From Vending Machine to Board of Advisors: AI Thinking Partners in the Learning Center

Doan Winkel's Board of Advisors framework gives learning centers a concrete way to teach students to use AI as a thinking partner instead of an answer machine. And honestly, we're better positioned to teach it than almost anyone else on campus.

Most of us have seen it in sessions. A student opens a chatbot, pastes in the assignment, tweaks a few words in the output, and calls it done. Winkel calls this the vending machine approach. The student gets a product and skips the thinking the assignment was supposed to build.

This past summer, Winkel published a seven-part series on a different model. Each student builds a small panel of AI expert personas for each class, then uses that panel to argue, practice, and pressure-test their work all semester. He wrote it for faculty. But a lot of it looks like what learning centers already do.

Why learning centers? Three reasons:

• We work across every discipline, so we see a student's whole semester, not one course.
• We already teach the habits this framework runs on: metacognition, deliberate practice, and self-testing.
• We meet students one-on-one and in small groups, which is where a habit like this actually gets built.

The framework in brief

The system has seven layers, each building on the last. Here's the short version:

1. Build the board. For each class, the student sets up a saved AI project with 4 to 7 personas: a few core experts in the field, one or two from related fields, and one or two skeptics who push back on the mainstream view.
2. Add friction. Before asking the board anything, the student writes down their own answer and predicts what each advisor will say. They also argue the other side at full strength and keep a weekly thinking journal.
3. Drill. The student breaks the course into 10 to 20 specific skills ("compare two theories," not "Chapter 4") and practices one at a time with feedback, tracking results in a practice log.
4. Get reviewed before the professor does. Before turning in a big assignment, a harsh reviewer tries to tear it apart, a set of graders scores it against the real rubric, and a pre-mortem asks why it might fail.
5. Practice out loud. Using voice mode, the board grills the student live. They practice what to do when stumped, then debrief the transcript.
6. Connect the classes. A "Meta-Board" sits above all the class boards and helps the student work through problems that cross disciplines.
7. Become the board. The student predicts each advisor's response before asking, retires advisors whose thinking they've absorbed, and works on problems with no AI at all.

That last step is the point of the whole thing. The AI is scaffolding, and the goal is for the student to carry those voices in their own head.

Where it fits in learning center work

You don't need to run the full system to get value from it. Different pieces map onto services most centers already offer.

Academic coaching

The friction layer (step 2) is basically good coaching with a new tool. Asking a student to write their answer and predict the board's response before they open the AI is a metacognition exercise. The thinking journal gives you something concrete to review together at each check-in.

The skill breakdown from step 3 also makes a great coaching activity, with or without AI. Sit down with the syllabus and turn the learning objectives into verbs the student can practice. The practice log that follows gives both of you real data on progress.

Tutoring

Tutors can use the last 10 minutes of a session to model a drill session (step 3). The student leaves with a plan for practicing between sessions instead of just rereading notes.

The rubric graders from step 4 work well in a session too, as long as the course allows it. Run them together, then talk through which feedback is actually right. The tutor's job becomes helping the student judge the AI's feedback, not just accept it.

Presentation and oral exam prep

Step 5 is a natural fit for students facing oral exams, clinical check-offs, presentations, or interviews. Voice mode lets them rehearse with a tough questioner as many times as they need.

The recovery drill is worth teaching even without AI. When stumped, the student names it honestly, restates the question, reframes it toward what they do know, and owns what they don't. Most students have never been taught what to do in that moment.

Study strategies courses and workshops

If your center runs a first-year seminar or study skills course, building a small board for one class is a solid in-class activity. These courses also see students juggling several classes at once, which makes them a good place to introduce the Meta-Board idea from step 6.

Tutor training

Have tutors build their own board for their content area, plus one for tutoring itself. A board with a learning scientist, a skeptical faculty member, and a struggling first-year student is a good way to role-play tough sessions. It also means tutors have used the method themselves before they teach it.

What to adapt for learning centers

Winkel's version assumes a motivated student with time to spare. Most of the students we see need a lighter, more careful version.

• Scale it down. The full weekly routine runs about two hours. That's a lot for a student who's already behind. Start with one move, like predicting before asking or a harsh review before submitting.
• Check the syllabus first. AI policies vary from course to course and even assignment to assignment. A rubric review might be fine in one class and an integrity violation in another. Make "What does your professor allow?" the first question every time.
• Use the tools your campus provides. Winkel's examples use Claude and ChatGPT, but students shouldn't need a paid subscription. Most campus-licensed tools have a way to save custom instructions (Gemini's Gems, for example) and many have a voice mode.
• Remind students the personas aren't real experts. An AI "economist" can sound confident and still be wrong. The board is a sparring partner, and claims still get checked against course materials.
• Offer space for voice practice. Rehearsing out loud needs a quiet, private spot. Not every student has one, so consider making study rooms available for it.
• Keep the human in it. The board doesn't replace the tutor or coach. Our role shifts toward helping students decide what to trust, what to push back on, and what to do next.

A four-week starter sequence

This condensed version works for a coaching caseload or a short workshop series. Pick the student's hardest class and stick with it for all four weeks.

1. Week 1: Build a mini board. Set up three advisors: one core expert, one from a related field, and one skeptic. The student brings a real question from the course, writes their own answer first, then asks the board.
2. Week 2: Break down the skills. Turn the syllabus into 5 to 10 verb-based skills. Do one short drill session together, and have the student do two more on their own and log the results.
3. Week 3: Review before submitting. On an upcoming assignment (if the course allows it), run a harsh review and a rubric check. The student decides which feedback to act on and explains why.
4. Week 4: Go solo. The student works on a course problem with no AI. Afterward, talk about which advisor voices showed up in their thinking and which didn't.

Here's a starter prompt students can adapt for week 1:

I'm a student in [course name]. Set up a board of three advisors I can consult all semester:

1. A core expert in [discipline] who teaches this material
2. An expert from a related field: [related field]
3. A skeptic who questions the mainstream view in [discipline]

For each advisor, write a short brief: who they are, what they believe, and what they're quick to doubt. Don't answer any questions yet.

When I bring a question, I'll share my own answer first. Each advisor should respond to my thinking, disagree where they would, and ask me questions. Never write my assignment for me.

The bottom line

Winkel's core claim is that AI plus real teaching produces a stronger student than either one alone. That's our lane. Learning centers have always been about helping students think better, not just get the assignment done.

You don't have to roll out a new program to try this. Pick one move, try it with one student this week, and see what happens.

Sources

• The Complete Playbook: How to Teach Your Students to Use AI Without Losing Them to It, Doan Winkel, How to Teach With AI
• Teach Your Students to Build a Board of Advisors for Every Class They're In (Part 1 of 7), Doan Winkel, How to Teach With AI

[[shortcode.authors.mark-woolwine]]