AI Tools

80% to Claude: How I Hand Off Most of My Business Work to AI (And Actually Send It)

The difference between people getting real leverage from Claude and people who aren't isn't intelligence — it's one system decision they made once. Here's what that looks like.

MurphMarch 23, 20268 min read

Here's a thing I've noticed talking to freelancers and consultants who use Claude.

The ones getting real leverage aren't running more prompts. They're not spending more time on AI. They made one structural decision — usually about two hours of setup work — and now Claude handles most of the execution.

The ones stuck are still writing prompts from scratch every time, rewriting half the output, and quietly concluding that "AI is useful but not that useful."

This post is about the structural decision. And what changes once you make it.

The Problem With How Most People Use Claude

You open a new chat. You write a sentence or two describing what you need. You get something that's 60% there. You rewrite the rest yourself.

That's not bad. But it's not leverage.

The issue isn't Claude. It's context. Every time you start a new chat, Claude knows nothing about you — your business, your clients, your voice, your standards, your specific terminology. You're starting from zero and hoping a short prompt bridges the gap.

It can't. Not reliably.

The people getting dramatic results from AI have solved the context problem. They've given Claude a stable, rich picture of who they are and what they do — and now every prompt starts from that baseline instead of a blank slate.

The 80/20 of What Claude Can Own

Before building a system, you have to audit your work.

Most people's instinct is to use AI for "the creative stuff" — copy, ideation, first drafts. That's fine. But that's not where the leverage is.

The real leverage is in the high-frequency, medium-difficulty tasks. Not the simple stuff you could dash off in two minutes anyway. And not the deeply strategic work that requires your specific judgment. The stuff in between:

  • Client proposals and project scopes
  • Follow-up sequences
  • Research summaries and competitive briefs
  • Meeting prep notes
  • Service page copy for new offerings
  • FAQ and support documentation
  • Social posts and newsletter sections
  • SOPs for recurring processes

These tasks have a few things in common: they take 30–90 minutes each, they follow patterns (even if the content varies), and they require you to sound like yourself — not a generic AI.

That last part is where most people get stuck. And it's exactly what the context file system solves.

Context Files: The One-Time Setup That Changes Everything

A context file is a document you give Claude at the start of a session (or store in a Project) that contains everything it needs to work like you.

Here's what mine includes:

Business context. What I do, who I serve, what problems I solve, what I don't do. Written in plain English like I'm explaining to a smart friend.

Voice and tone. Examples of my actual writing — emails I've sent, proposals I've written, posts that landed well. Claude doesn't analyze these. It absorbs them.

Client archetypes. The three or four types of clients I work with most. Their language, their concerns, their decision-making patterns.

Standards. Things I never do. Things I always do. Language I avoid. How I handle pricing conversations, scope changes, hard questions.

Setup time: about 90 minutes the first time. After that, you paste the file into every new session and Claude is immediately operating at full context.

The results are immediate and obvious. Output stops sounding like "an AI trying to write like you" and starts sounding like you.

The 7 Core Workflows

Once you have context, the next step is building workflow templates — structured prompts for the tasks you do most often.

The ones I use constantly:

1. Proposal generator. Input: project type, client industry, key requirements, budget range. Output: full proposal draft in my format, with my pricing language, at my level of specificity.

2. Follow-up sequence. Input: what was discussed, where we left things, the next step I'm trying to drive. Output: 2–3 follow-up messages, spaced and calibrated to the situation.

3. Content brief. Input: topic, target audience, key points to hit. Output: research summary + outline + SEO notes, ready for me to write from or hand off.

4. SOP draft. Input: a process I do regularly, described in rough terms. Output: a clean step-by-step SOP I can share with a contractor or VA.

5. Service page copy. Input: service name, who it's for, what they get, price, key objections. Output: a full page draft in my voice.

6. Social post batch. Input: a topic, angle, or piece of existing content. Output: 5–7 posts adapted for different platforms and formats.

7. Client debrief summary. Input: notes or transcript from a client call. Output: clean summary with key takeaways, next steps, and open questions.

Each template is a 150–300 word structured prompt. Each one took me 20–40 minutes to build and test. Together they cover about 80% of what I produce in a given week.

The Trust Ladder

Here's the mental model most people are missing.

Most people operate at trust level 1: Claude writes something, I rewrite most of it, I send what I wrote. AI saves maybe 20% of the time.

Trust level 2 is: Claude writes something, I review and edit lightly, I send. AI saves 50–60% of the time.

Trust level 3 is: Claude writes something, I read it once, I send. AI saves 80–90% of the time.

Most people are stuck at level 1 or trying to jump to level 3 too fast. The move is deliberate iteration at level 2.

For each workflow, I ran it 10–15 times before I trusted it. I edited less each time. I tracked what kept needing to be fixed and updated the template. Eventually the template was good enough that I barely touched the output.

That's when leverage kicks in. Not when you first try Claude. When you've iterated a specific workflow to the point where the output is reliably good enough to send.

What This Actually Looks Like in Practice

On a typical Monday, here's what Claude handles for me before I've had lunch:

  • Three client follow-up emails (two minutes of input each, zero editing)
  • A full project proposal for a new inquiry (15 minutes of input, light edits)
  • A week's worth of social posts (30 minutes of input, done)
  • Research summary for a new client's industry (one prompt, 8 minutes)

That's work that used to take most of my day. Now it's done by 11 AM.

The remaining 20% — the strategic thinking, the relationship work, the stuff that actually requires my judgment — gets my full attention instead of my last-hour leftovers.

That's the shift. Not "AI does my work." AI does the execution so I can focus on the parts only I can do.

The Guide

I packaged all of this into a $79 guide: the audit framework, the context file template, all seven workflow prompts, and the trust ladder methodology.

It's not a course. It's a system — the exact documents and prompts I use, adapted for you to use. Setup is a few hours. The leverage is permanent.

Get 80% to Claude →

If you're already using Claude and not getting dramatic results, the system is the thing you're missing — not a better prompt, not a different AI. The system.

Want to see how your business stacks up?

Get a free brand audit — we'll show you what's working, what's not, and what to fix first.

Free Brand Audit →

Frequently Asked

What is the 'context problem' that limits most people's results with Claude?

Every new Claude chat starts blank — no knowledge of your business, voice, clients, or standards. Without a system that gives Claude rich context upfront, every output requires heavy rewriting. Solving this once, by building a persistent context document, is what separates people getting real leverage from people who think AI is only marginally useful.

How long does it take to set up a Claude workflow that runs on its own?

Most people are up and running in about two hours. The setup involves writing a master context document about your business, building a library of prompt templates for your most common tasks, and establishing a simple workflow for how Claude plugs into your work. The payback starts immediately.

What does '80% to Claude' actually mean in practice?

It means handing off the first 80% of any task — first draft, research, structure, formatting — to Claude, then spending your time on the 20% that requires your judgment: positioning, relationship nuance, final editing. The goal isn't to eliminate your input; it's to make your input the high-value kind.

Can freelancers and solo consultants actually use Claude the way agencies do?

Yes, and solo operators often have the advantage — they can build a single, deeply personalized Claude context rather than trying to standardize across a team. The workflow scales down to one person just as effectively as it scales up.

Jason Murphy

Written by

Murph

Jason Matthew Murphy. Twenty years building digital systems for businesses. Former CardinalCommerce (acquired by Visa). Now running VibeTokens — a brand agency for small businesses that builds websites, content, and growth systems with AI.

Your brand is your first impression.

Find out if it's costing you customers.

Free brand audit. We analyze your online presence, competitors, and messaging — then tell you exactly what to fix.

Get Your Free Brand Audit →