Step 01
Pick one class to test it on
For example: one teacher uses our AI detector to check essays for one class, or tries auto-marking on one assignment. You also pick one person (like a department head) who will make the final decision.
Institutional assessment pilot
Run a governed assessment pilot before a platform-wide decision.
Start with one bounded workflow, an agreed baseline, named reviewers, and the evidence your institution needs to make a responsible next decision.
How the pilot works
Try it in one class first.
A pilot means testing AI Sentinels in one real class before your whole school commits — for example, checking one teacher's essays for AI writing, or auto-marking one assignment. You pick the class, we help you run it, and at the end you decide if it's worth doing school-wide. Here's how that works, step by step.
Step 02
Write down how it works today
Before switching anything on: how long does grading normally take? What problems come up now, like suspected AI-written essays? This becomes your "before" picture.
Step 03
Turn it on for that class
We set it up just for the class you picked — who can log in, what it checks for, and how student data stays private and secure.
Step 04
Run it for a few weeks
Your teacher uses the tool as normal, with our support if anything comes up. A real teacher always reviews the results — the tool never decides on its own.
Step 05
Compare and decide
Look at the "before" picture next to what actually happened. Then choose: stop, tweak the setup, try it in more classes, or roll it out school-wide.
Evidence frame
Measure change against the process you have now.
Measures are agreed during qualification. The pilot does not promise a predetermined improvement; it creates a fair way to discover whether the workflow is valuable and workable in your context.
See the evidence publication standard →Educator time
Compare preparation, review, and follow-up effort against the agreed baseline.
Consistency
Review how reliably the agreed criteria are applied across the pilot sample.
Review exceptions
Count and examine cases that require academic judgement or process changes.
User confidence
Gather structured feedback from the staff operating and reviewing the workflow.
Implementation readiness
Record integration, governance, training, and support work needed to scale.
Operating guardrails
Governance is part of the pilot, not paperwork after it.
Human decision ownership
Academic and institutional owners retain responsibility for consequential decisions.
Documented criteria
The pilot begins with an agreed workflow, review points, and evidence questions.
Privacy by scope
Only the information required for the agreed test should enter the pilot workflow.
Permission before publication
Institution names, quotes, and outcomes stay private unless written publication permission is granted.
Pilot application
Is there a bounded workflow and a decision owner?
Use the form to outline the evaluation. The first review is about fit and readiness, not a commitment to buy.
Before applying, identify:
- • The workflow or course in scope
- • The academic and operational owners
- • The current problem or decision question
- • The likely timing and review requirements
Pilot questions
- What is an AI Sentinels institutional pilot?
- A bounded test of the platform in one real class before a platform-wide decision. A useful pilot starts with one course, assessment, or integrity workflow, a named academic owner and operating team, a documented current-state baseline, agreed evidence and decision criteria, and a final review offering stop, adapt, extend, or procure options.
- How does a pilot run, step by step?
- Five steps. Pick one class to test on and one person who will make the final decision. Write down how the workflow runs today, including grading time and current problems. Turn the platform on for that class, covering access, scope, and how student data stays private. Run it for a few weeks with support. Then compare against the baseline and decide.
- What does an institutional pilot measure?
- Change against the process you have now, with measures agreed during qualification. The pilot does not promise a predetermined improvement; it creates a fair way to discover whether the workflow is valuable and workable in your context. The outcome review offers four options: stop, adapt the setup, extend to more classes, or procure.
- Does the platform make decisions about students during a pilot?
- No. A real teacher always reviews the results, and the tool never decides on its own. Each pilot also names an academic owner, such as a department head, who makes the final decision. Detection output is a review signal that supports the institution's own process, responsibilities, and appeal or review safeguards.