Prompt Engineering for Product Work: A 2026 Guide

Start with what the task needs

A useful prompt states what you want, supplies the relevant context, and explains any limits or required format. Reasoning and tool use can help a model complete the task, but they do not replace that information.

For a simple request, a sentence may be enough. For a task you will repeat, write down the requirements and test the prompt on representative examples.

The CRISP Framework

Use CRISP when it helps you specify a task clearly.

Context: who you are, what the product is. Role: who the model should play. Instructions: what you want done. Specifics: format, length, constraints. Path: how to handle edge cases.

A bad prompt:

Help me with this PRD

A CRISP prompt:

Context: PM at a B2B SaaS in early-stage growth.
Role: Senior PM coach with 10 years of experience.
Instructions: Review the PRD below. Find three weaknesses.
Specifics: One sentence per weakness. Cite the section.
Path: If a section is missing entirely, say so.

[PASTE PRD]

The CRISP prompt specifies what the critique should cover. Check whether the response addresses those requirements; a more detailed prompt does not guarantee a useful answer.

When to Skip CRISP

CRISP is often unnecessary in these three cases:

1. A clear question with enough context. You do not need every CRISP label if the model already has the information and constraints needed to answer.

2. Quick lookups. "Summarize this email" does not need a role and a path. Just ask.

3. Tool-heavy workflows. Tool definitions explain how to call tools. Your prompt still needs to state the goal, relevant limits, and what a successful result looks like.

A useful heuristic: if the prompt is for a one-shot task and you'll scan the answer in seconds, skip CRISP. If it's a high-stakes prompt you'll save and reuse, use CRISP.

The Hallucination-Aware Pattern

AI can produce convincing claims that are wrong. In OpenAI’s 2025 PersonQA evaluation, hallucination rates were 33% for o3 and 48% for o4-mini¹. Those are results from a specific factual test, not a general rate for reasoning models or a prediction for your task.

For work based on supplied material, add instructions such as:

- Support factual claims with quotes or specific references to the input.
- If you cannot find evidence, say "no clear evidence."
- Keep exact quotes separate from your interpretation. Do not invent details.

These instructions make the response easier to review. Check that the cited material exists and supports the claim; asking for citations does not guarantee accuracy.

Five Templates That Work

Adapt these templates to your product and the evidence you have.

1. PRD Review

Context: PM at [STAGE] [PRODUCT TYPE].
Role: Senior PM coach.
Instructions: Review the PRD below. Find:
- Three weaknesses with section reference
- Three missing pieces
- One question I should ask before shipping

[PASTE PRD]

2. Feedback Synthesis

Context: Analyzing customer feedback for [PRODUCT].
Instructions: Find the top 3 complaints, top 3 requests, and one non-obvious pattern.
Format: Bullet list with verbatim quotes and counts.
Report fewer items if the input does not support the requested count.
Grounding: Every claim must cite a verbatim quote from the input.
If you cannot find evidence, say so.

[PASTE FEEDBACK]

3. Competitive Teardown

Context: PM at [PRODUCT]. Competitor is [COMPETITOR URL].
Instructions: Read the competitor pages I'll link or paste. Identify:
- Three things they do better than us
- Three things we do better
- One opportunity their positioning leaves open

Cite the page or quote for each point.

[PASTE OR LINK COMPETITOR MATERIAL AND YOUR PRODUCT MATERIAL]
If the material does not support a comparison, say what is missing.

4. Eval Design

Context: Need an eval for [AI FEATURE].
Instructions: Help me design a 50-example golden set.
Specifics:
- What dimensions should the examples cover?
- What edge cases should I include?
- What metric matches this task type?
- Show me 5 example entries in the format I should use.

5. Stakeholder Update

Context: [STAKEHOLDER] asks about [METRIC] in 30 minutes.
Data: [LAST 7-30 DAYS]

Prepare 5 talking points, one sentence each:
- Current value with trend
- Biggest issue
- What we're doing about it
- Expected impact of the fix
- One useful finding supported by the data

Use plain language. Cite the data, preserve uncertainty, and say when the planned action or expected impact is not supplied.

Tool Use: The Real 2026 Skill

Connecting a model to the systems used for a task can extend what it does beyond drafting a response.

Three patterns to know:

Web search. Claude, ChatGPT, and Gemini all ship web search. For any question that depends on current info, enable it. The prompt stays simple.

MCP servers. The Model Context Protocol provides a way for compatible AI applications to connect to tools and data sources². Available actions depend on the server, client, and permissions you configure.

Structured outputs. Both Anthropic and OpenAI SDKs let you specify a JSON schema for the response. The schema constrains the response format, but the application still needs to handle refusals, incomplete responses, and request failures. This is essential for any prompt that feeds a downstream system.

If your prompts are doing string manipulation that a tool could do, switch to a tool. A tool can make the operation repeatable and its result easier to check.

Common Mistakes

1. Prompt theater. Long preambles like "you are a world-class expert who is extremely meticulous" add little about the task. Use that space for relevant context and requirements.

2. Telling reasoning models to think step by step. Redundant. They already do.

3. No evidence to check. Ask for source references on factual tasks, then verify that they support the claims.

4. Conflicting or unnecessary instructions. Keep requirements that affect the result. Remove repetition and resolve contradictions; there is no universal rule count that guarantees compliance.

5. Not iterating on the eval, only the prompt. You change the prompt, the answer changes, you call it "better." Without an eval you do not know.

How To Improve Your Prompts

Test reusable prompts. Start with a set of reviewed examples, such as 20 cases, and expand it as you find gaps. Track accuracy on each prompt revision.

Save what works. Keep a personal prompt library in a Notion doc, an Obsidian vault, or our /prompts page. Tag by use case.

Learn from published examples. Review prompts and guidance vendors make public. Test useful ideas on your task rather than assuming they will transfer unchanged.

Revise the specific problem. Remove confusing instructions or add missing context, then check whether the change improves the result.

Action Plan

This week: pick three prompts you reuse. Add CRISP structure where the stakes are high. Add grounding lines where facts matter.

This month: build a 20-example eval for your most-used prompt. Run it every time you change the prompt.

This quarter: move two of your manual workflows from "long prompt" to "agent with tools." Tool access can support workflows that require more than a text response.

Related

• Prompt Library for the full template catalog
• Building Agentic Products for tool use and orchestration
• AI for User Research Synthesis for grounded prompting

Sources