Free Guide

The practical playbook I teach my clients on day one. Working modes, plans, platforms, the PACT prompting framework, twelve field-tested tips, and an honest list of what Claude still cannot do.

What Claude Offers

Three Modes, Three Jobs

Most teams plateau on Claude because they use one mode for every task. The first move is matching the surface to the job.

Chat

Fast back-and-forth in the browser or app. The default surface for quick answers, drafts, and brainstorms.

Use it when you need a one-shot answer or a first draft.
Skip it when the task spans multiple files or sessions.

Cowork

Claude takes the work off your plate. Reads files, runs tools, takes multi-step actions, hands you back the result. Less "ask and answer," more "delegate and review."

Use it when you want a task executed, not discussed.
Skip it when you just need a quick answer, or you need to be in the loop on every step.

Code

Claude Code in the terminal - agentic, file-aware, runs commands, edits real repositories. The serious development surface.

Use it when you are shipping software.
Skip it when you only need an explanation.

Plans & Platforms

Pick the Right Tier and the Right Surface

Plans

Five Tiers, One Decision

Free

Try it. Good enough to feel the model, light enough to outgrow.

Pro

Daily power use for one person. The first real tier.

Max

For heavy individual users - long sessions, big files, frequent extended thinking.

Team

Multi-seat with admin tooling. The right tier when more than two people are using it daily.

Enterprise

Org-wide deployment with SSO, advanced controls, and procurement-friendly contracts.

Platforms

Four Surfaces, Different Strengths

Web

The default. Fastest to share a link, works anywhere, no install.

Desktop

Better for long sessions, quick keyboard launch, and screen-aware work.

Mobile

For voice notes, photo questions, and offloading thoughts while you walk.

API

When Claude becomes part of your product - automation, agents, custom apps.

How Claude Thinks

Instant vs Extended Thinking

Claude can answer fast or reason slowly on hard problems. Knowing which mode to reach for is the second-biggest skill, right after writing the prompt itself.

Default

Instant

Quick, conversational, no visible reasoning step. Claude responds immediately and you iterate by chatting back.

Reach for it when:

  • Rewriting an email
  • Summarising a meeting transcript
  • Drafting a tagline or first paragraph
  • Quick lookups and definitions
Deliberate

Extended Thinking

Slower and more methodical. Claude reasons step-by-step before answering, and the quality jump on hard problems is real.

Reach for it when:

  • "Should we restructure this pricing page?"
  • "What is wrong with this database schema?"
  • Comparing three vendor proposals against your real constraints
  • Anything where being wrong costs more than waiting thirty seconds

Default to Instant. Reach for Extended when the cost of being wrong is higher than the cost of waiting.

The Bentata Method

The PACT Framework

Most prompts fail for three reasons: vague intent, missing context, and no defined output shape. PACT fixes all three in about sixty seconds of writing time, and saves twenty minutes of back-and-forth.

P

Purpose

The business or creative goal. Why does this output exist, and what changes after Claude produces it?

"Convince a busy founder to book a discovery call by the end of this week."

A

Assets

Files, links, examples, prior work, real numbers - anything Claude needs to read instead of guess.

"Use the attached pitch deck and the three case studies in /case-studies/."

C

Constraints

Tone, length, audience, deadlines, regulatory limits, what to avoid. Everything that scopes the answer.

"Under 150 words. Friendly, not salesy. No emojis. Don't promise specific outcomes."

T

Template

The exact shape of the output: headings, table, JSON, document structure, word count.

"Subject line, then three short paragraphs, then a single CTA link on its own line."

× Weak Prompt

Write a launch email for our new feature.

✓ PACT Prompt

Purpose: get existing Pro customers to try our new bulk-export feature this week. Assets: the feature spec at /docs/bulk-export.md and the last launch email I sent (attached). Constraints: under 180 words, our usual warm-and-direct voice, no superlatives, single CTA. Template: subject line, three short paragraphs, one CTA button text on its own line.

PACT takes sixty extra seconds to write. It saves twenty minutes of back-and-forth.

From the Field

Twelve Tips I Give Every Client

Not philosophy. Habits and moves that change Monday-morning output, gathered from real consulting work.

Setup & Habits

1. Start a fresh chat for new work

Long chats accumulate noise, stale assumptions, and old constraints. New task means new chat.

2. Set Custom Instructions once

Who you are, what you do, how you like answers. Five minutes saved on every session, forever.

3. Save your best prompts

Skills, text expanders, a notes file - anywhere. The good ones get reused dozens of times.

4. Treat Claude like a sharp junior

Brief it like you would brief a real person. Real context in, real work out.

Quality & Verification

5. Use PACT for anything non-trivial

The sixty extra seconds always pays back. Vague prompts produce vague output.

6. Audit every output before using it

Read it the way you'd read a junior's draft. Confidence is not the same as correctness.

7. Run a second-opinion pass

Paste the output back into a fresh chat and ask "what is wrong with this?" The list is usually real.

8. Verify dates, prices, and numbers

Claude does not know what week it is and does not always remember the last decimal. Check the facts that matter.

Power Moves

9. Reach for Extended Thinking on big calls

Pricing, architecture, hiring, strategy. Anywhere wrong is expensive, slow is worth it.

10. Use Projects to keep context warm

A persistent workspace for files, instructions, and history - works inside both Chat and Cowork. Context arrives for free.

11. Connect only what is needed

Drive, Slack, GitHub - useful, but a smaller surface produces sharper answers. Add connectors deliberately.

12. When stuck, ask Claude what it needs

"What additional information would let you do this better?" The answer reveals what you forgot to give it.

Honest Limits

What Claude Cannot Do (Yet)

The kind of person who reads honest limitations is the kind of person who hires good consultants. Here is the short list.

Real-time data

Live prices, current news, today's weather, this week's launches. Claude has a knowledge cutoff. For anything time-sensitive, give it the data or use a tool that fetches it.

Reliably perfect math

Long calculations, tax-grade arithmetic, anything where the eighth decimal matters. Use a calculator or hand it a tool. Trust language, not silent maths.

Access your private systems

Your CRM, your database, your file server - Claude sees none of it unless you explicitly connect it through Projects, the API, or an integration.

Remember across sessions

By default, every fresh chat starts from zero. Use Projects, Custom Instructions, or paste-back to carry context forward intentionally.

Replace high-stakes judgment

Hiring, firing, legal calls, medical calls, anything regulated. Claude is a thinking partner that sharpens the decision - it is not the decision-maker.

Want this as a workshop for your team?

I deliver the full playbook as a 90-minute hands-on session. Your team leaves with PACT prompts and habits they actually use Monday morning - not slides they forget by lunch.

Book a Call