Behind the Feed: My Weekly Workflow for Learning Center AI

Every Monday at 8 a.m., a new article goes up here. A few people have asked how I keep that going while also running a learning center. The short answer: AI does a lot of the legwork, and I make the calls.

Here's the whole workflow, start to finish, including the tools. If you put out a staff newsletter, a tutor training blog, or a monthly update for your dean, most of this carries over.

1. Find the sources

Most of what ends up on the site starts with my own reading. I follow a handful of AI newsletters and sites, and when something looks useful for learning centers, I grab it.

I also have help. In Muse, Meta's AI assistant, I built an agent I call Elvis. Elvis's job is to find and curate articles about AI that might matter to people who run tutoring, coaching, and academic support programs. It catches things I'd miss and saves me from scanning everything myself.

Elvis doesn't decide anything, though. It's a scout, not an editor. Whatever it flags still has to get past me, and plenty of it doesn't.

There's a second reason I use Muse. I try to use and test a lot of AI tools, because that's how I understand what our students are actually using. Meta's AI shows up in apps students already have open all day, so it's worth knowing firsthand what it does well and where it falls short.

2. Clip it, with a note

When something is worth keeping, I save it in a Craft space set up just for the site. Then I write a short note on why I clipped it: what caught my eye, and who in a learning center would care.

Those notes matter more than the links. They're my thinking on the page. They're also what lets AI help me later without guessing what I meant. A link with no note is just a bookmark. A link with "this would change how we train new tutors on AI policy" is the start of an article.

3. Wednesday prep with Claude

On Wednesdays, I go through the week's clips and notes with Claude. We look at what connects, what's strong enough to carry a full article, and which of the site's four categories it fits: Coaching & Pedagogy, Leadership & Operations, Data & Dashboards, or Tutor Training.

Claude suggests. I pick. Once I've chosen a topic, Claude helps me outline and draft the article and the prompts that go with it.

4. Revise and check

Every draft gets my edit. I check each claim against its source, cut anything that doesn't sound like me, and add the details only someone who has worked in a center for years would know. If I can't verify something, it comes out.

This is the step I'd never hand off. AI can write a confident sentence about a study it misread. Catching that is my job.

5. Test the prompts

Every article ends with prompts learning center staff can use to do whatever the article describes. Before a post goes live, I test those prompts in ChatGPT, Claude, Copilot, Gemini, and Grok. Your campus might provide any one of those, so the prompts need to work in all of them.

I keep the prompts in a separate Craft collection. Once a post goes live, its prompts get added to the site's prompt library.

6. Publish through Studio

The site is built in RapidWeaver Elements. Behind it is LCAI Studio, a content manager I built so I'd have one place to run everything: articles, scheduling, authors, and the newsletter.

Studio connects to Claude, so Claude can save a draft straight into Studio instead of me copying and pasting between windows. It saves drafts only. Nothing goes live until I review it and schedule it for Monday morning.

7. Say how AI was used

Every article ends with an AI use statement written for that specific piece: which tools I used, what each one did, and an APA reference for each. You'll find one at the bottom of this post.

We ask students to be open about how they use AI. I think we should model that ourselves.

8. The Friday email

On Fridays, the newsletter goes out with the week's three most-read articles and a few extra links under "Also worth your time." Studio builds the draft. I read it, tweak it, and hit send.

Why I keep a human in the loop

AI saves me hours every week. It finds sources, drafts, and handles the busywork. But it doesn't own the decisions: what to cover, whether a claim holds up, and what goes in front of readers. Those stay with me.

That's the same thing we ask of tutors. Let AI take the friction that doesn't matter, and keep the thinking that does.

Try It

These prompts work in whatever AI tool your campus provides. Fill in the bracketed parts before you paste. None of them need student data, so keep student names and ID numbers out.

1. Write instructions for a source-scouting agent. For directors and coordinators who want help keeping up with AI news for a staff newsletter or training updates. Paste the result into any tool that lets you build a custom assistant (Muse, a custom GPT, a Gemini Gem, a Claude project, or a Copilot agent).

You are an expert at writing instructions for custom AI assistants.

Goal: write the instructions for an assistant that finds and curates articles about AI that matter to my learning center staff.

Context:
- My role: [e.g., director of a tutoring center, academic coaching coordinator]
- Who will read what I share: [e.g., professional staff, peer tutors, my dean]
- Topics I care about most: [e.g., AI and tutoring, academic integrity, tutor training, student data privacy]
- Topics I want it to skip: [e.g., stock news, product launch hype, K-12 classroom tips]
- Sources I already trust: [LIST NEWSLETTERS OR SITES, or write none yet]
- How often I want results: [e.g., weekly, 5 to 8 items]

Write instructions that tell the assistant to:
1. Search for recent articles that fit my topics and audience.
2. For each article, give the title, source, link, date, and one sentence on why it might matter to a learning center.
3. Flag anything that looks like opinion, marketing, or an unverified claim.
4. Never summarize an article it couldn't actually open.
5. Leave the final decision about what to share to me.

Keep the instructions under 300 words, in plain language, written directly to the assistant.

2. Turn your clipped notes into an outline. For anyone writing a staff newsletter, blog post, or training update from links they've saved.

You are an editor who helps learning center leaders turn their reading into short, practical articles for staff.

Goal: help me choose a topic from my saved links and outline one article.

Context:
- Audience: [e.g., peer tutors, professional staff, other learning center directors]
- Article length: [e.g., 800 to 1,000 words]
- Tone: [e.g., direct and informal, plain language]

Here are my saved links, each with my note on why I saved it:
[PASTE EACH LINK WITH YOUR NOTE]

Give me:
1. Two or three possible article topics that connect at least two of my links, with one sentence each on the angle.
2. Ask me which topic I want before outlining.
3. Once I choose, an outline with section headings, the main point of each section, and which link supports it.

Base the outline on my notes, not on what you assume the articles say. If my note and a link seem to disagree, ask me about it instead of guessing.

3. Draft an AI use statement for a piece of work. For staff who want to be open about how they used AI in a report, newsletter, or training material.

You are helping me write a short, honest AI use statement for something I created.

Goal: draft a two- to four-sentence statement that says which AI tools I used and what each one did, plus an APA 7 reference for each tool.

Context:
- What I made: [e.g., a staff newsletter, a tutor training handout, an annual report section]
- Tools I used and what each one did: [e.g., ChatGPT brainstormed section headings; Gemini summarized two articles I then read in full]
- What I did myself: [e.g., chose the topic, checked every claim, wrote the final version]
- Version and year for each tool, if I know them: [LIST]

Write the statement in first person and plain language. Only include tools and uses I listed. Do not add or guess at anything I didn't tell you. Then list the APA 7 references, and flag any detail you need from me to finish them.

AI use statement

Claude (Anthropic, 2026) helped me turn my notes on how I run the site into an outline and draft of this article and its prompts. I revised the draft and decided what got published.

References

• Anthropic. (2026). Claude (Opus 5.5 version) [Large language model]. https://claude.ai

About the author

[[shortcode.authors.mark-woolwine]]