ERA Framework
Start with where you want to end up. ERA flips the typical prompt structure by leading with your Expectation — the desired outcome — before assigning a Role and defining the Action. When the destination is clear, the AI takes a more direct path to get there.
Introduced: The ERA Framework (Expectation, Role, Action) emerged from the prompt engineering community in 2024 as a deliberate inversion of the role-first approach used by most structured frameworks. While frameworks like CRISP and CO-STAR typically begin by assigning the AI a persona, ERA argues that defining the expected outcome first produces more goal-aligned responses. The reasoning: when you tell an AI what success looks like before telling it who to be, every subsequent decision — including how to inhabit the role — is anchored to the end goal.
Modern LLM Status: ERA remains a practical and effective framework for goal-oriented prompting. Its expectation-first ordering aligns with how modern LLMs process context: early tokens in a prompt have disproportionate influence on output direction. By placing the desired outcome at the beginning, ERA leverages this primacy effect to steer the AI before role or task instructions add complexity. Whether you use Claude, GPT-4, or Gemini, leading with a clear expectation consistently produces outputs that are more tightly aligned with your actual goals.
Define the Destination Before the Journey
Most prompts follow a natural but suboptimal order: “Be a [role] and do [action].” The problem is that without knowing the desired outcome, the AI interprets the role broadly and the action generically. A “marketing expert” could write anything from a billboard tagline to a 50-page strategy document — the role alone does not constrain the output enough.
ERA reverses the sequence. By stating the Expectation first — what the final output should look like, who it serves, and how success is measured — you give the AI a target to aim at before it even knows what role to adopt. The Role then becomes purpose-built for that specific outcome, and the Action becomes the precise steps needed to deliver it. Every component serves the goal.
Think of it like briefing a consultant. You would not start by saying “Be an analyst” — you would say “I need a board-ready slide deck showing our market position by Friday” (expectation), then specify “approach this as a competitive intelligence analyst” (role), then detail the specific research tasks (action). The expectation shapes everything that follows.
In natural language processing, earlier tokens in a prompt carry more weight in shaping the model’s response trajectory. This is known as the primacy effect. By placing your expected outcome at the very beginning of the prompt, ERA takes advantage of this phenomenon — the AI’s entire processing of the subsequent role and action is already conditioned by the goal you stated first. This is not just a theoretical benefit: prompts that lead with outcomes consistently produce more focused, goal-aligned results than those that lead with roles or tasks.
The ERA Process
Three components, outcome-first ordering
Expectation — What Does Success Look Like?
Define the desired outcome before anything else. What should the final output be? Who will use it? What format, quality, and characteristics make it successful? The Expectation is not just a format specification — it is a complete picture of what “done right” looks like. Be specific about measurable criteria: length, structure, audience, tone, and the standard the output must meet.
“I need a 500-word blog post introduction that hooks technical readers in the first sentence, establishes credibility within the first paragraph, and ends with a clear thesis statement about why microservices are not always the right architecture choice.”
Role — Who Should Deliver This?
Now that the outcome is clear, assign the AI a persona that is specifically suited to deliver it. The Role should be chosen because it serves the Expectation — not as a generic identity, but as the exact expert whose voice, perspective, and knowledge would produce the best version of the expected output. A role without a goal is vague; a role calibrated to a specific outcome is powerful.
“Write as a senior software architect with 15 years of experience who has led both monolith-to-microservice migrations and teams that chose to stay monolithic — someone who argues from experience, not ideology.”
Action — What Specifically Should Be Done?
With the outcome defined and the role assigned, specify the concrete steps or behavior the AI should follow. The Action is the execution plan — the specific task, methodology, or approach that connects the role to the expectation. Because both the goal and the persona are already established, the action can be precise and purposeful rather than open-ended.
“Open with a provocative question about a common microservices assumption. Follow with a brief personal anecdote about a migration that taught a hard lesson. Close the intro by framing the rest of the post as a balanced analysis, not an anti-microservices polemic. Verify all technical claims before publishing.”
See the Difference
Why leading with expectations produces more targeted results
Role-First Prompt
You are a financial advisor. Help me plan for retirement.
Planning for retirement involves several key steps. First, determine how much you will need by estimating your annual expenses in retirement. Consider factors like inflation, healthcare costs, and lifestyle goals. Start by maximizing your 401(k) contributions, especially if your employer offers matching. Diversify your investments across stocks, bonds, and other assets based on your risk tolerance and time horizon...
ERA Prompt
Expectation: I need a personalized retirement savings action plan for a 35-year-old earning $85K/year with $40K already saved, targeting retirement at 60. The plan should include specific monthly savings targets, account type recommendations, and three milestone checkpoints at ages 40, 50, and 55.
Role: Approach this as a certified financial planner who specializes in middle-income retirement planning and favors practical, actionable advice over theoretical models.
Action: Calculate the savings gap, recommend a specific monthly contribution amount, suggest an account allocation strategy (401k, Roth IRA, taxable), and create a milestone tracker with specific portfolio targets for each checkpoint age. Note that I should verify all financial calculations and consult a licensed advisor before making investment decisions.
Your Retirement Action Plan: Age 35 to 60
The Gap: To maintain your current lifestyle in retirement, you will need approximately $1.4M by age 60 (assuming 4% withdrawal rate and moderate inflation). With $40K saved, you need roughly $1.36M more over 25 years.
Monthly Target: $1,850/month across all accounts
• 401(k): $1,150/month (max employer match first)
• Roth IRA: $550/month ($6,500 annual limit)
• Taxable brokerage: $150/month (overflow + flexibility)...
Important: These projections use historical averages and simplified assumptions. Consult a licensed financial advisor for personalized investment advice.
Natural Language Works Too
While structured frameworks and contextual labels are powerful tools, LLMs are exceptionally good at understanding natural language. As long as your prompt contains the actual contextual information needed to create, answer, or deliver the response you’re looking for — the who, what, why, and constraints — the AI can produce complete and accurate results whether you use a formal framework or plain conversational language. But even in 2026, with the best prompts, verifying AI output is always a necessary step.
ERA in Action
See how expectation-first ordering shapes better outputs across different scenarios
Expectation: I need a 300-word press release announcing our new AI-powered inventory management system for small retailers. It should be publication-ready for industry trade outlets, include one quote from our CEO, and end with a boilerplate company description. The tone should convey innovation without hype.
Role: Write as a seasoned technology PR specialist who has placed stories in Retail Dive and TechCrunch and understands that journalists skip anything that reads like an advertisement.
Action: Lead with the customer problem (inventory waste costs small retailers $X annually), introduce the solution in the second paragraph, embed the CEO quote in the third paragraph, and close with availability details and the boilerplate. Avoid superlatives like “revolutionary” or “game-changing.” All statistics should be verified against actual data before distribution.
FOR IMMEDIATE RELEASE
[Company Name] Launches AI Inventory System to Cut Small Retailer Waste
Small retailers lose an estimated 8% of revenue annually to overstock, stockouts, and manual counting errors — a problem that enterprise-grade inventory systems have solved for large chains but left unaddressed at the small business level.
[Company Name] today announced [Product Name], an AI-powered inventory management platform built specifically for independent retailers with 1 to 10 locations. The system uses demand forecasting to automate reorder points, reducing excess stock by an average of 34% in pilot testing...
“Independent retailers deserve the same demand intelligence that big-box chains have had for a decade,” said [CEO Name], CEO of [Company Name]. “We built [Product Name] so a shop owner can stop guessing and start knowing.”...
Note: Verify all statistics, pilot results, and quotes with actual company data before distribution.
Expectation: I need a comprehensive 30-60-90 day onboarding checklist for new software engineers joining a mid-size SaaS company. It should be organized by week, include both technical and cultural milestones, and have a “manager checkpoint” at the end of each 30-day period. The checklist should be immediately usable without customization for the technical setup items.
Role: Approach this as a VP of Engineering who has onboarded over 100 engineers and has learned that the first 90 days determine whether a new hire becomes a productive team member or quietly disengages.
Action: Structure each 30-day block with specific, actionable items (not vague goals). Include technical setup, codebase orientation, first contribution targets, relationship-building tasks, and cultural integration milestones. Add a “red flag” section for each checkpoint that helps managers identify when onboarding is going off track. Adapt this to your specific company tools and culture before distributing.
Software Engineer Onboarding: 30-60-90 Day Playbook
Days 1–30: Foundation
Week 1: Environment & Access
• Complete all IT access requests (GitHub, Jira, Slack, AWS console, CI/CD pipeline)
• Clone the main repository and run the full test suite locally
• Pair with assigned buddy for a 90-minute codebase walkthrough
• Attend or watch recordings of the last 3 sprint retrospectives...
Day 30 Manager Checkpoint
• Has the engineer shipped at least one merged PR?
• Can they explain the high-level architecture to a peer?
Red Flags: No PR attempts yet; avoids asking questions; has not met with anyone outside their immediate team...
Note: Customize tool names, team structures, and cultural milestones to match your specific organization before using.
Expectation: I need a customer apology email that acknowledges a 4-hour service outage affecting our project management SaaS platform. The email should restore trust, not just say sorry. It must include what happened (without jargon), what we are doing to prevent recurrence, and a concrete goodwill gesture. Under 200 words. The customer should feel heard, not patronized.
Role: Write as a Head of Customer Success who takes outages personally and believes transparency is the only path to retained trust — someone who would rather over-communicate than hide behind corporate language.
Action: Open with a direct acknowledgment (no “we apologize for any inconvenience” hedging). Explain the root cause in one plain-language sentence. List two specific preventive measures being implemented. Offer a concrete goodwill gesture (credit, extended trial, or direct access to engineering for questions). Close with a personal commitment. Review and customize with actual incident details before sending.
Subject: We Let You Down Yesterday — Here Is What We Are Doing About It
Hi [Name],
Yesterday between 2:00 PM and 6:00 PM EST, [Product] was down. Your team could not access your projects, and that is on us. I am not going to dress this up: it should not have happened.
What happened: A database migration failed during a routine update, and our rollback process did not trigger as designed.
What we are fixing:
• Implementing staged rollouts so updates hit 5% of users before full deployment
• Adding automated rollback triggers that activate within 60 seconds of anomaly detection
For you: We are crediting your account with one full month free. No action needed — it is already applied.
If you have questions or want to speak directly with our engineering team, reply to this email and I will set it up personally.
[Name], Head of Customer Success
Note: Customize with actual incident details, times, and credit amounts before sending.
When to Use ERA
Best when outcomes matter more than process
Perfect For
When you know exactly what the final product should look like — a report, a plan, a specific document format — and need the AI to work backward from that vision.
Emails, reports, and presentations where the recipient’s needs and the expected impact should drive every word — not the author’s process for getting there.
When the output will be reviewed by decision-makers or published externally — ERA’s expectation-first approach ensures the quality bar is set before a single word is generated.
Building reusable documents where the expected structure, completeness criteria, and audience needs must be crystal clear from the start.
Skip It When
When you do not yet know what the outcome should look like — brainstorming, ideation, or creative exploration where rigid expectations would limit discovery.
Logic problems, code debugging, or complex reasoning where the process matters more than a predefined output format. Chain-of-Thought or Self-Ask techniques serve these better.
When audience, tone, style, and format all require independent specification, richer frameworks like CO-STAR or CREATE provide the additional dimensions ERA intentionally omits.
Use Cases
Where ERA delivers the most value
Business Reports
Generate quarterly reports, market analyses, and executive summaries where the deliverable format and audience expectations are clear before a single word is written.
Professional Email Drafting
Craft emails where the desired recipient reaction and outcome drive every word — from partnership proposals to difficult feedback conversations.
Training Materials
Build course outlines, workshop agendas, and learning assessments where the learning outcomes are defined first and the content is engineered to achieve them.
Customer Communication
Draft outage notifications, feature announcements, and support escalation responses where the expected customer response shapes the message strategy.
Policy and Compliance Documents
Create security policies, compliance checklists, and governance frameworks where the regulatory expectation is the non-negotiable starting point for all content.
Project Planning
Generate project plans, sprint breakdowns, and resource allocation proposals where the success criteria and deliverables are defined before the work plan is constructed.
Where ERA Fits
ERA occupies the goal-first position on the structured prompting spectrum
ERA’s key insight is that the sequence of information in a prompt is not arbitrary. Placing expectations first primes the AI’s entire response generation toward that goal. This is analogous to writing a thesis statement before the body of an essay — everything that follows is automatically organized around the central claim. If you find that role-first frameworks produce outputs that are well-written but miss your actual goal, try leading with your expectation instead. The role and action will naturally sharpen to serve that outcome. And as always, verify the final output before relying on it.
Related Techniques & Frameworks
Explore complementary approaches to structured prompting
Try the Expectation-First Approach
Build your next prompt with ERA’s outcome-first structure or explore other frameworks to find the best fit for your workflow.