Slate Reader AI — The Toggle 2,000 Colleges Can Flip

Slate Reader AI: The Toggle 2,000 Colleges Can Flip

When you read that "colleges are using AI to read essays," the examples are always a short, nameable list—Virginia Tech built its own model, UNC ran one for years, a handful of others. That framing is comforting because it's checkable: you can look up whether your school is on the list. But it misses how AI application-reading actually spreads. It isn't built school-by-school and announced. It's shipped vendor-by-vendor as a feature inside the admissions software thousands of schools already run—and switched on with no announcement, no roster, and no notice to you.

The clearest example is Slate.

The substrate: one CRM, most of American admissions

Slate, made by Technolutions, is the dominant admissions CRM in US higher ed. By the company's own count it powers more than 2,000 colleges and universities—including, per industry write-ups, the overwhelming majority of the top 100 national universities and liberal-arts colleges. (Treat that number as the vendor's marketing claim, not an audited figure; Technolutions doesn't publish a client list. But even discounted, the install base is enormous.) If your application lands at a mid-to-large US college, there's a very good chance it lands inside Slate.

In 2025, at its annual Slate Summit, Technolutions unveiled Reader AI—a feature inside the application Reader that summarizes the documents under review. In the company's own words, it will "summarize what a reviewer needs to know about application materials, like a letter of recommendation or college essay," read a transcript and list the relevant courses, and highlight passages on demand. Admins can set "directives" to tailor those AI pre-reads to specific goals.

Here's the detail that turns a product into a structural story.

Why this is the "silent scaling vector"

The DIY builders make headlines precisely because they're rare and discrete. Virginia Tech trained its own model and announced it; each homegrown system comes with its own press cycle, FAQ, and a name you can put on a roster of confirmed schools. That's the visible path.

The toggle is the invisible one. There's no press release when an office enables Reader AI, no public log, and—because it ships inside a CRM the school already licenses—no procurement record to find. That last point is the sharp contrast with the other way AI enters admissions: schools buying detection tools through public RFPs you can actually trace. Procurement is the vector you can FOIA. The CRM toggle is the one you can't.

UNC-Chapel Hill is the proof of the pattern: it ran a vendor AI tool on applicant essays for roughly five years before its student newspaper, not the university, made it public. The honest summary, echoing the chair of NACAC's admissions-practices committee, is that the prevalence of AI in review is genuinely hard to gauge from the outside—because the mechanism was designed to require nothing from the outside.

Slate summarizes. Some vendors go further.

A crucial distinction, and one worth keeping crisp: Slate's Reader AI is documented as summarization—it produces notes and pre-reads, and nothing in its knowledge base says it assigns a score or ranks applicants. Conflating "an AI summarized your file" with "an algorithm rejected you" would be wrong.

But the vendor layer doesn't stop at summaries. Element451, an AI-native admissions CRM that took a $175 million investment in late 2024 and reports 350+ institutions, ships an "Application Reader Agent" that does the genuine first read: it applies the institution's rubric, scores the components, and summarizes for a human—claiming 95%+ agreement with human reviewers and hundreds of staff-hours saved per cycle. Its agents are built with human approval in the loop, though the company's documentation also allows a "self-approval" mode that can act without it. Liaison, which runs the centralized application services behind much of health-professions admissions, added predictive holistic scoring in 2026. The toggle that summarizes today sits on the same shelf as the agent that scores.

The asymmetry that's actually the story

Stack the two sides of this up and the problem isn't that AI summarizes essays. It's the disclosure gap.

What applicants are told about their AI use is explicit and enforceable: Yale calls submitting "the substantive content or output of an artificial intelligence platform" application fraud that "may result in admission revocation or expulsion"; the Common App's fraud policy carries that language across its 1,000+ members; medical and other applications make you certify your writing is your own. The penalty is real.

What schools owe you about their AI use is, by contrast, voluntary. The AAMC's responsible-AI principles only recommend that institutions "provide notice and explanation" of how AI is used; NACAC's ethics guidance is similarly non-binding. So you can be expelled for using AI on an essay that an undisclosed AI then summarized for the committee. That's the same one-sided standard we've documented when schools ban student AI while using AI themselves—just moved down into the infrastructure.

The honest counterarguments

To be fair, and because the vendors deserve their steelman:

• Summarizing isn't deciding. Humans still read and decide; Virginia Tech is explicit that "AI will not make admissions decisions." A summary that helps an overwhelmed reader is not an algorithm issuing verdicts.
• CRMs always shaped review. Pre-reads, flags, fit scores, and yield models predate generative AI. The novelty here is the instrument, not the fact that software has long mediated how your file reaches a human.
• The need is real. Test-optional volume buried admissions offices; AI summarization is a response to a genuine workload, not a gratuitous one.
• No-retention, if true, matters. Technolutions says your data is "never stored, pooled… or used for AI training." That's a meaningful privacy claim (a vendor assertion, not an audit)—though no-retention is not the same as no-disclosure.
• Off-by-default is arguably responsible design. Shipping the feature as opt-in puts the choice—and the duty to disclose—where it belongs: with the school.

All fair. None of it changes the architecture: AI essay-reading no longer has to be built or announced. It's a switch inside software 2,000+ schools already own.

What this means for you

Practically, you can't look up whether your application was AI-summarized, and chasing that certainty isn't worth your energy. The summarizer rewards the same thing a good human reader does—a clear, specific, genuinely-yours essay—so the move is unchanged: write authentically, and don't write to an imagined algorithm. The systemic fix isn't ours to make, but it's a modest one: GradPilot's position is not "ban the AI"—summaries are fine and the workload is real—it's tell the applicant. If a school flips the toggle, disclosure should flip on with it. Until it does, the most accurate answer to "is AI reading my essay?" isn't a list of schools. It's: it's a setting, and you'll probably never be told.

---

Sources and notes

Technolutions Slate AI / Reader AI knowledge base and product pages (capabilities, OpenAI model, "Early Access" opt-in status, data-handling claims, "2,000+ institutions" — all vendor self-reported); Slate Summit 2025 coverage (Carnegie) for the June 2025 unveiling; Element451 / PSG materials on the $175M investment and the Application Reader Agent (vendor-reported accuracy figures); Liaison's WebAdMIT Holistic launch (2026); the Daily Tar Heel / EdScoop on UNC's years-long undisclosed use; Yale and Common App AI/fraud policies; AAMC "Principles for Responsible AI in Selection" and NACAC ethics guidance on voluntary notice. Install-base and accuracy figures are the vendors' own claims, attributed as such and not independently audited. Nothing here is legal advice.

Related Reading

• Which Colleges Use AI to Read Essays?
• Which Colleges Use AI to Read Your Essays? (Virginia Tech, UCLA, Penn State)
• What Universities Spend on AI Detection — the Procurement Trail
• Colleges Ban Student AI but Use AI to Read Your Essays
• Med Schools Ban Your AI — Then Use AI to Screen You