AI
Your AI Guide
Practical AI for University Staff

Just Start Talking. Then Iterate.

You don't need to write perfect prompts. Modern AI is smart enough to work with simple questions. The real skill? Knowing what to say next.

For Systems & Automation

Engineered Prompts

When AI runs automatically — in apps, pipelines, or integrations — every prompt must be carefully tested and optimized. There's no human in the loop to fix a bad output.

  • Untested prompts waste money (you pay per request)
  • Bad outputs can reach users or break workflows
  • Security risks if inputs aren't controlled
  • This is prompt engineering — a technical skill
For You — Everyday Use

Start Simple. Iterate.

When you're chatting with AI directly, you don't need a perfect prompt. Just ask, look at what comes back, and guide it with follow-ups. You're in the loop — you are the quality control.

  • Feels like talking to a colleague
  • Anyone can do it — no technical skills needed
  • Each follow-up makes the result better
  • You discover what you want along the way
You I need to create a syllabus for my course
AI I'd be happy to help! Could you tell me the course name, how many weeks the semester is, and who your students are?
The AI asks for what it needs. You don't have to think of everything upfront.
You Introduction to Sociology, 15 weeks, 2nd year BA students. About 40 students.
AI Here's a draft syllabus with 15 weeks of topics, learning outcomes, and suggested readings...
You Good start, but weeks 8-12 are too theoretical. I want more interactive sessions — workshops, debates, case studies. Also add a group project.
This is where the magic happens. You react to what you see and steer.
You Almost perfect. Move the exam prep to week 14 and add a Lithuanian case study in week 6.
AI Done! Here's the updated syllabus with those changes...
Happy with the result? Ask the AI to turn this whole conversation into a reusable template. ↓ See "Save It" section below.

Build a Course Syllabus

Any LLM

Start with your course basics, then shape the syllabus through a few back-and-forth exchanges.

1
Start simple
Help me create a syllabus for my [course name] course. It's [X weeks], for [student level] students.
2
After you see the first draft
Weeks 5-8 are too lecture-heavy. Replace some with interactive formats — seminars, workshops, or group activities. Also the topics in weeks 10-12 overlap, merge them and free up a week for student presentations. React to what you see. Tell it what's wrong, what's missing, what to change.
3
Polish
Add a learning outcome for each week using Bloom's taxonomy verbs. And suggest one reading per week — prioritize open-access sources.
When you're happy
This is great. Now analyze our entire conversation and create a reusable prompt template I could use next semester for a different course. Include all the context and constraints we discovered along the way.

Design an Engaging Lecture

Any LLM

Turn a dry topic into a 90-minute session students actually enjoy. Start with the basics, refine the activities.

1
Start simple
I have a 90-minute lecture on [topic] for [student level] students. Help me plan it so it's not just me talking for 90 minutes.
2
Shape the activities
The group activity in the middle is too complicated for 40 students in a lecture hall. Give me something simpler — like a think-pair-share or a quick poll they can do from their seats.
3
Get specific
Write out the exact discussion questions I should ask after the activity. And give me a strong opening — something that grabs attention in the first 30 seconds.

Generate Discussion Questions

Any LLM

Get questions for your seminar. Start rough — then calibrate difficulty and add Lithuanian/European context.

1
Start simple
Give me discussion questions about [topic] for a university seminar. Students have read [source/chapter].
2
Calibrate
Questions 1-3 are too basic — my students already know this. Replace them with questions that require comparing perspectives or applying theory to a real case. Add at least one question about the Lithuanian context.
3
Add facilitator notes
For each question, add a short note in brackets — what kind of answers to expect, and a follow-up question if the discussion stalls.

Create Exam Questions

Any LLM

Get a balanced exam draft, then adjust difficulty, fix ambiguous wording, and add your own twist.

1
Start simple
Create exam questions for my [course name] course. Topics to cover: [list 3-5 topics]. Mix of multiple choice and essay questions.
2
Adjust difficulty
The multiple choice questions are too easy — a student could guess without studying. Make the distractors more plausible. And the essay questions are too broad — narrow them so a student can answer in 300 words.
3
Add grading info
Add point values and a marking guide for each essay question — what should a 9/10 answer include vs a 6/10 answer?

Build a Grading Rubric

Any LLM

Get a first rubric draft, then sharpen the criteria until each cell clearly distinguishes one level from another.

1
Start simple
I need a grading rubric for a [assignment type, e.g., research essay] in my [course]. Total points: [e.g., 100].
2
Sharpen distinctions
The descriptions for "Good" and "Excellent" are too similar — I can't tell them apart. Make each level clearly different. Use specific observable criteria, not vague words like "demonstrates understanding."
3
Add student version
Now make a student-facing version — same criteria but with friendlier language and a short explanation of what I'm looking for in each category.

Write Student Feedback

Claude / GPT

Give the AI your quick notes on a student's work, iterate until the feedback sounds like you — not like a robot.

1
Start simple
Write feedback for a student's essay. Grade: [e.g., 7/10]. Good parts: [what worked]. Problems: [what didn't].
2
Fix the tone
Too formal and generic. Make it sound warmer — like I'm talking to the student in my office, not writing an official report. Use "I noticed" instead of "The student demonstrates."
3
Make it actionable
Add one specific thing the student should do differently in their next essay — not generic advice, something concrete based on what they got wrong here.

Draft a Research Abstract

Claude

Start with your messy notes, iterate until the abstract is tight enough for journal submission.

1
Start simple
Help me write an abstract for my paper. Topic: [your topic]. Method: [what you did]. Main finding: [key result]. Word limit: [e.g., 250].
2
Tighten it
The background section is too long — cut it to 2 sentences max. Put more weight on the results. And include the specific numbers from my findings, not just "significant differences were found."
3
Match the target
I'm submitting to [journal name]. Check if the structure matches their abstract requirements. Also suggest 5 keywords for indexing.

Organize Literature Review

Claude / GPT

Paste your messy reading notes, then iterate to find themes, gaps, and structure for your review.

1
Start simple
I'm writing a literature review on [topic]. Here are my notes on the sources I've read so far: [paste rough notes]. Help me organize this.
2
Find the gaps
Good overview. Now identify what's missing — what topics or perspectives aren't covered by these sources? What gap could my research fill?
3
Structure it
Create a comparison table: Author/Year, Method, Key Finding, Limitation, How it relates to my research. Then suggest a thematic structure for writing up the review.

Prepare Conference Presentation

Any LLM

Go from paper to slides. Start with your key points, iterate on the story arc and visuals.

1
Start simple
I have a [15/20-minute] conference talk about [topic]. My main argument is [1-2 sentences]. Help me structure the slides.
2
Fix the story
Too many slides on background — my audience already knows the context. Cut that to 2 slides and add more on my results. I need a stronger opening — something that makes people put down their phones.
3
Get speaker notes
Write speaker notes for each slide — what I should actually say, in a natural conversational tone. Not a script, just key talking points.

Clean Up Meeting Notes

Any LLM

Paste your messy meeting scribbles, iterate until you have clean minutes with clear action items.

1
Start simple
Here are my rough notes from today's [meeting name] meeting. Turn them into proper meeting minutes. [paste notes]
2
Add structure
Pull out all action items into a separate table at the bottom: What needs to be done, who's responsible, and by when. Mark anything with an unclear deadline as "[to confirm]."
3
Polish for distribution
Make the tone more neutral — remove any editorializing. And add a one-paragraph executive summary at the top for people who won't read the full minutes.

Draft a Policy Document

Claude

Start with your key points, iterate on structure and legal precision. Great for AI use policies, assessment rules, etc.

1
Start simple
I need to draft a policy on [e.g., AI use in student assessments] for our faculty. Key rules I want to include: [list 3-5 points].
2
Tighten the language
Section 3 is too vague — "appropriate use" means nothing. Replace with specific examples of what's allowed and what isn't. Each rule should be one clear sentence with no room for misinterpretation.
3
Add enforcement
Add a section on what happens when someone violates the policy — who reports it, what's the process, what are the consequences. Keep it proportional, not punitive.

Write a Formal Email

Any LLM

Get a draft email, then adjust the tone and sharpen the ask until it's exactly what you'd send.

1
Start simple
Help me write an email to [who, e.g., a partner university in Sweden] about [what, e.g., proposing a joint summer school].
2
Fix the tone
Too stiff. Make it collegial — like I've met this person at a conference before. And get to the point faster — the ask should be in the first paragraph, not buried at the end.
3
Sharpen the CTA
End with a specific next step — suggest a meeting date or ask them to reply by a certain date. "Looking forward to hearing from you" is too passive.

Reply to Student Emails

Any LLM

Paste the student's email, get a draft reply, then adjust empathy and specificity.

1
Start simple
A student sent me this email: [paste email]. Help me write a clear, helpful reply. The relevant rule/info is: [what they need to know].
2
Adjust tone
The student sounds stressed. Make the reply warmer — acknowledge their situation before jumping to the answer. But keep it under 150 words, they won't read more than that.
3
Add specifics
Add the exact deadline, the office they need to visit, and the form they need to fill out. Number the steps so it's impossible to miss anything.

Write Event Descriptions

Any LLM

Create compelling event copy for workshops, guest lectures, or open days that people actually want to attend.

1
Start simple
Write a description for our event: [event name], happening [date] at our university. It's a [type] about [topic], aimed at [audience].
2
Make it compelling
Too dry — reads like a bureaucratic announcement. Lead with why someone should come, not when and where. What will they learn or gain? Make me want to go.
3
Get both versions
Give me two versions: the full description for the website (200 words) and a short teaser for email/social media (50 words max). Both should have a clear call-to-action.

Write Recommendation Letters

Claude

Give the AI your honest impressions of the student, iterate until the letter sounds like you wrote it yourself.

1
Start simple
Write a recommendation letter for [student name] who's applying to [what]. I taught them in [course], they got [grade]. They were [your honest impression].
2
Add a real story
Too generic — could be about anyone. Include this specific story: [describe a moment, project, or interaction that shows the student's quality]. That's what makes a letter memorable.
3
Match my voice
This sounds like AI wrote it. Shorter sentences. Less flowery language. I'd write "She's one of the best students I've had" — not "It is with great enthusiasm that I recommend this exceptional scholar."

Summarize a Long Document

Claude / Gemini

Paste a long report, regulation, or strategy paper. Get the gist first, then drill into what matters for you.

1
Start simple
Summarize this document for me. I'm a [your role] and I need to understand how it affects my work. [paste document]
2
Focus on what matters
I only care about sections related to [specific area]. What are my specific action items? Are there any deadlines I might miss?
3
Get the briefing
Write me a 5-sentence briefing I can forward to my team. Include only what they need to know and what they need to do. Skip the background.

Plan a Presentation

Any LLM

From vague idea to slide outline. Start with "I need to present about X" and iterate from there.

1
Start simple
I need to give a [X-minute] presentation about [topic] to [audience]. My goal is to [inform / persuade / get approval]. Help me plan the slides.
2
Reshape the story
The structure is too predictable — intro, body, conclusion. Start with a surprising fact or a question instead. And slides 5-7 have too much text — one idea per slide, max 3 bullet points.
3
Get talking points
For each slide, write 2-3 sentences I should say — conversational, not a script. And suggest what visual would work best on each slide (chart, photo, diagram, or just a bold quote).

Polish or Translate Text

Claude / GPT

Paste your draft, get it polished. Works for English, Lithuanian, or translating between them.

1
Start simple
[Choose one]: "Polish this text for a journal submission" / "Translate this from Lithuanian to English" / "Make this simpler — it's for a general audience." Here's the text: [paste]
2
Protect your voice
You changed my meaning in paragraph 2 — I said X but you made it sound like Y. Fix that. Also, keep my sentence style — don't make everything long and formal. I write in short sentences on purpose.
3
Review changes
Show me only the changes you made — list each one with a brief reason why. I want to learn from the edits, not just accept them blindly.

Custom GPT

Save as a Custom GPT in ChatGPT. It becomes a reusable assistant you can open anytime — no re-explaining needed.

Google Gem

Save as a Gem in Gemini. Same idea — your refined workflow becomes a one-click tool in Google's ecosystem.

Claude Project

Save as a Project in Claude. Add your template to the project instructions — Claude will follow it every time.

The Magic Prompt — paste this after a successful conversation
Look at our entire conversation from the beginning. Analyze what worked: - What context did I give you that was essential? - What follow-up instructions improved the output? - What constraints or preferences did we discover along the way? Now create a reusable prompt template that captures all of this. It should: 1. Include all the essential context as fill-in-the-blank variables 2. Bake in the constraints and preferences we discovered through our back-and-forth 3. Be written so that anyone could use it and get a similar quality result on the first try Format it so I can paste it directly into a Custom GPT / Gem / Claude Project as system instructions.

Google NotebookLM is a free AI research tool that works only with the documents you upload. Unlike ChatGPT or Gemini, it doesn't make things up from the internet — it reads your sources and answers based on them. Think of it as an AI assistant that actually did the reading. Open it at notebooklm.google.com — it's free with your Google account, including Workspace for Education.

🎧

Audio Overviews

Upload any document and get a podcast-style discussion between two AI hosts explaining the content. With Interactive Mode, you can interrupt mid-conversation to ask questions — the hosts pause, answer you, then continue.

🎬

Video Overviews

Generates narrated videos with visuals, diagrams, and animations from your documents. The new Cinematic mode creates dynamic, engaging explainer videos — not just narrated slides.

📚

Studio Tools

One-click generation of quizzes, flashcards, mind maps, slide decks, infographics, and data tables — all from your uploaded sources. Slides export to PPTX for editing.

📊

Data Tables

Upload multiple documents and generate comparison tables automatically. Export to Google Sheets. Great for comparing research papers, policies, or student submissions.

🔎

Source-Grounded Chat

Ask questions and get answers with inline citations pointing to exact passages in your sources. Every claim is traceable — no hallucinations from outside data.

🔗

Gemini Integration

Connect your notebook to Google Gemini. Ask Gemini to "answer based on my notebook" from anywhere in Google's ecosystem — even in Docs or Gmail.

How University Staff Can Use It

1
Prepare lectures faster

Upload your textbook chapters, past slides, and articles. Ask NotebookLM to summarize key concepts, generate discussion questions, or create a quiz — all grounded in your actual materials.

2
Create study materials for students

Generate Audio Overviews of your lecture notes — students can listen on their commute. Create flashcard sets and practice quizzes from your course content with one click.

3
Compare research papers

Upload 5-10 papers into one notebook. Ask "What methods were used across these studies?" or "Where do the authors disagree?" — get answers with citations to specific papers.

4
Digest long policy documents

Upload an EU regulation or ministry report. Ask "What does this mean for our faculty?" — get a plain-language summary based only on the actual document, not AI guesses.

5
Onboard new colleagues

Upload department handbooks, procedures, and policies into a shared notebook. New staff can ask questions and get answers grounded in your actual documents.

6
Generate presentations from documents

Upload a report or research paper, generate a slide deck in Studio, then export to PPTX and edit in PowerPoint or Google Slides. First draft in seconds.

Output is too generic Give it specifics from your real situation
"Use this real example: ..."
Output is too long Set a hard limit
"Max 200 words"
Tone is wrong Describe how it should feel
"More casual, like talking to a colleague"
Some parts are good, some bad Point to specific parts
"Keep section 2, rewrite section 4"
Wrong format Show what you want
"Put this in a table instead"
Missing your context Add the specifics
"I should mention that at our organisation we..."
Too complex Ask for simplification
"Explain like I'm a first-year student"
Need a different angle Tell it what perspective to take
"Now approach this from a student's view"
Almost perfect Fine-tune the last details
"Just fix the intro and shorten the conclusion"
Want options Ask for variations
"Give me 3 versions: formal, friendly, brief"
You can't break it

There's no wrong way to ask. If the result isn't good, just tell the AI what's wrong. It won't judge you, get offended, or remember your awkward first attempt.

Paste your raw material

AI is much better at improving your draft than creating from nothing. Paste your messy notes, half-written text, or rough ideas — even in Lithuanian.

3 follow-ups is normal

Getting a perfect result on the first try is rare. Most good results take 2-4 back-and-forth exchanges. That's not failure — that's how it works.

Always verify facts

AI can confidently present wrong information — especially dates, statistics, and citations. If it mentions a study or gives a number, double-check it yourself.

Don't share sensitive data

Never paste student grades with names, personal information, or confidential university documents into public AI tools. Anonymize first.

The AI isn't the author — you are

Use AI as a starting point, not a replacement. Your expertise, judgment, and voice are what make the final output trustworthy.