Prompts, agents and workflows — tested in the open

You're a Medical Science Liaison classifying field insights for the Medical Affairs team.

I'm going to paste one or more raw insights, exactly as I captured them after an interaction with a healthcare professional. For each one:

1. Summarise the insight in one clear, actionable sentence (no jargon, no filler)
2. Classify it by:
   - Therapeutic relevance: High / Medium / Low, and why
   - Urgency: Immediate (days) / Next cycle (weeks) / To monitor (months)
   - Stakeholder type: KOL / DOL / General HCP / Patient / Payer / Other
3. Flag whether it needs escalation to Medical, Regulatory or Pharmacovigilance
4. Propose the concrete next action (who to inform, what decision it could change)

Don't invent data I haven't given you. If there's not enough context to classify confidently, say so explicitly instead of assuming.

Here are the insights:
[paste your field insights here, one per line]

Act as an assistant that structures my congress notes into organised takeaways and next steps.

I'm going to paste my raw notes from the sessions I attended at a congress (abstracts, talks, posters, informal conversations). From them, generate:

1. Executive summary (max 5 lines): what actually changes for our therapeutic area
2. Findings by session, as a table: Session | Key finding | Relevance to our product/competition | Evidence level
3. Insights to classify: the 3-5 most actionable points, ready to go through an insight-classification process
4. Pending follow-ups: contacts I should write to, open questions, literature to review
5. Alerts: any safety data, competitive signal or guideline change that needs escalating now

Be concise and concrete — no generic summaries. If my notes are ambiguous or incomplete anywhere, flag it instead of filling the gap yourself.

Here are my congress notes:
[paste your notes here]

Help me prepare a repeatable brief before a scientific call with a healthcare professional.

I'll give you the stakeholder's name/profile and the context of the visit. Generate a brief with this structure:

1. Stakeholder profile: specialty, role (KOL/DOL/HCP), affiliation, known scientific interests
2. Interaction history: what we know from previous contact (summarise what I give you, don't invent)
3. Visit objective: in one sentence — what I'd want to walk away with from this conversation
4. Likely questions they might raise, with key response points based on the evidence I give you
5. Relevant evidence to have on hand: which studies, guidelines or data I should review beforehand
6. What I should NOT do: a reminder of the compliance boundaries relevant to this type of conversation (no promotion, no unsolicited off-label)

Here's the context:
- Stakeholder: [name/specialty/affiliation]
- Visit context: [reason, what's been discussed before, what new evidence exists]

I want to turn a new scientific publication into a structured medical insight, ready to share with my team.

I'll paste the abstract (or full text) of an article. Generate:

1. Plain-language summary (3-4 lines): what was studied, in whom, and what was found
2. Why it matters for our therapeutic area or product — be specific, not generic
3. Evidence level: study type (RCT, observational, real-world, meta-analysis...), sample size, key limitations
4. Contrast with existing evidence: does it confirm, nuance or contradict what we already knew?
5. Practical implications: what could change in conversations with HCPs, or in the Medical Plan
6. Open questions this study doesn't resolve

Base everything strictly on the text I give you — if you need additional context (comparators, prior guidelines) you don't have, say so explicitly instead of assuming.

Here's the article:
[paste the abstract or text here]

Help me put together structured research on a KOL or DOL before my first scientific engagement with them.

From the public information I give you (publications, professional profile, congress/social activity), organise:

1. Scientific profile: area of specialisation, active research lines, affiliations
2. Publication footprint: recurring topics, usual co-authors, journals they publish in
3. Likely positioning: which debates or controversies in the specialty they might have a formed opinion on
4. Points of connection with our therapeutic area — where their interests overlap with what I can offer
5. Opening questions that show I've done my homework, without sounding scripted

Don't invent biographical or publication data I haven't given you — if the public information you have is limited, say so and suggest what I should look up myself.

Here's what I know about this stakeholder:
[paste what you know: name, specialty, publications, profile]

Help me organise my Medical Affairs knowledge into a coherent, reusable structure.

I'm going to give you a messy list of topics, notes, documents or insights I have scattered around. From that:

1. Propose a folder/category structure that makes sense for my territory (e.g. by therapeutic area, by content type, by product lifecycle stage)
2. For each item I give you, indicate where it would fit in that structure and why
3. Flag gaps: what type of content seems to be missing for a complete knowledge base
4. Suggest a simple naming convention so future documents are easy to find

The goal is a system I can maintain on my own, without complex tools — think of something that works with normal folders or a structured note.

Here's my list of scattered content:
[paste your topics/notes/documents here, one per line]

Help me do a structured debrief after a congress, faster and more consistent than a free-form summary.

I'll give you my raw notes (or a transcribed recording) from the congress. Generate a debrief in this exact format, so it's comparable across congresses:

Congress: [name, date]
Objectives met: (of the ones I set before going, which I achieved)
3 key scientific findings: with relevance and implication for each
Stakeholder interactions: 1-line summary per relevant contact, with proposed follow-up
Comparison with previous congresses (if I give you prior debriefs): what's changed

Be honest about limitations: if my notes don't support completing a section, say so instead of inventing content.

Here are my congress notes:
[paste your notes or transcript here]

Help me build a safe, effective prompt for a specific Medical Affairs task.

I'll describe a task I do regularly (or want to partly automate). From that description:

1. Identify the role the assistant should take on for this task
2. Propose a clear output structure (sections, table, list) suited to how I'll use the result
3. Add explicit guardrails: what the assistant must NOT do (invent data, give clinical recommendations, bypass compliance)
4. Leave a clear placeholder where I'll paste my real information each time I use the prompt
5. Give me the final prompt, ready to copy and reuse

Here's the task I want to turn into a prompt:
[describe your task: what information you receive, what you need to get out of it, who the result is for]

Help me design what my week as an MSL could look like if I leveraged AI at every stage — without it replacing my scientific judgement.

I'll describe my typical activities for a normal week. For each one, suggest:

1. Whether AI could help with that activity, and exactly how (what specific task I'd delegate to it)
2. What part of the activity must stay 100% mine, no exceptions (clinical judgement, human relationship, final decision)
3. An example prompt or process I could try this week for that specific activity
4. The main risk of using AI there, and how to mitigate it

At the end, give me a summary of my "augmented" week as a table: Activity | AI-assisted | Stays 100% mine.

Here's my typical week:
[describe your usual activities for a week, day by day or by task type]