Lewis & Clark College logo

How to Evaluate AI Vendors in Higher Education: A 2027 Scorecard

Illustration of a campus committee at a table reviewing an AI vendor checklist
Written by
Yousuf Peshawarwala
Published on
November 18, 2026

TABLE OF CONTENTS

What AI committees should ask, what evidence to request and which red flags to watch for, with insight from Liz Young of Lewis & Clark College.

Book my team's walkthrough

Vendor scorecard

Weighted score: 3.6 of 5. Pricing needs follow-up.

The short answer: Score every AI vendor on six areas: data handling, agent controls, pricing you can predict, fit with your systems, accessibility and the effect on staff roles. Ask for evidence on each one, run a one-term pilot with success measures agreed up front, and give low-risk tools a fast review path so people have less reason to turn to shadow AI.

Why does AI procurement feel different from other software purchases?

AI changes how the work itself gets done, so a feature checklist misses most of what matters.

Liz Young of Lewis & Clark College sees AI as a change in how work gets done and in how people behave at work. Her team's governance questions start with who brings tools forward, how they're vetted, who approves pilots, and who gets a seat at the table for those decisions. In 2027 the stakes are higher, because agents can take actions in campus systems on their own.

Liz Young, Lewis & Clark College: AI as a process shift (video, 0:40). Liz Young on why AI changes how work gets done, and the governance questions that follow.

What should an AI evaluation scorecard include?

Six areas, each with a question to ask, the evidence to request and a red flag to watch for.

AI vendor scorecard: suggested weights

Data handling and security (weight 25%)

Agent controls (weight 20%)

Pricing you can predict (weight 15%)

Fit with your systems (weight 15%)

Accessibility (weight 15%)

Staff impact (weight 10%)

Weights are a starting point. Adjust them to your institution's risk profile. To see how CampusMind answers each area, visit our Trust Center.

What should you ask about student data and security?

Ask whether campus data trains models, where it's stored, and how the vendor prevents prompt injection and data leaks. Then ask for documents that prove it.

Liz Young, Lewis & Clark College: Building trust and security in AI (video, 1:10). Liz Young on the questions her team asks vendors about model training, storage and prompt injection.

Questions to send every vendor

How CampusMind answers: Campus data never trains models. CampusMind is FERPA compliant, with SOC 2, HECVAT, VPAT and TX-RAMP.

FERPA compliant · SOC 2 · HECVAT · VPAT · TX-RAMP

It runs in your institution's Azure tenant or is hosted by CampusMind.

Visit our Trust Center to review our security documents

What controls should an AI agent have?

Its own identity, a narrow scope, approval steps for high-impact actions, a spending cap and a complete log.

This update adds agents, since they can take actions in campus systems. Ask each vendor which of these levels an agent supports, who sets the level, and how you can see every action it took.

Four levels of agent permission

For a deeper look at ownership, identity and caps, read AI governance for CIOs.

How can you tell when AI is bolted on?

Look at what you're paying for. If the renewal price rises for AI features your team doesn't use, that's a sign it was added without a clear use on your campus.

Liz Young, Lewis & Clark College: AI bolted on, rising costs and budget cuts (video, 1:09). Liz Young describes a software provider that raised its price after adding AI her team wasn't using, and what that means when budgets are being cut.

Ask for the full cost in writing: usage or credit fees, connector and integration work, training, staff time to review AI output, and how the price can change at renewal. See CampusMind pricing

How should you run an AI pilot?

One term, one workflow, a small group, and success measures agreed before day one.

A one-term pilot

Success measures to agree up front

Want to try before a formal pilot? Campus Chat has a 2-week trial for 5 people with $10 in AI credit. Start a Campus Chat trial

How will AI change staff roles?

Job descriptions shift as repetitive steps get automated and new review duties appear. Plan for that during procurement.

Liz Young, Lewis & Clark College: AI's impact on staffing and job roles (video, 0:32). Liz Young on job descriptions changing as parts of the work get carved away and new duties come in, and why that makes vetting harder.

How one role can shift

Gets automated

Stays with staff

New duties

How do you keep a slow review from pushing people to shadow AI?

Run two lanes: a fast review for low-risk tools and a full review for anything that touches student records.

Two review lanes for a new AI tool

Fast lane: no student records, campus sign-in

Full review: student records, agents or campus-wide use

Publish both lanes and the expected timing, so people know what to expect from the approved path.

Frequently asked questions

Ask whether campus data is used to train models, where prompts and outputs are stored and for how long, who can access them, how the vendor supports FERPA, and how it defends against prompt injection and data leaks. Ask for documents, such as a current HECVAT and SOC 2 report, to back each answer.

Many institutions require one for any tool that handles institutional data. HECVAT 4 includes a dedicated AI section covering training data, transparency and human oversight, so ask for the current version.

One academic term is a practical length for most workflows. It covers a full cycle of real work, and gives you enough data to compare against a baseline you measured before the pilot started.

AI built for higher education connects to systems like the SIS, LMS and CRM, supports FERPA, and fits academic calendars, roles and approval paths. General tools usually need that work done by your team.

Plan for usage or credit fees, integration and connector work, training and change management, staff time to review AI output, and price increases at renewal. Ask each vendor to put these in writing.

For public institutions, web tools offered to students and the public fall under Title II, including tools provided through a vendor. The standard is WCAG 2.1 AA. The DOJ's April 2026 interim final rule set compliance dates of April 26, 2027 for public entities with a population of 50,000 or more, and April 26, 2028 for smaller entities and special district governments.

Evaluating AI with your committee this year?

We'll walk your team through how CampusMind answers each part of the scorecard, with documents you can check. Book my team's walkthrough or visit our Trust Center.