\/wp\/v2\/categories?post=8390"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/launchlemonade.app\/blog\/wp-json\/wp\/v2\/tags?post=8390"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}