Module 04A · Lesson 05

Capstone — Your Bench Scientist's AI Playbook

Reading time: 18 minutes (plus exercise time) Track: Role Path — Bench R&D · Capstone Prerequisites: Module 04A · Lessons 01-04

What this lesson does

This is the capstone of Module 04A. It pulls together everything from Lessons 01-04 into a personal operating system you'll actually use.

By the end of this lesson, you'll have produced:

A personal prompt library — 8-12 prompts for the work you do most
A workflow map — your top 3-5 recurring lab science workflows with AI integration
A verification checklist — your habitual quality controls for AI-assisted work
A 30-day plan — what you'll practice over the next month to internalize the capability

This lesson is more workshop than reading. Set aside 90 minutes when you can do it without interruption.

01 · Why a personal playbook matters

You've now finished 13 lessons of curriculum. You know the frameworks, the workflows, the failure modes. You could explain all of this to a colleague.

But knowing doesn't translate to doing. Without a personal operating system, you'll default to old habits within a week. The lessons will fade. You'll use AI occasionally but not systematically.

The personal playbook is the bridge from learning to practice. It externalizes what you've learned into specific tools you reach for repeatedly. After 30 days of using it, the playbook becomes invisible — the practices are habits, the prompts are remembered, the workflows are reflexive.

This is the work of building durable capability, not just episodic knowledge.

02 · Part One — Your prompt library

Build a file (Markdown, Word, OneNote, Notion — wherever you'll find it) called "My AI Playbook" or similar.

The first section is your prompt library. Below is the structure to use.

Structure for each prompt entry

## [Prompt name]

**Use case:** [When this prompt is appropriate — 1-2 sentences]

**Status:** [Draft / Tested / Reliable]

**Prompt:**
[Full prompt text]

**Common refinements:**
- [Refinement 1 you typically need]
- [Refinement 2]

**Verification steps after running:**
- [What to check]

**Last updated:** [Date]

This level of structure feels excessive at first. After a month it pays for itself.

The 8-12 prompts to start with

Build these now. They cover most of what bench scientists do with AI.

Prompt 1 — Initial orientation to a new area

Start from Lesson 02's orientation template. Customize for your field.

Prompt 2 — Synthesis from provided papers

Start from Lesson 02's synthesis template. Add your team's preferred citation format.

Prompt 3 — Specific question with literature backing

Start from Lesson 02's specific-question template.

Prompt 4 — Adversarial review of a draft

Start from Lesson 02's adversarial-review template. Add the perspectives most relevant to your work (your team's senior scientists, target journal reviewers, regulatory reviewers).

Prompt 5 — Design dialogue

Start from Lesson 03's design dialogue template. Add the specific considerations relevant to your model system (cell line quirks, animal model considerations, assay-specific factors).

Prompt 6 — Protocol drafting

Start from Lesson 03's protocol drafting template. Reference your lab's protocol template.

Prompt 7 — Adversarial protocol review

Start from Lesson 03's adversarial review template. Add specific failure modes you've seen in your own lab.

Prompt 8 — Data analysis code

Start from Lesson 04's basic pattern. Specify your typical tools (Python with which libraries, R with which packages, etc.).

Prompt 9 — Methods section drafting

Start from Lesson 04's methods template. Add your target journals' format requirements.

Prompt 10 — Discussion section drafting

Start from Lesson 04's discussion template.

Prompt 11 — Response to reviewers

Start from Lesson 04's reviewer-response template.

Prompt 12 — Lab notebook entry standardization

Build a prompt based on your lab's notebook conventions that helps you produce clean, complete entries.

How to use this section

Do the prompt-building work this week. Each prompt takes ~10 minutes to customize from the templates. Total time: ~2 hours.

After building, you'll iterate. Use each prompt 2-3 times in real work over the next month. Note what worked, what needed refinement. Update the library. By end of month 1, you'll have 8-12 reliable prompts that match your specific work.

03 · Part Two — Your workflow map

The prompts are the units. The workflows are how you string them together for actual work products.

Identify your top 3-5 workflows

What are the recurring multi-step work products you produce? Examples:

Writing a paper section
Designing and executing an experiment
Reviewing a colleague's manuscript
Preparing a presentation
Writing a grant or progress report

For each, map the steps and the AI prompts involved.

Workflow map template

For each workflow:

## Workflow: [Name]

**Trigger:** [When this workflow starts]
**Output:** [What this workflow produces]
**Time:** [Typical duration without AI / with AI]

**Steps:**
1. [Step name] — [What you do] — [AI prompt from library, if applicable]
2. [Step name] — [What you do] — [AI prompt]
3. ...

**Verification gates:**
- [What gets checked at which step]

**Common variants:**
- [Variant 1 and when it applies]

Example workflow — writing a paper introduction

## Workflow: Writing a paper introduction

**Trigger:** Manuscript is ready for introduction drafting
**Output:** 500-800 word introduction, fully cited and verified
**Time:** 2 days with AI (vs. 5-7 days without)

**Steps:**
1. Define the introduction's argument — manual — write a 200-word "story" of what the introduction must accomplish
2. Identify the literature to draw on — manual — reference list from comprehensive search (Module 04A Lesson 02)
3. Build synthesis matrix — Prompt #2 (synthesis from provided papers) — populate matrix
4. Verify matrix entries — manual — check each cell against the actual paper
5. Draft introduction — Prompt #2 again with the matrix — get ~700 word draft
6. Verify all citations — manual — PubMed search + read where claims are substantive
7. Add specific framing — manual — rework opening and transition to make the argument uniquely yours
8. Adversarial review — Prompt #4 — get reviewer-style critique
9. Address real critiques — manual — selective revision
10. Final polish — manual — voice, tone, format compliance

**Verification gates:**
- Step 4: matrix accuracy before drafting
- Step 6: every citation verified before circulating
- Step 9: critiques addressed or explicitly dismissed with reasoning

**Common variants:**
- Short introduction (250 words): skip the matrix, go directly from outline to draft
- Review-paper introduction: more matrix entries, more synthesis emphasis

This level of explicitness feels heavy. Build 3-5 of these for your most common workflows. Inside a month, they become invisible — you stop reading the workflow document because the steps are habit.

04 · Part Three — Your verification checklist

The third element of the playbook: the verification habits that distinguish responsible AI use from rubber-stamping.

Build a single master checklist

## Pre-finalization AI verification checklist

For any work product where AI was substantially involved:

### Citations and references
- [ ] Every cited reference verified to exist in PubMed
- [ ] Author/year/journal verified
- [ ] For substantive claims, the actual paper read to confirm the attribution

### Quantitative content
- [ ] All numerical values cross-checked against primary sources
- [ ] All statistical methods appropriate for the data
- [ ] All units, doses, concentrations verified

### Technical content
- [ ] Methodology suggestions cross-checked with current best practices
- [ ] No fabricated technical claims
- [ ] No outdated information presented as current

### Voice and framing
- [ ] Output reworked from generic AI voice to authentic voice
- [ ] Specific intellectual content added beyond what AI provided
- [ ] No verbatim large blocks of AI text

### Documentation
- [ ] AI use documented in artifact metadata or version history
- [ ] Tool, model, date noted
- [ ] Verification steps documented

This is a generic version. Customize for your role and the artifacts you produce most often.

Use the checklist mechanically

For high-stakes artifacts (papers, grants, regulatory documents), literally walk through the checklist. Don't skip items because they "feel fine."

For lower-stakes artifacts (internal memos, exploratory work), the checklist is mental. You internalize it; you run through it implicitly.

After 30 days of using it mechanically, the internal version becomes reliable.

05 · Part Four — Your 30-day plan

The final element: a specific plan for what you'll practice over the next month to internalize the capability.

Week 1 — Library building

Goal: Build out the 8-12 prompts from Section 02. Save them in your playbook.

Daily commitment: 30 minutes/day to customize prompts and test them on simple tasks.

End-of-week check: Do you have 8-12 prompts that work for your real work? Have you tested at least 3 of them on real tasks?

Week 2 — Workflow mapping

Goal: Build out 3-5 workflow maps (Section 03). Use them on real work.

Daily commitment: Use at least one workflow per day on real work. Note what works and what doesn't.

End-of-week check: Are you using AI more deliberately than you were two weeks ago? Are your outputs noticeably better or faster?

Week 3 — Verification habituation

Goal: Run the verification checklist (Section 04) on every AI-assisted work product. Track adherence.

Daily commitment: Don't ship anything AI-assisted without a verification pass.

End-of-week check: How many AI errors did the verification catch this week? (Some — that's good; verification is working. None — be skeptical, you may be verifying too lightly.)

Week 4 — Workflow refinement and the day-30 wall

Goal: Refine your playbook based on three weeks of practice. Address the day-30 wall (from Module 01 Lesson 03) — the temptation to stop investing in this capability.

Daily commitment: 15 minutes/day reviewing what's working and what isn't. Update the playbook.

End-of-week check: Has the capability become reflexive? Are you faster than you were a month ago? Are you more careful, not less?

Beyond 30 days

After the first month:

The playbook stops being a daily resource and becomes a reference
New prompts get added as new work types emerge
Workflows get refined as you learn what actually works
The verification habit becomes invisible — you can't skip it even if you tried

You're now at competent. The path to fluent is another 60-90 days of regular practice. The path to expert is 12-18 months of practice plus deliberate teaching others.

06 · The personal capstone exercise

Spend the next 90 minutes building your playbook.

Allocate the time as:

30 minutes: prompt library (Section 02)
30 minutes: workflow maps (Section 03)
15 minutes: verification checklist (Section 04)
15 minutes: 30-day plan (Section 05)

Don't aim for perfection. Aim for a starting version you'll iterate from. Version 1 of your playbook is worse than version 5. Version 5 only happens if you build version 1 today.

If you do this exercise, you finish Module 04A with an actual capability, not just understanding. If you skip it, you finish with understanding that will fade. The exercise is the deliverable.

07 · A meta-observation

A specific thing worth noting as you finish your role-path module:

You started this curriculum because some combination of curiosity, urgency, and ambition brought you here. You're now five modules in. Three more required modules and (depending on your path) some advanced modules remain.

Look back at what you knew when you started.

If you started this curriculum unable to write a specification, unable to classify data correctly, unable to identify AI failure modes confidently — and you can now do all of those — you've already made the change that 95% of biotech professionals haven't made yet.

The remaining work (advanced modules, capstone) sharpens and extends the capability. But the foundational shift has happened. You're now equipped to use AI as the productivity multiplier it can be, while staying inside the compliance and quality standards that biotech requires.

This is what fluency feels like. Not "AI can do my job." Not "AI is dangerous and I should avoid it." But "I know how to use AI as a tool to do my work better, faster, and with appropriate caution."

That's the goal. You're there. Keep going.

08 · Knowledge check

Three questions to lock in this lesson.

Q1. What's the difference between knowing the curriculum content and having durable capability?

a) Knowing is sufficient; capability follows automatically b) Durable capability requires externalizing the learning into specific tools (prompts, workflows, checklists) that you reach for repeatedly — without this, the lessons fade c) Capability requires a year of practice d) There's no difference

Q2. Which of these is NOT one of the four elements of the personal playbook?

a) Prompt library b) Workflow map c) Verification checklist d) Performance metrics dashboard

Q3. Why is the day-30 wall specifically called out as a risk?

a) Most people quit at day 30 because they're tired b) Day 30 is when initial easy wins have plateaued but deeper capability is still being built; people are tempted to declare success and stop investing, missing the compounding gains that come later c) Day 30 is when AI tools change d) Day 30 is when verification fatigue sets in

Answers: Q1: b · Q2: d · Q3: b

09 · End of Module 04A

You've finished the bench R&D role path.

You can now:

Identify the eight core workflows where AI helps bench scientists most
Use AI for literature review without producing fabricated citations
Use AI for protocol design and methodology critique
Use AI for data analysis with appropriate verification
Use AI for scientific writing while preserving your voice
Maintain a personal playbook that compounds your capability over time

Path forward:

Advanced modules (Modules 05-10) — Connectors, Agent Design, Skills, Operating Model, Cross-Functional Excellence, Capstone
Cross-role learning — if your role spans multiple paths, consider another role-path module relevant to your work
Practice — the playbook you built today only matters if you use it. Use it.

Whatever you choose next, the work of becoming AI-fluent has begun. The next 60-90 days of practice will harden the capability. The next 12-18 months will distinguish you in your field.

Good luck.

End of Module 04A.