Ten minutes of setup so the next ninety are productive.
Which Claude plan you actually need and why, what a Project is and how it differs from an ordinary chat, and the one document you will be filling in for the rest of the course. You finish this lesson with an empty Project named after your business, ready to receive everything you build.
The whole setup done on screen: which Claude plan you actually need and why, what a Project is and how it differs from an ordinary chat, and the one document you will be filling in for the rest of the course. You also run a single prompt before anything is configured and keep the answer — it is the before shot you compare against at the end of Part 3. Follow along or watch first and build after. Either way you finish with an empty Project named after your business, ready to receive everything you come to build.
The problem isn't your prompting. It's that Claude has never met you.
Every chat starts from zero, so Claude fills the gaps with the most average consultant it can imagine. This lesson reframes that as a setup problem rather than a wording problem, introduces the four pillars a knowledge base has to carry, and gets you a seven-section skeleton plus a sharpened consultant profile and a tone-of-voice extract taken from your own writing.
The specific beats the polished. Write what is true.
Work through the five remaining sections: two or three genuinely different ideal clients, the problems people actually pay to remove, your positioning, your offers, and the proof Claude is allowed to point at. You bring the truth about your business; Claude sharpens the wording and interrogates the gaps. A rough version in twenty minutes is enough to carry on with — this document is meant to be revisited, not perfected on day one.
Watch the whole thing built, loaded and deliberately stress-tested.
The full Part 1 prompt pack run end to end on Clarion Consulting, then loaded into a Project and attacked with three sharpening questions. You will learn to read the three failure signatures — generic, contradictory, and confidently wrong — and what each one tells you to go back and fix.
Every prompt from Part 1, run end to end on a real consultancy — including the ones that came back wrong.
Watch the full Part 1 prompt pack executed on Clarion Consulting: the knowledge base skeleton, the consultant profile sharpened after a first draft, the tone of voice pulled from three real writing samples, then the ideal clients, problems, positioning, offers and proof. You see every prompt, every output, and every point where a human judgement call replaced what came back. The failures stay in — a workshop where everything works first time teaches you nothing about what to do when yours does not. Follow along with your own document, or watch it through once and build after.
Two slots in a Project. Two completely different jobs.
A Claude that knows your business still sounds wrong, pads answers with caveats, and invents case studies when pushed. That is a behaviour problem, not a knowledge problem, and loading more into the knowledge base never fixes it. Knowledge is the brief; Instructions are the manager.
Tone, format, no-go, push-back. Short list, big difference.
Four short rule sets: how Claude should sound, how it structures and delivers answers, what it must never do regardless of how you ask, and when it should disagree with you. The push-back rules are the ones most consultants skip and the ones that turn Claude from a content machine into something worth arguing with.
The rules block written live, installed, then deliberately broken.
Clarion's four rule sets assembled into a 250–400 word Project Instructions block, pasted in, and then attacked with three awkward tests — one on tone, one too vague to answer, and one inviting invented client results. Including the first version that failed and exactly what changed.
The four rule sets assembled into a single Instructions block, installed, and then deliberately attacked. You will see the word count checked against the 400 ceiling, the placeholder brief from Part 1 replaced, and three awkward tests run in a fresh chat — one on tone, one too vague to answer, and one that openly invites Claude to invent a case study. The first version failed two of the three, and you see exactly which line fixed each one. You finish with a rules block that has been tested rather than assumed.
Where the setup pays for itself in week one.
A pre-call brief that maps a prospect to your core problems and predicts their objections; a first-pass proposal drawn from your own offers, pricing and proof; one long-form piece turned into a week of on-brand content; and Claude used as a critic on the decisions you are least sure about. That fourth one is where most consultants realise what they have actually built.
Three small prompts that compound, and four lines you do not cross.
A morning priority brief, a post-call debrief and a Friday pattern review, each under two minutes. Alongside them the guardrails: nothing client-facing goes out unedited, no number goes out without a source, no confidential client data goes in. You stay accountable. Claude is a tool, not a deputy.
Everything you have built, pointed at real work.
Discovery prep from a live prospect profile, a proposal drawn from messy call notes, one talk repurposed into ten assets, and a genuine critique session — plus the three daily-habit prompts. Closes with the 30-day check: time, quality, confidence, and what a disappointing answer is actually telling you.
Everything you have built, pointed at real work. Watch all four use cases run end to end on Clarion Consulting: a pre-call brief from a real Operations Director profile, a Delivery Reset proposal drawn from messy call notes, one conference talk turned into three posts and a newsletter intro, and a genuine critique session on a pricing decision Priya was unsure about. The three daily-habit prompts run back to back at the end so you can see how short they actually are. Watch the follow-up prompts rather than the first drafts — the cut, the sharpen, the pressure-test. That second pass is the step almost everybody skips, and it is the difference between output you edit heavily and output you approve.