BROKE Framework
Five components that bring OKR discipline to prompt engineering. BROKE ensures every AI prompt has measurable goals — Background, Role, Objective, Key Result, and Expectation — so outputs are judged against clear success criteria, not gut feeling.
Origin: The BROKE framework emerged in 2024 from the intersection of Objectives and Key Results (OKR) methodology and structured prompt engineering. While frameworks like CO-STAR and CRISP focus on communication dimensions, BROKE introduces the concept of measurable outcomes directly into the prompt. The Key Result component — borrowed from OKR practice — forces the prompt author to define what success looks like in quantifiable or verifiable terms before the AI generates a single word. The Expectation component adds a quality gate, specifying the standards the output must meet.
Modern LLM Status: BROKE is increasingly relevant in enterprise and professional AI workflows where accountability matters. Modern LLMs like Claude, GPT-4, and Gemini produce higher-quality outputs when given explicit success criteria. The Key Result and Expectation components address a common failure mode: prompts that produce plausible-sounding but unmeasurable outputs. BROKE is particularly valuable for teams that need to evaluate AI outputs against defined KPIs, quality standards, or compliance requirements.
Define Success Before You Prompt
Most prompts tell the AI what to do but never define what “done well” looks like. The result is output that seems good on first read but fails to meet actual business requirements — the report is well-written but does not address the right KPIs, the email sounds professional but misses the key message, the analysis is thorough but answers the wrong question.
BROKE solves this by embedding success criteria into the prompt itself. The Objective states the goal. The Key Result defines the measurable indicator of success. The Expectation sets the quality bar. Together, these three components transform the prompt from an open-ended request into a specification with verifiable acceptance criteria — just like a well-written OKR transforms a vague intention into a trackable commitment.
Think of BROKE as writing a user story for AI: “Given this Background and this Role, when I request this Objective, the Key Result should be measurable by X, and the Expectation is that quality meets standard Y.”
A prompt that says “write a marketing report” gives the AI no way to know if it succeeded. A BROKE prompt that specifies “Key Result: the report identifies 3 actionable insights with supporting data, Expectation: each insight includes a specific recommendation with estimated impact” gives both the AI and the reviewer clear criteria for evaluation. The output is no longer subjectively “good” or “bad” — it either meets the defined criteria or it does not.
The BROKE Process
Five components that build a goal-driven prompt specification
Background — Set the Context
Provide the relevant background information, situation, and constraints that frame the task. Background grounds the AI in the specific domain, industry context, and any facts it needs before generating a response. Include what has happened, what is happening now, and any relevant constraints.
“We are a B2B SaaS company with 2,000 enterprise customers. Our annual churn rate increased from 8% to 14% over the past two quarters. Customer success interviews indicate onboarding complexity is the primary driver. We have a 90-day onboarding window.”
Role — Assign a Persona
Define the expert perspective or professional identity the AI should adopt. Role shapes the lens through which the AI approaches the problem — a data analyst will frame recommendations differently than a customer success manager, even given the same background information.
“Act as a VP of Customer Success with 10 years of experience in B2B SaaS. You specialize in reducing churn through onboarding optimization and have successfully implemented onboarding programs at three previous companies.”
Objective — State the Goal
Define the specific outcome you want the AI to produce. The Objective should be clear, focused, and actionable. Unlike a vague request, the Objective states exactly what deliverable is expected — not just the topic, but the specific artifact or analysis needed.
“Create a 90-day onboarding improvement plan that addresses the three most impactful friction points in the current customer journey. The plan should include specific interventions, a timeline, and resource requirements.”
Key Result — Define Measurable Success
Specify the measurable indicators that will determine whether the output is successful. Key Results are the signature component of BROKE — they bring OKR accountability to the prompt. Define what specific, quantifiable outcomes the plan should target and what metrics the recommendations should move.
“The plan should target: (1) reducing 90-day churn from 14% to below 10%, (2) increasing onboarding completion rate from 60% to 85%, and (3) reducing time-to-first-value from 45 days to under 21 days. Each intervention must include its projected impact on these metrics.”
Expectation — Set Quality Standards
Define the quality criteria, constraints, and standards the output must meet. Expectation acts as the quality gate — specifying format requirements, depth of analysis, evidence standards, and any non-negotiable constraints. This prevents outputs that hit the metrics on paper but fail on execution quality.
“Each recommendation must include: a clear owner, estimated implementation effort (S/M/L), a rollout timeline, and a risk assessment. The plan should be presentable to the executive team without modification. Avoid generic advice — every recommendation must be specific to B2B SaaS onboarding. All claims should be verifiable.”
See the Difference
Why measurable success criteria produce more actionable outputs
Open-Ended Prompt
Help me improve our customer onboarding process.
Here are some tips for improving customer onboarding: Create a welcome email sequence, provide video tutorials, assign a dedicated success manager, build an in-app walkthrough, send regular check-in emails, and gather feedback surveys at key milestones.
BROKE Prompt
B: B2B SaaS, 2,000 customers, churn up from 8% to 14%, onboarding is the primary driver.
R: VP of Customer Success, B2B SaaS specialist.
O: 90-day onboarding improvement plan targeting top 3 friction points.
K: Reduce churn to <10%, completion rate to 85%, time-to-value under 21 days.
E: Each recommendation needs an owner, effort estimate, timeline, and risk assessment. Executive-ready format.
Intervention 1: Guided Product Setup Wizard (Days 1–7)
Owner: Product team. Effort: Medium. Projected impact: +15% onboarding completion, –8 days time-to-value. Risk: Requires engineering sprint allocation...
Intervention 2: Milestone-Based Check-In Cadence (Days 7–30)
Owner: CS team. Effort: Small. Projected impact: +12% completion rate, early churn signal detection...
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.
BROKE in Action
See how OKR-driven prompts produce measurable, actionable outputs
Background: We are a fintech startup targeting small business owners. Our blog gets 12,000 monthly visitors but only 2% convert to free trial signups. Our top 3 competitors publish weekly and dominate search rankings for key terms like “small business accounting software” and “invoice automation.”
Role: Act as a Head of Content Marketing with deep expertise in B2B SaaS content strategy and SEO-driven growth.
Objective: Create a Q2 content calendar with 12 blog posts, each mapped to a specific keyword and stage in the buyer’s journey (awareness, consideration, decision).
Key Result: The strategy should target: (1) increasing organic traffic from 12,000 to 20,000 monthly visitors, (2) improving blog-to-trial conversion from 2% to 4%, and (3) ranking on page 1 for at least 4 new target keywords.
Expectation: Each post entry must include: title, target keyword, search intent, buyer journey stage, estimated search volume, CTA type, and a 2-sentence content brief. Format as a table. Verify all keyword assumptions before acting on this strategy.
Q2 Content Calendar — 12 Posts
Week 1: “How to Choose Small Business Accounting Software in 2026”
Keyword: small business accounting software | Intent: Commercial | Stage: Consideration | Est. Volume: 8,100/mo | CTA: Comparison guide download
Brief: Decision-stage guide comparing top 5 platforms on pricing, features, and ease of use. Position our product as the simplest option for non-accountants.
Week 2: “The True Cost of Manual Invoicing for Small Businesses”
Keyword: manual invoicing cost | Intent: Informational | Stage: Awareness | Est. Volume: 1,900/mo | CTA: Free calculator tool
Brief: Data-driven analysis of time and money lost to manual invoicing. Include a cost calculator embed to drive engagement and trial signups...
Background: Our engineering department of 45 people is transitioning from annual reviews to quarterly performance conversations. Managers currently struggle to provide actionable feedback — reviews tend to be vague praise or criticism without specific examples or growth paths. We use an OKR-based goal system.
Role: Act as a People Operations Director who has implemented OKR-based performance systems at technology companies.
Objective: Design a quarterly performance conversation template that managers can complete in under 30 minutes per direct report, producing feedback that is specific, growth-oriented, and tied to OKRs.
Key Result: The template should enable: (1) 90% of feedback statements to include a specific behavioral example, (2) every review to include at least 2 measurable growth goals for the next quarter, and (3) completion time under 30 minutes per report.
Expectation: Include the template structure, example fill-ins for a mid-level engineer, and a brief manager guide (under 200 words) explaining how to use it. Avoid HR jargon — write for engineering managers. All suggested metrics should be verified as relevant before deployment.
Quarterly Performance Conversation Template
Section 1: OKR Progress (5 min)
For each active OKR, rate progress: On Track / At Risk / Behind
Include one specific data point per OKR (e.g., “Shipped 3 of 4 planned API endpoints, fourth delayed due to dependency on auth service refactor”).
Section 2: Strengths — What to Keep Doing (10 min)
Cite 2–3 specific behaviors observed this quarter. Format: [Situation] → [Action] → [Impact]
Example: “During the Q1 incident response (Jan 15), you took ownership of the postmortem documentation within 2 hours. The postmortem was used as a template by two other teams...”
Background: We are preparing to launch a new AI-powered search feature for our e-commerce platform. Beta testing with 500 users showed 78% satisfaction, but page load time increased by 1.2 seconds and the feature occasionally returns irrelevant results for long-tail queries. Our production environment handles 50,000 concurrent users at peak.
Role: Act as a VP of Engineering responsible for production readiness decisions at a high-traffic e-commerce platform.
Objective: Create a go/no-go assessment framework for this feature launch, including a risk matrix and a set of launch criteria that must be met before proceeding.
Key Result: The assessment should: (1) categorize all identified risks by severity and likelihood, (2) define pass/fail thresholds for at least 5 launch criteria, and (3) provide a clear recommendation with conditions.
Expectation: Use a structured format with red/yellow/green indicators. Include rollback criteria. The assessment should be understandable by both engineering and business stakeholders. All performance benchmarks should be validated against your actual production data before making launch decisions.
Launch Readiness Assessment: AI Search Feature
Overall Status: YELLOW — Conditional Go
Launch Criteria Matrix
1. Performance: Page load <500ms added → FAIL (current: +1,200ms). Threshold: <300ms at P95.
2. Relevance: Top-3 result accuracy >90% → PASS for head queries (94%), FAIL for long-tail (71%).
3. Scalability: Handles 50K concurrent users → YELLOW (tested to 30K, projected OK).
4. Satisfaction: Beta NPS >70 → PASS (78%).
5. Rollback: Feature flag with <5min disable time → PASS.
Recommendation: Conditional go with phased rollout. Launch to 10% of traffic with feature flag. Gate full rollout on achieving <400ms load time and >85% long-tail relevance...
When to Use BROKE
Best for goal-driven tasks that need measurable outcomes
Perfect For
Quarterly plans, roadmaps, and initiative proposals where every recommendation must tie back to measurable business outcomes and defined success metrics.
Market analyses, competitive assessments, and operational reports where outputs must meet specific analytical standards and support data-driven decisions.
Any task where the output will be reviewed against formal criteria — compliance documents, audit responses, RFP submissions, or certification materials.
Performance reviews, process redesigns, and organizational development initiatives where outcomes must be tracked against baseline metrics.
Skip It When
Brainstorming, creative writing, or ideation tasks where rigid success criteria would constrain the creative process. Use simpler frameworks or free-form prompts.
Simple questions, lookups, or casual interactions where the overhead of defining Key Results and Expectations exceeds the value of the structured approach.
When you do not yet know what success looks like and need to explore before defining outcomes. Use Chain-of-Thought or Self-Ask to build understanding first.
Use Cases
Where BROKE delivers the most value
OKR Development
Draft quarterly OKRs with specific, measurable Key Results tied to business objectives — ensuring every goal has a clear definition of success.
Project Proposals
Build project proposals where every recommendation includes a measurable outcome, resource estimate, timeline, and quality standard that stakeholders can evaluate.
Process Optimization
Analyze workflows and generate improvement recommendations with baseline metrics, target metrics, and expected impact for each suggested change.
Hiring and Talent Strategy
Define hiring plans with specific headcount targets, timeline expectations, diversity metrics, and quality-of-hire indicators that align with company OKRs.
Compliance and Audit
Generate compliance assessments with specific pass/fail criteria, evidence requirements, and remediation timelines that auditors can verify directly.
Training Program Design
Create learning programs with specific competency targets, assessment criteria, completion rate goals, and knowledge retention benchmarks.
Where BROKE Fits
BROKE bridges structured prompting and measurable accountability
If your organization already uses OKRs, BROKE will feel immediately familiar. The Key Result component maps directly to OKR Key Results — specific, measurable indicators of success. The Expectation component adds quality criteria that OKRs sometimes lack. Teams that are fluent in OKR thinking can adopt BROKE with minimal training, and the prompts they write will naturally align with their existing goal-setting discipline.
Related Techniques & Frameworks
Explore complementary approaches to structured prompting
Build Your BROKE Prompt
Structure your next goal-driven prompt with measurable Key Results or find the right framework for your task.