FANO: From AI-Native Company to AI-Native School

The next frontier of AI is not the chatbot. It is the organisation itself.

For the last two years, most of the world has asked a tool question: what can one person do faster with AI? The deeper question is institutional: what becomes possible when an organisation is designed so AI agents participate in real work, inside real roles, through governed systems, with durable records and human accountability?

FANO exists to answer that question.

FANO stands for Foundations of AI-Native Organisation. It is a conceptual framework for moving from individual AI adoption to an AI workforce. It defines the foundations required before AI agents can safely operate as part of a school, company or institution: roles, authority, records, assurance and human governance.

It is also a challenge to the status quo. The existing model assumes that schools, companies and public institutions can keep their inherited structures and simply add AI tools around the edges. FANO argues the opposite. If AI changes who can do work, how knowledge is produced, how evidence is created and how decisions are supported, then the organisation itself has to change.

Tutorwise has already built a working form of this pattern inside its own operating model and product estate. The Build, Operate & Govern (BOG) Operating System is the practical implementation: AI seats, work routing, shared records, release gates, dashboards, approval controls and cross-agent coordination. Around it, Tutorwise has also built the marketplace layer, bookings, payments, referrals, VirtualSpace, Sage and Google Classroom integration. The AI-company dashboard exposes the operating evidence behind that claim. FANO is therefore not a speculative label attached to an empty idea. It emerges from a company that has already built both the internal AI workforce pattern and the education platform surfaces that make an AI-native school architecturally plausible.

FANO asks what the Tutorwise pattern means beyond one company. If an AI-native company can be built, an AI-native school becomes an architectural question: how would a secondary school operate if its structure were designed from the beginning to include an AI workforce supporting school-native roles: Headteacher, Deputy Headteacher, Assistant Headteacher, Director of Learning, Head of Department, SENCO, Designated Safeguarding Lead, Exams Officer and Data Manager?

This paper is a companion to The AI-Native Company, which makes the company-level argument. FANO generalises the pattern to organisations whose work depends on roles, trust, evidence and governance.

That question does not make the AI tutor less important. It makes the AI tutor more important. In this paper, an AI tutor or AI educator means the pupil-facing learning agent inside a governed school operating model, not a standalone chatbot. It supports the pupil directly, but its greater institutional value comes when that support improves the system around the pupil: the teacher's visibility, the department's evidence, the pastoral team's timing, the SENCO's record and the school's capacity to respond.

The question is urgent because the existing education model is being squeezed from both ends. On one side, schools face constrained budgets, rising SEND demand, teacher workload, attendance concerns, safeguarding complexity, assessment pressure and fragmented technology. On the other side, AI is challenging the assumptions under which modern education has been organised: what counts as knowledge, how work should be assessed, what teachers should teach, how learners should practise, and which tasks remain distinctively human.

FANO is not a call to abandon the school. It is a way to preserve the school by changing the institution around it.

AI adoption is not enough

Most organisations begin with AI adoption. They buy tools, issue guidance, train staff and encourage usage. A teacher drafts resources. A manager summarises meetings. A founder writes strategy. The gains are real, but the institution is unchanged.

AI adoption improves individual work. AI-native organisation redesigns the system in which work happens.

The difference is structural. In an AI-adopted organisation, the AI output usually lives beside the real workflow. The human asks, receives, edits and moves the result into the official system. The organisation remains human-only at the level of structure. AI is a helpful instrument, but not a participant in the institution.

In an AI-native organisation, AI agents receive work through the institution's own channels. They act in named roles. They use authorised sources. They update durable records. They escalate when authority is missing. Their output can be audited by humans and other AI agents. Their work is measured from systems of record, not from usage claims.

FANO names the transition from AI as tool to AI as governed workforce.

That transition is harder than procurement. It requires the organisation to decide what kind of institution it wants to become. The same AI tool can produce opposite outcomes depending on the operating model around it. In one school, it may become another ungoverned shortcut that weakens learning and increases risk. In another, it may become a governed assistant to teachers, leaders and pupils, improving preparation, feedback, evidence and intervention while keeping professional judgement intact.

The difference is not the model. The difference is the organisation.

The organisation becomes the platform

Software changed organisations once by giving them digital systems. AI changes them again by allowing those systems to act, provided the organisation is designed for that step.

A ticket is no longer only a description of work. It can assign work. A role is no longer only an HR title. It can constrain what an AI agent is allowed to do. A dashboard is no longer only a reporting surface. It can expose whether the institution is actually operating. A release gate is no longer only a checklist. It can halt unsafe change. A message bus is no longer only communication infrastructure. It can become the coordination layer of an AI workforce.

The architectural turn is clear: the organisation itself becomes the platform.

FANO begins from the idea that institutions need a new operating grammar. The old grammar assumed humans read documents, attended meetings, remembered decisions and filled gaps through judgement. That does not scale to AI agents. An AI workforce cannot safely infer the organisation from habit, corridor knowledge or private context. The organisation has to make itself legible to the actors it now expects to work inside it.

The more work an AI workforce can do, the more explicit the organisation must become. Roles must be clear. Authority must be clear. Records must be clear. Escalation must be clear. Assurance must be clear. Without that, capability outruns accountability.

The role is more durable than the model

The first foundation of FANO is role durability.

A model is not an organisation. A chat session is not a department. A prompt is not a governance structure.

Organisations are built around roles that persist beyond the individual holder. A school still has a Headteacher when the person changes. A department still has a Head of Maths when the timetable changes. A company still has a finance function when a tool changes.

AI-native organisation must follow the same principle. The durable unit is the role, not the worker, session, model or vendor.

An AI agent may occupy a role-support position, such as AI Head of Year support or AI Release Manager. The underlying model may change from one provider to another. The runtime may change. The agent may be unavailable for a period. But the role, remit, record and authority boundary remain.

This is the difference between an AI assistant and an AI workforce. A workforce requires continuity. It needs a durable organisational home.

FANO therefore avoids vendor-centred thinking. The institution should not be designed around one model provider. It should be designed around its own roles and responsibilities, then use models as replaceable execution capacity.

Capability is not authority

The second foundation is authority.

AI agents can already perform work that resembles professional output: drafting lessons, analysing data, reviewing code, generating policy text, summarising safeguarding notes, preparing parent messages and proposing interventions. Their capability will improve.

But capability is not authority.

An AI agent may be capable of drafting a safeguarding chronology. That does not authorise it to make a safeguarding decision. It may be capable of preparing a parent communication. That does not authorise it to send the message. It may be capable of identifying pupils who need intervention. That does not authorise it to decide the intervention outside the school's process.

This distinction will become one of the defining tests of mature AI-native organisation.

Weak AI adoption treats capability as permission. If the AI can produce the output, the organisation is tempted to use it. FANO rejects that. It treats authority as an institutional property. Authority belongs to roles, policies, governance and law. AI agents act inside that structure, not above it.

This does not slow the organisation down. It enables speed where speed is safe. When authority is explicit, an AI agent can complete routine work without repeated human clarification. It can also stop cleanly when work reaches a human-only boundary.

The result is not less autonomy. It is governed autonomy.

Records are how the organisation remembers

The third foundation is institutional memory.

Human organisations often rely on informal memory. People remember why a decision was made, where a document lives, who approved a change, which exception applied and what happened last time. That memory is powerful, but it is fragile.

AI makes the fragility visible.

An AI-native organisation cannot depend on private conversations as its memory. Work needs to land in durable systems. Tickets, registers, logs, assessment records, safeguarding systems, calendars, content systems, release histories, dashboards and message buses become the institutional memory of the AI workforce.

The record must answer practical questions. What was requested? Who or what claimed it? What evidence was used? What changed? What was blocked? What required human approval? What remains unresolved? What did the organisation learn?

Tutorwise's AI-company dashboard matters because it does not merely say that AI is being used. It reads operating evidence: code, documentation, commits, production releases, ticket movement, published content and safety signals. It turns an AI-native claim into something inspectable.

An AI-native school would need the same discipline in its own language. The question would not be "how many prompts did staff use?" It would be: which attendance issues moved, which intervention plans were prepared, which SEND evidence packs were updated, which parent communications were drafted and approved, which safeguarding escalations were supported, and which decisions remained human-owned?

This is the difference between activity and institutional progress.

Assurance is the architecture of trust

The fourth foundation is assurance.

AI-native organisation cannot rely on confidence. AI agents can be fluent and wrong at the same time. They can complete the visible task while missing the institutional consequence. They can update one surface and leave another stale. They can treat a field as proof when it is only a label. They can execute work on the wrong path and still produce a persuasive explanation.

Assurance must therefore be designed into the workflow.

A serious AI-native organisation does not ask only whether the AI answer was good. It asks whether the work passed through the right control, reached the right record, respected the right authority and created evidence that can be inspected.

For Tutorwise, that means release gates, migration checks, approval records, dashboard metrics and bus-based coordination. For a school, it would mean safeguarding escalation, SEND governance, assessment policy, data protection, parent communication review and leadership oversight.

The principle is the same in both cases. The guardrail must cover the actual path of work. A policy that AI agents never encounter is not assurance. A field that nobody reads is not authority. A dashboard that reports usage but not outcome is not proof. A review that happens after the work has already affected pupils, customers or production systems is too late.

FANO's claim is demanding: AI workforce design and assurance design are the same problem.

That claim matters because current education technology often treats governance as an administrative layer added after adoption. FANO reverses the order. Governance is part of the architecture from the beginning. The AI agent is not first granted a tool and then surrounded with policy. The AI agent is placed inside a role, and the role defines the work, authority, records, escalation and assurance before action begins.

Why schools are a defining use case

Schools are one of the defining applications of FANO because they expose the limits of tool-based thinking immediately.

Most discussion of AI in education starts with the learner-facing tool: the AI tutor, the homework helper, the marking assistant, the lesson generator. These tools may become important. Some already show promise in narrow tasks. But a school is not a classroom interaction. It is an institution.

Tutorwise already sits near that boundary: marketplace, bookings, referrals, trust scoring, VirtualSpace, Sage and Google Classroom integration. It is not yet an AI-native school, but it is adjacent infrastructure: marketplace, learning space, AI support, school-facing integration and operational evidence.

A secondary school is a complex operating system. It has a Headteacher or Principal, Deputy Headteachers, Assistant Headteachers, a Senior Leadership Team, Heads of Department, Heads of Year, Directors of Learning, a SENCO, a Designated Safeguarding Lead, pastoral teams, exams officers, data managers, form tutors and support staff. It runs assessment cycles, safeguarding processes, SEND reviews, attendance follow-up, behaviour systems, curriculum planning, parent communication and governance reporting.

The work of a school is instructional, organisational, pastoral, statutory and evidential.

The AI-native school is more radical than the AI tutor alone. The tutor helps a learner. The school strengthens the institution around every learner. The critical move is to connect the learner-level interaction back into the school system, so individual support becomes institutional intelligence rather than another isolated tool.

Imagine an AI workforce designed around school-native roles.

An AI Headteacher support seat could prepare governance evidence, track school improvement priorities and surface cross-school risks. An AI Deputy Headteacher support seat could coordinate operational follow-up and maintain escalation visibility. An AI Assistant Headteacher for Teaching and Learning support seat could review curriculum implementation and reduce planning friction. An AI Director of Learning support seat could connect assessment patterns to intervention planning. An AI Head of Year support seat could surface attendance, behaviour and pastoral patterns earlier. AI Head of Maths, English and Science support seats could help departments analyse misconceptions, review resources and maintain curriculum coherence.

An AI SENCO support seat could help maintain provision maps, prepare evidence packs and draft teacher guidance. An AI Designated Safeguarding Lead support seat could assist with chronology preparation, workflow prompts and escalation checks while leaving safeguarding authority firmly with the human DSL. An AI Exams Officer support seat could track deadlines, entries, access arrangements and invigilation planning. An AI Data Manager support seat could maintain reporting packs and connect assessment, attendance and intervention records.

The word support matters. FANO does not pretend that an AI agent becomes a Headteacher, SENCO or Designated Safeguarding Lead in the legal, professional or moral sense. It defines how AI agents can strengthen those roles inside the institution's own authority structure.

The opportunity is not an AI layer placed on top of school life. It is an AI workforce designed into the school's operating model.

The pressure on education is structural

The case for an AI-native school does not begin with fascination about AI. It begins with the condition of the education system.

In England, the fiscal picture is tight. The Institute for Fiscal Studies reports that total school spending per pupil fell by 10% in real terms between 2010-11 and 2019-20, with recent increases only bringing it back to around 2010 levels. The same analysis shows that SEND has absorbed over half of recent school funding increases and points to a possible £6bn gap between expected funding and SEND spending by 2028-29.

That matters because schools are not being asked to do less. They are being asked to support more complexity with limited capacity. SEND demand is rising. Attendance and behaviour require sustained follow-up. Assessment data has to be interpreted, not merely collected. Parent communication has to be timely and careful. Safeguarding records have to be accurate, sensitive and actionable. Curriculum quality has to be maintained while staff workload remains a central retention concern.

The Education Endowment Foundation identifies manageable workload, leadership and school climate among the highest-potential areas for improving teacher recruitment and retention. That is not a marginal management issue. It is a capacity question at the heart of school quality.

This is where FANO becomes more than a technology argument. If the constraint is institutional capacity, then a new institutional architecture is relevant. The AI-native school is not a way to replace teachers. It is a way to increase the operating capacity around teachers: the evidence preparation, pattern detection, follow-up, communication, drafting, checking and coordination that currently consumes too much of the school day.

The problem is not that schools lack software. Most schools already have too many systems. The problem is that those systems do not form a governed AI workforce.

The policy gap is institutional

Government cannot avoid the AI question by treating it as procurement, safety guidance or staff training alone.

Those matters are necessary. Schools need safe products, secure data handling, clear expectations, evidence of impact and protection against irresponsible use. DfE guidance on generative AI, support materials for education settings and data-protection advice all matter because the education system cannot adopt AI casually.

But safe use is not the same as institutional redesign.

The harder policy question is whether government is preparing schools to use AI inside the inherited model, or preparing the inherited model to change. If AI can support lesson preparation, assessment design, intervention planning, safeguarding chronology, SEND evidence, attendance follow-up, parent communication and school improvement, then AI is no longer only a classroom tool. It becomes part of school operating capacity.

That changes the policy unit. The question is not simply whether a teacher may use AI to draft a worksheet. The question is what a school operating model should look like when AI agents can support named roles, maintain records, escalate exceptions, prepare evidence and surface weak signals before a human leader has time to notice them manually.

This is where FANO challenges national strategy. An AI-native school cannot be produced by issuing tool guidance to a non-AI-native institution. It needs an operating model: role architecture, authority boundaries, record standards, assurance gates, workforce design and outcome measurement. Without that, the system may achieve compliant AI use while missing the deeper transformation.

The state therefore faces a strategic choice. It can regulate and advise at the edge of the old model, or it can help define the foundations of the next one.

AI challenges the curriculum itself

The second pressure is deeper than workload.

AI changes the meaning of learned work. If a pupil can generate a summary, essay plan, worked example, translation, quiz, image, code sample or revision sheet in seconds, the school has to rethink what learning is for, how understanding is demonstrated, and which tasks should be protected from automation.

The OECD has made this point directly: education systems must rethink what teachers teach and students learn as AI capabilities advance. This is not a small adjustment to homework policy. It is a challenge to the architecture of curriculum, assessment and credentialing.

For years, schools and universities have used written work as a proxy for knowledge, reasoning, effort and originality. AI weakens that proxy. The response cannot be only detection. Detection will be partial, contested and temporary. The deeper response is pedagogical and institutional: redesign tasks, assessment, feedback, classroom dialogue and evidence of learning for an age where AI assistance is normal.

FANO matters here because the answer cannot sit only with individual teachers. A school needs institutional coordination. Heads of Department need to decide which skills must be practised unaided and which can be extended with AI. Directors of Learning need to see whether assessment evidence still means what it used to mean. Senior leaders need policy that can be executed, not only announced. Governors need assurance that the school is adapting without lowering standards.

An AI-native school would not treat AI as a cheating problem alone. It would treat AI as a curriculum design problem, an assessment design problem and an institutional memory problem.

That is a larger and more demanding vision.

Qualifications need a new proof model

The same pressure reaches universities, awarding bodies and professional routes.

For generations, education has used produced work as evidence. Essays, reports, coursework, problem sets and projects have carried more than content. They have signalled effort, authorship, reasoning, independence, knowledge and readiness for the next stage.

AI does not make those forms worthless, but it makes their meaning unstable.

When a learner can generate a polished essay plan, a plausible argument, a worked solution or a literature summary in seconds, the institution can no longer assume that finished output proves the same things it once proved. The danger is not only cheating. The deeper danger is false confidence: a school, university or employer may believe it has evidence of understanding when it has evidence of production.

That forces a more demanding assessment settlement. Some work will need to be protected from AI so learners build fluency, memory, method and judgement. Some work should deliberately include AI so learners practise supervision, critique, verification and responsible extension. Some evidence will need to move into live settings: oral defence, supervised drafting, practical demonstration, classroom problem solving, structured viva, project walkthrough, peer explanation and iterative critique.

FANO frames this as an institutional problem because assessment is not only a teacher's technique. It is a trust system. Qualifications matter because other institutions rely on them: universities, employers, regulators, families and learners themselves. If AI changes the meaning of evidence, then the education system must redesign how evidence is created, stored, reviewed and trusted.

The future qualification should not merely certify that a learner produced an artefact. It should certify what the learner can understand, judge, explain, apply and defend in a world where production itself has become abundant.

The classroom ratio changes

The most immediate promise of an AI-native school is the classroom attention ratio.

The traditional classroom is constrained by a hard human limit: one teacher, often around thirty pupils, many levels of confidence, pace, need, language, attention and prior knowledge. The teacher can circulate, question, explain, check work and respond to misconceptions, but attention is finite. Even an excellent teacher cannot hold thirty simultaneous learning conversations.

A real AI tutor, or AI educator, in the classroom changes that constraint.

It does not replace the teacher. It changes what the teacher can see and support. Each pupil can have access to guided explanation, practice questions, hints, worked examples, retrieval prompts and immediate clarification while the teacher remains responsible for the lesson, the culture, the judgement and the intervention.

The effective ratio begins to shift from one teacher serving thirty pupils in sequence to one teacher leading the room while thirty pupils have supervised access to individual learning support. The teacher becomes more, not less, important: orchestrating the learning, deciding when AI support is appropriate, spotting over-reliance, protecting productive struggle, and using the AI-generated signals to intervene where human judgement matters most.

This is why the AI tutor is critical to FANO. Without it, the AI-native school risks becoming a back-office efficiency model. With it, the learning experience changes and the system improves at the same time. A pupil receives immediate support; the teacher receives better signals; the department receives better evidence; the school receives a clearer picture of need.

This is where the AI-native classroom differs from a laptop room. The AI tutor should not be an ungoverned companion whispering answers. It should be part of the school's learning architecture. It should know the lesson objective, the teacher's constraints, the pupil's context, the permitted level of help and the evidence that must return to the teacher.

Used badly, AI could weaken learning by making answers too easy. Used well, it could make the classroom more humane: fewer pupils waiting silently, fewer misconceptions hidden until a test, fewer teachers forced to choose between the pupil with a hand up and the pupil who has already given up.

The ratio does not disappear. The accountability does not move. But the distribution of attention changes, and that may be one of the most important educational shifts of the AI-native school.

The LMS becomes an operating surface

Google Classroom and other learning management systems remain relevant in an AI-native world, but not as the final architecture. They become transition layers: assignment, content and evidence systems that must become AI-readable, role-aware and action-aware.

The old LMS is a container. It holds classes, assignments, submissions, announcements, comments, files and grades. It is valuable because it gives school work a shared digital home. But in an AI-native school, a container is not enough.

The LMS has to become part of the institutional operating layer. Its future value is not that it stores worksheets. Its future value is that it becomes a controlled surface where AI agents can read context, prepare work, route evidence and respect school authority.

That means five changes.

First, the LMS must become more AI-readable. AI agents supporting teachers and leaders need structured access to assignments, deadlines, submissions, feedback, class materials and learning history, under school-controlled permissions.

Second, the LMS must become more role-aware. A Head of Department, Head of Year, SENCO, DSL, tutor, classroom teacher and pupil should not see the same AI surface or hold the same authority. The system must understand role, context and decision boundary.

Third, the LMS must become more action-aware. It should distinguish drafting from posting, recommendation from approval, feedback from grading, and support from decision. Google's Classroom app in Gemini already reflects this boundary: Gemini can use Classroom context to draft and summarise, but cannot directly enter grades, provide private feedback, delete, archive, directly post assignments or create rubrics. That limitation is not merely a product constraint. It is an architectural principle: capability is not authority.

Fourth, the LMS must become an evidence system. It should record what was generated, what was edited, what was approved, what was sent, what was assessed and what was escalated. Without that, AI assistance becomes invisible, and invisible assistance cannot be governed.

Fifth, the LMS must connect to the wider school operating model. Classroom data alone is not enough. Attendance, behaviour, SEND provision, safeguarding chronology, assessment cycles, interventions and parent communication sit across multiple systems. The AI-native school needs an operating layer that can coordinate across them without collapsing all authority into one tool.

This is why FANO is broader than an LMS strategy. Google Classroom, Microsoft Teams, Arbor, Bromcom, SIMS, Satchel One and other systems may all remain important. But in the AI-native world they cannot remain isolated containers. They must become interoperable records and action surfaces inside a governed AI workforce.

The future LMS is not only a place where work is assigned. It is a place where institutional memory, authority and AI support meet.

The impact question

How large could the impact be?

FANO as a named framework has not yet been proven across schools. That should be stated plainly. It is a conceptual framework emerging from a working AI-native company pattern. But the pressure it addresses is already visible.

Schools are constrained by workload, coordination, evidence, safeguarding, SEND, attendance, assessment and leadership capacity. These are not marginal issues. EEF's evidence summary on teacher recruitment and retention identifies manageable workload, leadership and school climate among the highest-potential areas for improving recruitment and retention. The Department for Education's 2024 work on generative AI in education also frames AI as both an opportunity and a risk, with adoption dependent on evidence, safety and implementation.

The early evidence is already enough to move the debate. EEF's ChatGPT lesson-preparation trial found a 31% reduction in planning time for participating teachers using ChatGPT with guidance, with lesson quality not appearing to fall in the sampled review. EEF's wider EdTech research agenda is more cautious: evidence remains mixed, especially for disadvantaged pupils, and careful implementation matters. That is exactly the opening for FANO. Saving time on lesson preparation is the first-order effect. The second-order question is larger: what happens when the school redesigns the operating model so AI support reaches the right role, at the right moment, with the right evidence and the right guardrail?

That is where the impact becomes institutional.

The AI-native school could reduce administrative drag, make weak signals visible earlier, support intervention planning, help leaders distinguish policy from execution, improve SEND and safeguarding evidence, and give Heads of Department more capacity for teaching quality.

None of this removes the human centre of education. It protects it. FANO reduces the institutional friction that prevents humans from applying care, judgement and responsibility where they matter most.

The ambition is justified. If AI-native organisation works in schools, it becomes social infrastructure: a way to increase institutional capacity without reducing human responsibility.

There is also a risk in not moving. If schools do not develop an AI-native operating model, AI will still enter the system. It will enter through pupils, parents, teachers, vendors, browsers, phones, homework, search, tutoring apps and workplace expectations. The choice is not AI or no AI. The choice is governed institutional adoption or unmanaged drift.

Unmanaged drift is the weak path. It produces hidden AI use, inconsistent classroom norms, uneven access, confused assessment, unmanaged data exposure, variable quality and more pressure on already stretched staff. It also risks widening inequality: confident families and well-resourced schools adapt first, while others inherit the risk without the infrastructure.

FANO is the stronger path. It says the institution should absorb AI deliberately, through roles, records, authority and assurance.

Measurement without theatre

FANO requires measurement, but not vanity measurement. Counting prompts, logins or generated documents may show adoption, but they do not show whether the institution improved.

AI-native measurement should come from systems of record. In a company, that may mean releases, commits, documents, tickets, published content, safety streaks and operating cost. In a school, it may mean intervention completion, attendance follow-up, assessment-cycle readiness, SEND evidence quality, parent communication turnaround, policy execution, workload reduction and safeguarding assurance.

The metric should describe institutional movement. Did the work reach the right role? Was the evidence complete? Was the decision within authority? Was the human approval recorded? Did the outcome change the system of record?

FANO rejects theatre. The organisation cannot simply say it is AI-native. It has to show the institutional record.

The standard FANO sets

FANO sets a high bar because the stakes are high.

An organisation is not AI-native because it has AI licences, private staff usage, a chatbot, a policy document or a public statement about innovation.

It should be able to show five conditions.

First, AI agents operate in named organisational roles or role-support positions.

Second, work reaches those AI agents through durable systems, not informal prompts alone.

Third, authority boundaries are explicit: what the AI agent may decide, what it may prepare, what it must escalate and what it must never do.

Fourth, outputs become institutional records that can be inspected, measured and challenged.

Fifth, assurance gates check the work before it affects pupils, customers, finances, production systems, compliance or reputation.

These foundations do not reduce ambition. They make ambition credible.

The world does not need more AI theatre. It needs organisations that can absorb AI capability without losing responsibility. It needs schools where AI strengthens teachers and leaders rather than adding another ungoverned surface. It needs companies where AI agents build and operate under evidence, not assertion. It needs institutions that move faster because they have become clearer.

That is FANO's promise.

The AI-native future will not be defined only by better models. It will be defined by better organisations.

For education, that is the essential claim. Better models will change what is possible. Better organisations will decide whether that possibility strengthens or weakens learning.

Five Questions FANO Forces Education To Answer

Is DfE policy managing safe AI use inside yesterday's school model?

Current policy rightly emphasises safety, infrastructure, data protection, product expectations, training and effective use. Those are necessary, but they are not sufficient. The harder question for government is whether it is making AI safe for the existing model, or asking whether the existing model is still adequate. If AI can support curriculum, assessment, SEND, safeguarding, attendance, parent communication and school improvement, then the policy unit is no longer only "technology use". It is school operating design. FANO asks whether national strategy is prepared to treat AI as institutional infrastructure rather than another EdTech category.

When production becomes abundant, what does a qualification certify?

The school problem does not stop at school. Universities, awarding bodies and professional routes have also relied on written production as evidence of knowledge, effort and originality. AI has broken the simplicity of that settlement. The hard question is not whether students used AI; that question will become less stable every year. The hard question is what evidence still deserves institutional trust: oral defence, supervised creation, live problem solving, practical application, research judgement, critique, apprenticeship-style observation or new forms of AI-assisted assessment. If higher education does not redesign proof of learning, it risks certifying output rather than understanding.

If explanation, practice and feedback are no longer scarce, what is the teacher's irreducible work?

The profession should not answer this defensively. AI can already explain, rephrase, scaffold, quiz and generate examples at a pace no human can match. The teacher's irreducible work is not to compete with that abundance. It is to decide what deserves attention, create the conditions for effort, read the human signals an agent may miss, protect productive struggle, build a learning culture, and judge whether understanding has become part of the pupil rather than the screen. FANO makes the teaching-method question unavoidable: which parts of the inherited classroom are educationally essential, and which parts are only artefacts of limited adult attention?

If school leaders adopt AI without redesigning authority, who is accountable when the institution acts?

The serious question is not whether AI can draft reports, prepare resources, analyse attendance, summarise safeguarding chronology or support SEND evidence. It can. The serious question is whether the school, trust, college or university has defined who authorised the work, where the record lands, what the agent may decide, what it must escalate, and which human role owns the final judgement. Institutions that add AI tools without redesigning authority create a dangerous ambiguity: the machine can act, the human remains accountable, and the organisation may not be able to prove what happened between the two.

If public institutions do not build the AI-native learning environment, who will?

AI will not wait for institutional readiness. It will enter through pupils, parents, tutors, homework apps, search, browsers, workplace expectations and consumer platforms. If schools, universities and government do not provide a governed AI-native environment, learners will still learn with AI, but under rules set elsewhere. The strategic question is severe: will public education shape AI-assisted learning through public purpose, professional judgement and institutional records, or will it become the late-stage administrator of a learning environment designed outside its authority?

---

This draft introduces FANO as the conceptual foundation behind Tutorwise's Build, Operate & Govern (BOG) Operating System and applies the model to the AI-native school as a defining use case.

Sources referenced: Department for Education, Generative artificial intelligence (AI) in education; Department for Education, Using AI in education settings: support materials; Department for Education, Data protection in schools: generative AI and data protection in schools; Education Endowment Foundation, ChatGPT in lesson preparation - Teacher Choices trial; Education Endowment Foundation, Research Agenda theme: EdTech; Education Endowment Foundation, Recruiting, retaining, and supporting teachers; Department for Education, Generative AI in education: educator and expert views; Institute for Fiscal Studies, Annual report on education spending in England: 2025-26; OECD, What should teachers teach and students learn in a future of powerful AI?; Google Workspace Updates, Classroom app in Gemini.