AI4RA Tip of The Week #17

What is context engineering? 
Context engineering is the practice of deliberately designing an information environment for an AI to operate within. The key distinction between prompt engineering (PE) and context engineering (CE) is that PE is centered around asking smart questions while CE is about building smarter environments for the AI to work in. 

When you type a prompt into an AI tool, the model doesn’t just respond to your words. It draws from everything it can “see” at that moment: your prior messages, any documents you’ve uploaded, instructions built into the tool, and more. That full set of information is called its context, and the quality of that context greatly impacts the quality of the response the AI generates. 

In practical terms, CE involves four skills working together:

• Intent — establishing the big-picture goal and making clear what success looks like (and what the AI should avoid, like guessing when information is missing)
• Information — providing the documents, policies, data, and examples the AI draws from to produce accurate outputs
• Instruction — structuring the task with step-by-step directions, formatting requirements, and constraints (this is also known as prompt engineering)
• Interaction — refining the result over multiple turns through feedback, corrections, and summarization 

Together, these determine whether an AI produces results you can trust or results you can’t. It is important to remember that CE doesn’t replace PE. Context is often built into the prompt itself, such as establishing the goal of the prompt before asking your question, which expands the scope of the prompt to include things beyond the instructions provided to the model. However, CE can exist outside the prompt as well, such as providing resources (documents, websites, etc.) to bound the AI within.  

Why this matters for Research Administrators 
Research administrators already think this way. When you onboard a new grant specialist, you don’t just hand them a task and say, “figure it out.” You give them institutional policies, sponsor guidelines, templates, historical examples, and the unwritten rules about how things work. You build an environment that makes them effective. 

CE applies the same principle to AI. Instead of asking “Can you draft a budget justification?” and hoping the AI guesses the right requirements, you give it the FOA, your institution’s rate agreements, the PI’s prior budgets, and the relevant sections of the Uniform Guidance before you ask. The prompt becomes simpler because the hard work has already been done. It is important to remember that a chatbot will always try to provide an answer even if it doesn’t have the right references and resources. Giving it the correct context will make it more likely that the answer it provides will be accurate and trustworthy. RAs are already masters at organizing information and designing processes, and CE allows them to use those skills to leverage AI as effectively as possible.  

Context engineering also improves traceability. When you ground an AI in specific documents and resources, its answers are tied to those sources, making it easier to verify where the output came from. This builds your own confidence in the result and gives colleagues or auditors a clear trail to follow if they need to check the AI’s work 

How to use this resource 
You don’t need to be a developer to start applying context engineering. Here are four places to start today:

1. Upload documents before asking questions. When you drop a FOA into your AI tool before asking about eligibility requirements, you’ve done the most important act of context engineering: giving the AI what it needs to stop guessing.
2. Set standing instructions. Many AI tools let you configure persistent context. Setting these once means you don’t re-explain your situation every session.
3. Be selective about what you include. More isn’t always better. Choosing the right three pages from a 200-page sponsor manual is itself context engineering. Overloading an AI with irrelevant documents can reduce its accuracy.
4. Build repeatable workflows. Don’t reinvent the wheel. When you develop a prompt and context that works for a task you do frequently, save it for next time. 

How can I learn more about this topic?  
For a deeper dive, read the full article by Nate Layman on the AI4RA blog: https://ai4ra.uidaho.edu/from-prompt-engineering-to-context-engineering/
AI4RA Office Hours – Occurs every fourth Tuesday at 11 AM PST and includes live Q&A and demos (https://ai4ra.uidaho.edu/#events). 

Share your questions or success stories  
Post them to the AI4RA community forum (https://groups.ai4ra.uidaho.edu/g/main/topics). Your tips help the whole research‑admin community thrive.