The 3D Framework

Define: Where Your Expertise Shines

Before you touch any AI tool, you need to get clear on three things. This is where your institutional knowledge matters most.

1. Specific Problem

What repetitive task drains your time but requires your knowledge? Be precise. Not "communication is hard" but "every NOFO analysis requires extracting the same 15 data points and I do 30 of these per year."

The more specific, the better the solution.

2. Outputs Desired

What does "done" look like? Who needs what format? If you're analyzing funding opportunities, does the PI want a one-page summary? A comparison table? Talking points for a meeting? Define the deliverable before you build.

3. Data Risk Levels

What information is involved? Public data like NOFOs and agency strategic plans? Internal documents? Anything with trainee information or proprietary research? Your risk level determines which tools you can use and what guardrails you need.

Veterans know this instinctively. The key is to document it explicitly so anyone can understand the workflow.

Develop: Building Your Knowledge Infrastructure

This is where you prepare everything the AI needs to actually help you. Think of it as training a new employee, except you're structuring information for a machine.

4. Capture Context

Document the knowledge that lives in your head and your colleagues' heads. The examples, edge cases, exceptions, and preferences that make you good at your job. Record interviews, transcribe meetings, pull together the reference documents you actually use.

5. Clean and Structure

Make your information AI-ready. Create templates, decision trees, and checklists. Think: if I were training someone new, what would they need to see? Structure your context so it's searchable and usable.

6. Build Tools

Now you create the actual workflow: a custom GPT, a prompt template, or an automated process. If you're not comfortable building, partner with someone who is. Test with real scenarios. Your judgment validates whether the outputs are actually useful.

This is legacy building. You're not just solving today's problem. You're creating systems that can outlast you.

Deploy: You Stay in Control

The human-in-the-loop isn't a weakness. It's the entire point. AI scales your judgment. It does not replace it.

7. Generate Content

Let AI create the first draft based on your context. This is where you get consistency at scale: the same quality, every time, handling the repetitive heavy lifting so you can focus on the parts that need human thinking.

8. Expert Review

This step is non-negotiable. You are the quality gatekeeper. Check for accuracy, tone, completeness, and institutional fit. Refine the system based on what needs correction. Every review makes the workflow better.

9. Distribute

Deploy the approved output to the right audiences. Track what works and what doesn't. Iterate and improve. A good workflow gets better over time as you learn from each cycle.

Example: NOFO Analysis Workflow

Here's how the framework applies to a common research development task: analyzing funding opportunity announcements.

Define

Problem: Each NOFO takes 45 minutes to analyze, and you review 30+ per year. Faculty need quick summaries to decide whether to pursue.

Output: A one-page brief with eligibility, key dates, budget limits, review criteria summary, and fit assessment.

Data risk: Low. NOFOs are public documents.

Develop

Capture: Your mental checklist of what makes an opportunity worth pursuing for your institution. The questions faculty always ask.

Structure: A template with consistent sections. A list of your institution's strategic priorities to check against.

Build: A prompt template or custom GPT that extracts the key information and formats it consistently.

Deploy

Generate: Paste the NOFO, get a draft brief in 2 minutes.

Review: Verify accuracy, add your institutional knowledge, flag anything the AI missed.

Distribute: Send to relevant faculty. Track which briefs led to submissions.

Total time drops from 45 minutes to 10. Multiply that by 30 opportunities per year, and you've recovered meaningful capacity for higher-value work.

Getting Started

Pick one workflow. Something repetitive, time-consuming, and within your control. Work through the nine steps. Learn what breaks and what works. Then do it again with the next workflow.

If you're finding it hard to even start, you may need to address your context readiness first.