Turn Claude Into Your Personal AI Assistant (Outcome First Prompt Framework)

If you've ever gotten a mediocre response from Claude, the problem wasn't Claude — it was the prompt. Here's a four-part framework that fixes that, permanently, for every prompt you'll ever write.

Why "Be Specific" Is Useless Advice

Everyone tells you to be specific with your prompts. That's advice. What you actually need is a system — something you can use without thinking too hard about it. That system is what I call the Outcome First Prompt Framework. Four parts. Works for any AI, any task, any mode.

The four parts are: Outcome, Context, Constraints, and Format. Answer these four questions before you write a single word in the prompt box, and you will get a killer result every single time.

Part 1: Outcome — The Deliverable, Not the Request

This is the hardest part to nail, and also the most important. Most people write vague requests. "Help me with a LinkedIn post about AI trends." That's not an outcome. That's a direction. An outcome is a specific deliverable with a job to do.

Compare these two prompts:

• Vague: Help me with a LinkedIn post about AI trends.
• Outcome-driven: Create a LinkedIn post that positions me as an AI expert for my network, emphasizes practical applications over hype, and encourages engagement from other professionals in my industry.

The second one tells Claude exactly what to create, who it's for, what angle to take, and what it should accomplish. Anthropic's own guidance says it directly: be specific about the desired output. Know what you're asking for before you ask for it.

Part 2: Context — The Secret Ingredient

Once Claude knows what you want, it still doesn't know enough to do it well. Context fills that gap. It answers: who is the audience, what's the background, why are you doing this, and what's the bigger picture?

Here's a real example. Say you're writing a cold sales email. Without context, Claude invents a generic audience. With context, it writes to your actual person:

"This email is going to VP-level decision makers at mid-market SaaS companies. They are busy, skeptical of pitches, but interested in solutions that save time or money. The company is an AI automation tool. This is a cold outreach campaign — they have never heard of us."

Now Claude isn't writing to an imaginary person. It's writing to your person. Anthropic's guidance backs this up: adding context or motivation behind your instructions helps Claude better understand your goals. Context is not fluff. It is the secret ingredient that separates a generic response from one that actually works.

Part 3: Constraints — The Rails That Keep It On Track

Without constraints, you get bloated, off-target responses. With them, you get exactly what you need. Constraints are the rules: length, tone, style, what to include, what to exclude.

For that same sales email, the constraints would be:

• Length: 100–150 words
• Tone: professional but conversational
• Must include a specific pain point and one social proof point
• Must avoid jargon and generic phrases

Every constraint you add is a guardrail that keeps Claude from going off the rails. The more precise the constraints, the tighter the output. This is where most people skip a step and then wonder why they're getting three-paragraph responses when they wanted a single punchy email.

Part 4: Format — Tell Claude How to Package It

Format is the final piece, and it's the one that saves you editing time. Do you want a table? Bullet points? A document? Code? A script? Tell Claude exactly how the output should look.

For the sales email example: "Return as a single email with subject line plus body copy, ready to copy-paste into Salesforce. Separate subject from body with a blank line."

That one instruction means Claude delivers something you can use immediately — no extra explanation, no wrong structure, no reformatting required on your end.

Three Real Prompts Built Piece by Piece

Let me show you how this plays out in practice across three different types of work.

Example 1: Data Analysis

• Outcome: Analyze this month's customer churn data and identify the top three reasons people cancelled.
• Context: Our user base is mostly freelancers and small agencies. Churn usually happens in the first 30 days.
• Constraints: Focus on actionable insights, not just raw numbers. Exclude churned users who never got activated.
• Format: Return as a bulleted list with the reason, percentage of churn, and one recommended action per reason.

Every element is doing work. The outcome says what you want. The context tells Claude who your users are. The constraints cut out the noise. The format means the output is ready to act on.

Example 2: Creative Copy

• Outcome: Write five subject lines for a sales email about a new product feature.
• Context: The product is project management software for remote teams. The feature helps teams stay aligned without constant meetings. The audience is overworked project managers.
• Constraints: Curiosity-driven, not benefit-driven. Six to eight words each. No exclamation points.
• Format: Number them 1 to 5 and add a one-sentence note on why each works.

Notice how the constraints here are doing heavy lifting — "curiosity-driven, not benefit-driven" is the difference between a subject line that gets opened and one that gets ignored. Without that constraint, Claude defaults to the obvious angle. With it, you get something that earns the click.

Example 3: Code (The Same Logic Applies)

The framework doesn't change for technical work. Outcome is still the deliverable. Context is the codebase, language, or system. Constraints are the rules (no third-party libraries, handle edge case X, follow this naming convention). Format is the output structure (function only, no explanation, with inline comments). The principle is identical.

What Every Mediocre Response Has in Common

Here's the diagnosis I give every time someone shows me a bad Claude output: one of the four elements was fuzzy. Either the outcome was vague, the context was missing, the constraints were loose, or the format wasn't specified. Fix the fuzzy thing, you fix the response. Every time.

After a few prompts built this way, you stop thinking about the framework consciously. It becomes second nature. You'll find yourself mentally running through outcome-context-constraints-format before you even open a new chat. Quality follows clarity — that's the whole principle.

Your Next Step

Take one prompt you've been using — something you've gotten mediocre results from — and rebuild it using all four parts. Write out the outcome as a specific deliverable. Add the context your real audience needs. Set the constraints that define what good looks like. Specify the format so the output is ready to use. Run it. The difference in output quality will be immediate. Once you've done it once, you'll never go back to vague requests again. Next, we go deeper on the most powerful element — context — and how to give Claude the right files, links, examples, and reference material so it produces accurate, relevant outputs every single time.