Buyer glossary
Definitions that keep product claims, safety, enterprise operations, and market evidence in the same frame.
How to use it
These are working definitions for enterprise research, not legal, professional, or educational advice. Follow the linked category and market guides for context.
The knowledge and practical judgement needed to understand what an AI system does, what data it uses, where it can fail, and how people should challenge or govern it.
Why it matters: An AI literacy plan turns a product rollout into a capability programme for educators, staff, students, families, and decision-makers.
AI that produces new text, images, audio, code, or other content from a prompt or other input. Its fluent output can still be inaccurate, biased, or incomplete.
Why it matters: Buyers should define approved uses, human review, source checking, privacy, copyright, and age-appropriate access before enabling it.
The people, authority, information, and workflow needed to review, correct, override, escalate, or stop an AI-supported result.
Why it matters: A human-in-the-loop label is not enough; the buyer needs time, training, access to evidence, and a real route to change the outcome.
Information connected to a learner, family, educator, course, assessment, interaction, or education record, including derived profiles and activity signals.
Why it matters: Student data needs a purpose, access boundary, retention rule, security control, correction path, and contract that prevents unauthorised reuse.
The practices and expectations that support honest, fair, attributable learning, assessment, research, and academic work.
Why it matters: AI detectors and monitoring signals should support a fair review process, not become automatic proof of misconduct.
A systematic difference in an AI system's errors, treatment, or outcomes that disadvantages people or groups because of data, design, context, or use.
Why it matters: Buyers should test subgroups, accessibility, language, culture, and consequences, then give affected people a way to challenge harm.
The collection and analysis of learning-related data to understand activity, progress, support needs, or educational processes.
Why it matters: Analytics can help staff act earlier, but proxy measures should not become hidden labels or unreviewable decisions about a student.
An education or vocational-training AI use that falls within a high-risk category under applicable law, such as some systems affecting access, admission, outcomes, level, or test behaviour.
Why it matters: Risk classification depends on the exact purpose and role, so buyers need a written use case and legal review rather than a vendor-wide label.
Designing and testing a service so people with disability can perceive, understand, navigate, contribute to, and challenge it from the start.
Why it matters: Accessibility is part of educational quality and risk control; a general conformance statement is not a substitute for user testing in the real workflow.
The traceable record of where data or evidence came from, how it was changed, what was included, and how it informed an output or decision.
Why it matters: Provenance makes generated content, research summaries, analytics, and assessment signals easier to check, correct, and defend.
Why read
The short answer: shared definitions help a buyer separate a product's stated purpose from evidence, safety, market, and procurement questions.
Next step: choose a category, open its buyer questions, and ask the vendor to define any term that still means something different in practice.
A practical next step
Enterprise AI Group describes a 6–8 week path for a defined business process, with governance, policy management, enterprise security, and Microsoft-tenant deployment considered from the start.
Enterprise AI Group describes a 6–8 week path for a defined workflow. Timing and cost depend on scope, users, integrations, security, governance, and support. These research pages are published by Enterprise AI Group. The implementation links describe optional Enterprise AI Group services; they are not product endorsements or a replacement for local diligence.
Explore workflow automationDo not include student records, child data, assessment answers, or other personal information in an enquiry.
Keep the useful part
Send the teaching, learning, student-support, research, or institutional workflow you are assessing. We will use it to shape the next practical buyer brief.
Useful detail: include the market, workflow, or category behind a education AI term or category.
Please do not send student records, child data, assessment answers, or other personal information.