Cover of Agentic AI in Higher Education by Fernando Santamaría
Presentation dossier · Not the full book
Review dossierAgentic AI in Higher Education

What this publication is for

A framework for educational judgement

Agentic systems do more than answer. They can plan, use tools, preserve context and act. In education, that changes where decisions are made and what learners may stop doing for themselves.

The book connects technical architecture with pedagogy, evaluation, safety, ethics and institutional governance. Its purpose is not to promote adoption. It is to make the conditions, limits and responsibilities of adoption inspectable.

Readers are given a conceptual map, documented cases, decision frameworks and instruments that can be adapted to a course, programme or institutional pilot.

Operational autonomy can be delegated.
Epistemic agency cannot.
281pages
20chapters
73figures
322bibliographic entries

Seven connected lenses

Concepts

Definitions, genealogy and the limits of simulated agency.

Architectures

Models, instructions, memory, tools and orchestration.

Pedagogy

Roles, autonomy, instructions and deliberate disengagement.

Applications

Tutoring, feedback, simulation, writing and research.

002
Chapter-level contentsParts I–III

The structure of the book

From concepts to pedagogical design

IConceptual and ontological foundations14–47

  1. 01From conversational AI to Agentic AI 15
  2. 02The ontological problem of the agent: genealogy and contemporary appropriation 25
  3. 03Theoretical bases of agent-mediated learning 37

IIArchitectures, patterns and coordination48–73

  1. 04Anatomy and architecture of agentic systems 49
  2. 05Design patterns and multi-agent coordination 62

IIIPedagogical design with agents74–105

  1. 06Pedagogical roles of educational agents 75
  2. 07The Pedagogically Restricted Autonomy Framework 85
  3. 08Designing pedagogical instructions 93
  4. 09Personalisation and adaptation: promises and cautions 99
003
Chapter-level contentsParts IV–V

The structure of the book

Applications, evidence and responsibility

IVApplications in higher education106–149

  1. 10Agents as intelligent tutors 107
  2. 11Documented and verified cases 114
  3. 12Applications by field 126
  4. 13Agents for academic research 139

VReliability, evaluation and governance150–204

  1. 14Teaching skills to work with Agentic AI 151
  2. 15Evaluation of educational agents: how to know if they work 161
  3. 16Security and systemic reliability 173
  4. 17Ethics, biases and educational responsibility 184
  5. 18Institutional governance of agentic systems 195
Evaluation does not end with usability. It must ask what students learn, what they can later do unaided, who benefits, who is excluded, and what happens when the system fails.
004
Chapter-level contentsParts VI–VII

The structure of the book

Recommendations and working apparatus

VIPerspectives and recommendations205–224

  1. 19Towards a reflective agentic pedagogy 206
  2. 20Future scenarios and lines of research 215

VIIAnnexes and apparatus225–280

A. Glossary · B. Autonomy decision matrix · C. Evaluation template for agentic tools · D. Questions for institutional reflection · E. Agent Learning Canvas · F. Application maturity framework · G. Annotated references. Also includes an index of figures.

What the progression makes possible

Name the system

Distinguish a conversational assistant from an agentic architecture.

Inspect the design

See where instructions, memory, tools and authority enter.

Restrict autonomy

Relate agent initiative to competence, risk and supervision.

Evaluate learning

Look beyond performance towards transfer and self-regulation.

Govern the lifecycle

Assign owners, records, stop conditions and incident procedures.

Preserve agency

Keep justification and epistemic responsibility with people.

005
Analytical contributionFrameworks

From principle to decision

Four ways to make an agent inspectable

Chapters 1 and 4

The agentic cycle

Perceive, plan, act and evaluate. Oversight covers the sequence of actions, not just the final response. Self-review does not guarantee correctness.

Chapter 7 and Annex B

Four design questions

What must the learner not delegate? What self-regulation support is needed? What are the domain risks? What human review is available?

Chapter 7 · Levels 0-4

Pedagogically Restricted Autonomy

PRA (APR in Spanish) is the author’s conceptual and design proposal for contextual autonomy, not a claim of demonstrated comparative effectiveness.

Annex F

Application maturity

Distinguishes documented, emerging, plausible and speculative applications. Technical feasibility is not evidence of educational impact.

The levels describe contextual fit, not quality or progress. More autonomy is not automatically better pedagogy.
006
Bonus track · IncludedAgentic AI Teaching Toolkit

Practical workbook

A toolkit for decisions, not just ideas

Cover of the Agentic AI Teaching Toolkit

A separate workbook helps educators document a pilot from readiness and pedagogical alignment to evaluation, incident response and deliberate disengagement.

24pages
117editable fields
10adaptable scenarios
0connections required
  • Readiness diagnosis
  • Opportunity filter
  • Agent profile
  • Pedagogical alignment
  • Activity canvas
  • PRA matrix
  • Delegation contract
  • Pedagogical instruction
  • Data and privacy check
  • Reference verification
  • Uncertainty register
  • Human-supervision plan
  • Evaluation rubric
  • Incident protocol
  • Disconnection plan
  • 30-day implementation pathway

Local by design

The standalone HTML can be completed in a browser. Responses remain in that browser unless the reader chooses to export them. The workbook can also be printed.

007
Editorial methodEvidence and audience

Application maturity · Annex F

Not every promise has the same support

The book does not treat every technical result, educational application and future proposal as if they had the same evidential status.

01

Documented

Public evidence establishes the scope of the claim.

02

Emerging

An observable implementation exists; evidence remains limited.

03

Plausible

Technically feasible; educational impact still needs to be tested.

04

Speculative

Depends on capabilities not yet reliably available. Categories must be reviewed as the evidence changes.

Primary readers

University educators · Academic and learning-design teams · Institutional leaders · Researchers and reviewers of educational technology.

Not a programming manual, a list of prompts, a product ranking or a promise that autonomy automatically improves learning.
008
Invitation to readIndependent review

For reviewers, educators and community editors

Request the complete edition if it fits your work

This dossier is intended to let you judge the scope and orientation of the publication before receiving the complete files.

If the subject is relevant to your research, teaching or editorial community, contact Fernando Santamaría through LinkedIn and ask for a complimentary full review copy.

Editorial status: this first-release English edition is undergoing a further language and copy-editing pass. Textual corrections will be issued over the coming weeks and may affect pagination. Please request the latest revision before citing or reviewing the book.

01 · REVIEW

Inspect this dossier

Check the thesis, complete chapter structure, frameworks and practical component.

02 · REQUEST

Send a brief message

Indicate the context in which you may read or discuss the book.

03 · READ

Form your own judgement

A critical or mixed assessment is as legitimate as a positive one.

Request a review copy: linkedin.com/in/fernandosantamaria

Public edition: payhip.com/b/UOPxr

Formats: 281-page book + 24-page teaching toolkit · PDF and offline HTML

A complimentary copy creates no obligation to publish a review or to offer a favourable assessment. Independent critical reading is welcome.