Beat the Game of Telephone Before AI Amplifies It

When I managed analyst teams, most of my job was not analysis. It was managing the game of telephone.

Someone asks for a number. What they need is a decision. Between those two things sits the work of extraction and distillation. That means figuring out what is actually being asked, how it plugs into the broader system, and what success looks like to the person who will act on it. That translation depends on knowing the stakeholders and the goals, and it is the part that goes missing first.

AI amplifies this. An LLM will produce fluent, confident, well-formatted output from a thin spec. The distortion arrives looking finished. Fluency hides the gap, so nobody catches it until it is in front of a student, or an executive.

Here's the tactic. Build pseudo-personas of the people on the other end, and save them as Skills. For us in higher ed, that means your prospect types. Think the rising senior, the adult learner, the transfer, the parent paying the bill. Encode what they care about and the questions they'd ask of your email before they act on it.

Then run the persona in both directions.

Before you write, the persona is the spec. What would make a working adult learner actually open this? What is she deciding, and what would earn her trust and her click? Draft that with real context: her motivations, her constraints, how she reads a message at 9pm after her kids are down. Feed that in before you generate a single line.

After you write, run it through the gauntlet. Pass the draft back through the persona and let it interrogate the copy. Does this speak to where I am, or is it a generic blast? Do I know the next step and why it's worth my time? Does it see my real barriers, cost, time, being first in my family to do this?

It's the same move I use for a CEO deck review, pointed at the person you're actually trying to enroll. It won't replace a real prospect, but it catches the telephone distortion while it's still cheap to fix.

One rule applies here. You sent it, so you're accountable for it. AI is a force multiplier. A 2% miss across 5x the volume is a lot of wrong emails to real students.

Steal the Skill — prospect-review-persona

Drop this into a .md file, swap in your prospect type, and run it against any draft email or campaign. Build one per persona once, reuse it all cycle.

---
name: prospect-review-persona
description: >-
  Pressure-test an email or campaign the way a [PROSPECT TYPE] would actually
  read it - hunting for generic copy that ignores their moment, next steps that
  aren't clear or worth it, tone that sounds like a mass blast, and barriers
  the message pretends don't exist. Triggers on "review this as a [prospect
  type]", "how would an adult learner read this", or any campaign being
  finalized for a send.
---

# Prospect Review Persona: [e.g., Adult Learner, 32, working full time, first-gen]

Read the draft as this specific person, not as a marketer. Find every place
it loses them before they'd ever click.

## Context (fill this in - this IS the spec)
- Who they are, life stage, what they're deciding right now
- Motivations: what they actually want out of this
- Constraints: cost, time, family, confidence, first-gen, work schedule
- How they read: on a phone, at night, skimming, skeptical of hype

## Workflow
1. Read the full draft - subject line, body, CTA, and any personalization.
2. Run it through the four lenses below. Don't invent problems.
3. For each real weakness, write the exact thought this person would have.
4. Output a ranked list plus a 3-line summary and one rewritten subject line.

## The four lenses
1. Relevance to my moment. Does this speak to where I am today, or could
  it have been sent to anyone? Generic is a fail.
2. Clear, worth-it next step. Do I know exactly what to do and why it pays
  off for me - not for the school?
3. Tone I trust. Does this sound like a human who gets my situation, or a
  funnel talking at me?
4. Barrier awareness. Does it see my real constraints, or pretend they
  don't exist?

## Output
List ranked Critical / Should-fix / Minor. Every note quotes the offending
line and carries a concrete fix. A criticism without a fix is noise. End with
the one question this prospect needs answered before they'll act.

## Tone
Honest and specific. No "this looks great."

Higher-ed news worth flagging

If you want more on where higher ed and AI are headed, these stories are worth a look.

1. The AI adoption gap is now a student-experience problem. 88% of students use AI in their learning and 77% of faculty use it in teaching (up 16 pts on 2025). But 57% of students say their assessments come with inadequate AI guidance, and only 29% think their instructors are equipped to guide them. The tools already won. The gap is guidance and governance, not adoption. (EdTech Innovation Hub, Stanford HAI 2026 AI Index)

Podcast pairing: Higher Ed Pulse Ep. 120 - The SUNY AI Mandate

2. The demographic cliff starts this year - lean teams are the norm now. 2026 kicks off a projected 15-year undergrad enrollment decline, and new international is down ~17% this fall. Same goals, smaller teams. This is the case for AI as a force multiplier, done right. (Cengage, Tyton Partners)

Podcast pairing: Higher Intelligence Ep. 10 - The AI Operating System

3. Real playbook, not theory: Kettering's GPT went from experiment to trusted teammate. A team turned a custom GPT into an embedded part of their marketing workflow, brand rules and all. Maps directly to the Skills tip above.

Podcast pairing: Higher Ed Pulse Ep. 114 - Higher Ed Pulse Ep. 114