C1
Clear
What exactly do you want, in what shape?
- Output format
- Length and structure
- Deliverable type
The Chugai Marketing AI Foundation. A practical playbook for getting elite-level work out of Claude Enterprise, grounded in the 4Cs framework from PACE training.
Why this exists
Most teams treat AI like a search engine. Elite users treat it like a junior analyst. The output gap between those two approaches is bigger than the gap between Claude Opus and ChatGPT 3.5. The skill that creates that gap is not paying for the better model. It is prompting.
The 4Cs — which you already know from PACE — applied as your prompting spine.
That compound across every chat from today onwards. Save once, use forever.
A starter promptbook calibrated for Chugai Oncology Marketing. Copy, adapt, save.
Your prompting spine
From Chugai PACE training. Every prompt scored against four questions. The weakest C is where the next correction round will come from.
What exactly do you want, in what shape?
What does Claude need to know about the situation?
What must (and must not) be true about the output?
How will you verify the output is right?
The single biggest unlock
Stop typing the same context into every new chat. Save it once inside a Claude Enterprise Project, in the Custom Instructions field. Every chat you start inside that Project inherits it automatically. Six messages of correction become one good first draft.
Paste this into Claude Enterprise → Projects → Custom Instructions. Save once. Apply forever.
Before you hit send
Run this seven-point check before any prompt that matters. If all seven are answered in the prompt, you have already won most of the iteration rounds before they start.
Tap each item as you confirm it. Print the pocket card (below) and pin it next to your monitor for the first two weeks. After that it is reflexive.
Section II
Eight techniques to compound your output quality. Pick two, build the habit, add more.
Critique drafts in a NEW chat, never the same one. Fresh context means no anchoring. The same chat is committed to the framing it just produced. A new chat is brutal.
You are a sceptical ExCo member. You have five minutes before the meeting. Find three weaknesses in this paper, and the one question that would derail the recommendation in the room.
Show the format, do not describe it. Claude follows literal templates almost perfectly. It interprets prose descriptions loosely.
Output exactly as below. No preamble. No closing remarks. TITLE (max 8 words): BLUF (max 30 words): KEY POINTS (3, one sentence each): RISKS (max 3, flagged commercial / ABPI / comp law): DECISION NEEDED:
The hallucination killer. Generic "where did this come from?" produces plausible-sounding citations. Force it to quote, verbatim.
For each claim above, quote the exact sentence from the source document. If you cannot quote it, mark the claim as inferred.
Tell it what NOT to do. When you ask Claude for an ExCo paper, the first draft typically comes back padded with hedging phrases — "it is worth noting that", "while it is important to recognise", "this is not without its challenges". A short negative-prompt block stops Claude producing them in the first place. You save five minutes per paper.
Constraints: - No marketing buzzwords - No "consult your legal team" caveats - No invented statistics - No qualifying phrases like "it is worth noting" - No bullet points longer than one line - No closing sentence that summarises the document
Pharma-specific pre-mortem. Internal documents leak. ExCo decks get forwarded. Run this on more than just press materials.
If this content leaked to PMLive or Pharma Letter tomorrow, what is the worst possible headline a journalist could write? Rewrite to eliminate that risk.
Imagine the decision failed in twelve months. What would the post-mortem say? Run this on the strong cases too, not just the weak ones.
Imagine this recommendation was approved, then failed twelve months later. What is the single most likely reason it failed? What are we missing now that we will regret then?
One image is worth a thousand instructions. For format-sensitive output, paste a screenshot of "here is what good looks like". Eliminates iteration rounds on layout.
Here is what good looks like for this slide format. Match this layout, this density of text, and this style of heading. [paste screenshot]
For complex business cases. Reserve for prompts where assumption errors will be expensive. Catches the silent assumptions before the first draft.
Before you start, ask me up to five questions where my answer would materially change your output. Wait for my answers before drafting.
Section III
The constraints AI doesn't know by default. Non-negotiable for any output that could face external eyes.
Treat the chat window like an email you might forward by mistake. Before pasting, check the source against Chugai's data handling policy.
Do not paste:
When in doubt: redact first, prompt second.
Three passes, three clean contexts. Catches what a single chat misses.
Build these into your context block so they apply automatically:
Section IV
Ten prompts you can adapt this afternoon. Calibrated for Chugai Oncology Marketing. Click to expand, copy to clipboard, save to your Project.
Interactive tool
Paste a prompt below and score it against the 4Cs framework. The scorer is heuristic, not authoritative — useful as a teaching tool and a sanity check. Trust your own judgement on the borderline calls.
Section V
Embedding the practice across the team. This is not training — it is practice. Each week has one focus and one outcome.
Aim: median team member reaches Capable by week 6. Expert by month 6.
Print and pin
Print one per attendee for the kick-off session. Pin one next to your monitor for the first two weeks. After that the habits are reflexive.
Pocket reference card
The three habits that matter most
Take it with you
The full Foundation Kit. All formats, ready to share with your team and stakeholders.
We come back. We compare promptbooks. We see what changed.
