An AI-powered assignment should not simply ask a student to use a chatbot to finish ordinary homework. It should give the student a task in which AI makes a deeper investigation possible. Students can use AI to compare explanations, generate hypotheses, explore unfamiliar examples, test a model or challenge an argument, while remaining responsible for the question, evidence, decisions and final explanation. The teacher’s job is to design the assignment so that using AI properly expands the learner’s intellectual work rather than removes it.
This changes the meaning of homework. Instead of measuring how many paragraphs a student can produce or how quickly a familiar exercise can be completed, an assignment can measure whether the student can investigate something uncertain. A learner might begin with a personal observation, use AI to identify possible explanations, check those explanations against reliable sources, collect evidence, revise the original idea and explain what was discovered. The finished work matters, but the thinking journey matters too.
- Start with a question that requires thought, not a prompt that invites AI to deliver a finished answer.
- Require the student to record an initial view before using AI.
- Use AI to generate alternatives, counterarguments, examples, simulations or testable hypotheses.
- Verify important claims using appropriate sources, data, experiments or teacher guidance.
- Assess decisions, revisions, evidence and explanations, not only presentation quality.
Choose your reading route
Read from the beginning, or start with the question that matters to you. The full teaching chapters remain open below.
Chapter index
Chapters 1–3 · Design the assignment
Chapters 4–6 · Explore worked examples
Chapters 7–9 · Assess and support
Chapters 10–12 · Introduce the approach
Chapters 13–18 · Investigate and review
CHAPTER 1 OF 18 · Design the assignment
1. Why ordinary homework needs a new design
Homework has several legitimate purposes. It can help students retrieve what they learned, practise a method, prepare for a lesson, develop reading habits and build independence. Not every assignment needs AI, and not every worksheet needs to become a research project. Short, deliberate practice remains valuable, particularly when learners are acquiring foundational skills that need fluency.
The difficulty appears when the assignment is treated as proof of understanding simply because a student submits a complete product. A language model can create an essay, a spreadsheet can calculate totals, a translation tool can produce a sentence and a coding assistant can generate a program. If the student submits these outputs without being able to explain or verify them, the teacher has little evidence of learning.
There are two poor responses. The first is to prohibit every AI tool and pretend that students will not encounter them outside school. The second is to allow AI to produce the work and assume that a polished answer means the student has learned. A better approach is to decide what the assignment should develop and then design a responsible role for AI where it adds value.
CHAPTER 2 OF 18 · Design the assignment
2. What makes an AI assignment a learning task
The useful distinction is between completing a task and designing for learning. These are not standardized categories called “AI-assisted” and “AI-powered”; both expressions can describe many kinds of work. A student might request a summary and submit it without understanding the source. Another might use a summary to identify disputed points, inspect the original passages and explain what the summary omitted. The same type of tool use can serve different purposes. What matters is the intellectual work the assignment requires and the evidence that the learner has done it.
A useful test for educators is this: which part of the intended learning does the student still have to perform? More difficulty is not automatically better. AI may appropriately remove formatting or transcription effort so a learner can concentrate on the subject. It becomes a problem when the tool performs the very reasoning the task was meant to develop and the teacher has no other evidence of that capability. Good design makes that boundary visible.
Part one: define the learning purpose
Write down the capability the task should develop. It may be the ability to interpret a graph, construct an argument, understand a scientific variable, evaluate a historical source or explain a mathematical relationship. If the purpose is unclear, AI use will also be unclear. State what students must know or be able to do before they begin, and distinguish the part that must be demonstrated independently from the part that can be supported by AI.
Part two: ask the learner to make an initial attempt
Before opening an AI system, the student writes an initial explanation, prediction, question or solution plan. This is not a trap. It is a record of the learner’s starting point. It allows both student and teacher to see whether the investigation changed the student’s understanding. The initial attempt can be short: a few sentences, a sketch, a worked example or a prediction.
Part three: use AI to widen the possibility space
The student asks AI for alternative explanations, counterexamples, possible variables, different solution methods or questions that might reveal a hidden assumption. The prompt should explain the learning objective and request reasons, uncertainty and testable suggestions. A learner should avoid asking only for a final answer. The teacher can provide age-appropriate prompt structures that teach students to ask for options rather than unquestioned conclusions.
Part four: check the information
Students verify claims according to the subject. They consult official documents where the task concerns rules or syllabuses, inspect original sources where historical evidence matters, calculate independently where arithmetic is involved, and perform safe experiments where physical claims can be tested. AI-generated references should be checked rather than copied. A source list must identify the actual sources used, not merely repeat citations invented by a chatbot.
Part five: test, revise or reject
The learner records which suggestions were useful, which were wrong, and what evidence changed the initial view. A strong assignment rewards justified disagreement with AI. If the model proposes an inaccurate explanation, identifying the error is part of the achievement. Students should be encouraged to preserve uncertainty when the available evidence is insufficient instead of forcing a confident answer.
Part six: create a result for a real audience
The final output should have a purpose. It might be a clear explanation for a younger pupil, a prototype, a data visualisation, a policy recommendation, a short documentary, an experiment report or an improved procedure. Students should state what they contributed, what AI contributed and what remains unresolved. This makes authorship transparent without treating all assistance as misconduct.
Part seven: explain the learning
A short oral discussion, live demonstration or written explanation checks whether the student can account for the result. A teacher can then change one condition and ask how the reasoning changes. This supplies evidence of transfer to that changed case, not proof that the student can transfer the idea everywhere. A later check is needed to examine retention. These checks should preserve agreed accessibility accommodations; independence from generative assistance need not mean withdrawing reading, communication or other authorized support.
English: from generated essay to argument investigation
A conventional assignment might ask students to write an essay about whether social media helps teenagers. An AI-powered version asks students to investigate a narrower question, such as whether recommendation feeds affect the range of news a teenager encounters. The student begins with a position, asks AI for competing arguments, searches for credible evidence, identifies unsupported claims and develops a conclusion that reflects the evidence. The final essay is still assessed for structure, grammar and vocabulary, but also for source judgement and the quality of the reasoning.
For a younger learner, the same principle can be scaled down. A student compares two descriptions of a familiar object, asks AI to explain a difficult word, checks the meaning in an age-appropriate dictionary and then writes an original sentence. The important work is deciding whether the explanation fits the context, not copying an impressive definition.
Mathematics: from answer production to model testing
A Mathematics assignment can ask students to model a real question: how much water might a household save by changing a routine, or how does travel time vary with distance and average speed? Students first write down their assumptions and construct a simple model. AI can suggest a different equation, identify possible variables or produce a table of hypothetical cases. The student checks calculations, tests whether the assumptions are realistic and explains where the model stops being useful.
Students still need to practise calculations without AI when the learning objective is computational fluency. They also need to understand why a method works. AI assistance is appropriate when the assignment is explicitly about comparing methods, exploring a pattern or modelling a new scenario. A tool should not conceal a gap in foundational knowledge.
Science: from memorising explanations to testing hypotheses
A Science task can begin with an observation: two plants placed in different conditions grow differently. Students propose possible explanations, ask AI to identify variables and suggest fair-test designs, then compare the suggestions with what they learned about plant growth. They decide which variables can be controlled, collect observations and interpret the results. AI can help organise a table or explain a concept, but it cannot replace the actual measurement or make an experiment fair by assertion.
The teacher must set safety boundaries. Students should never use AI as the authority for handling chemicals, electricity, heat, biological materials or equipment. The experiment must follow school procedures and teacher approval. Discovery is not an excuse to take unsafe risks.
History and humanities: from summary to source interpretation
An AI system can summarise a historical event in fluent prose, but that summary may hide disagreement among sources or omit the context needed to interpret them. An assignment can ask students to compare two primary or authoritative sources, ask AI to suggest possible differences in perspective, and then inspect the original documents. Students should identify who produced each source, for what purpose, under what conditions and what the source cannot tell us.
The student can then present a reasoned account that distinguishes evidence from interpretation. A good result may include a conclusion that remains provisional because the sources are incomplete. That is not a weak answer. It is an honest account of what can and cannot be established.
Computing: from generated code to responsible system design
Students can ask AI to explain a small program, suggest test cases or identify a possible bug. The assignment should require them to describe inputs and outputs, inspect the generated code, run tests in an authorised environment and explain any changes. A project could design a simple club inventory system using fictional data. Students must consider duplicate records, missing values, accessibility and how a user corrects an error.
The purpose is not to train students to copy code from a chatbot. It is to develop systems thinking: what problem is being solved, whose needs are represented, what could fail, and how do we know the result works? A program that runs once is not necessarily a reliable solution.
Art and design: from imitation to expanded exploration
AI can help students explore styles, generate design alternatives or explain visual principles, but the task should preserve the student’s own intention and decision-making. A student might investigate how public signs remain readable at a distance, generate alternative layouts, test legibility with classmates and produce a final design. The assessment can examine the brief, the reasoning behind choices and the response of actual users, not only the visual polish.
Teachers should also make authorship and copyright expectations clear. Students need to identify generated components where required, follow the conditions of approved tools and avoid presenting another person’s work as their own. Creative exploration should increase agency, not erase it.
CHAPTER 5 OF 18 · Explore worked examples
5. Worked mathematics task: when equal percentages do not cancel
This classroom example uses invented amounts and exact arithmetic. Its purpose is to understand percentage bases and multiplicative change. The teacher first asks: “A quantity increases by 20%, then decreases by 20%. Does it return to its original value?” Each learner gives a prediction and a short reason before seeing an AI explanation. A student who thinks the changes cancel has provided a useful starting point for instruction.
AI may then be used to suggest two representations: a numerical example and a multiplication-factor explanation. The student checks both. Starting from 100, a 20% increase gives 120. The subsequent decrease is 20% of 120, which is 24. The final quantity is therefore 96. In factors, the calculation is 100 × 1.20 × 0.80 = 96. The conclusion is a 4% decrease from the original amount. The second percentage operates on a different base, so subtracting the percentage labels alone is not valid.
The learner’s submission should mark the exact step that corrects the original misconception. If AI claims the quantity returns to 100, the student identifies the base error and supplies the calculation. If AI is correct immediately, the student still explains why the factors are multiplied. There is no need to manufacture a machine error or reward students merely for disagreeing. Agreement supported by a valid check is also judgement.
The teacher now changes the question: “A quantity decreases by 20%. What percentage increase restores it?” Starting from 100 gives 80 after the decrease. Restoring 20 from a base of 80 requires 20 ÷ 80 = 0.25, or 25%. The learner must explain the changed denominator. A separate follow-up can start from 240 to test whether the reasoning depends on memorizing the original numbers.
For learners ready for algebra, let the initial positive quantity be x and the fractional decrease be d, with 0 < d < 1. The decreased quantity is x(1 − d). If r is the fractional increase that restores x, then x(1 − d)(1 + r) = x. Dividing by the positive x(1 − d) gives r = d ÷ (1 − d). This is a fraction; multiply by 100 to express it as a percentage. When d = 0.20, r = 0.25. At d = 1 the reduced quantity is zero, so this restoring-percentage expression is undefined.
The immediate check asks students to explain the percentage base without a generated solution. A later lesson uses a fresh decrease, such as 25%, and asks for the restoring increase: 1/3, or 33⅓%. The assisted solution, independent explanation, delayed check and changed question answer different assessment questions. Together they reveal more than a correctly formatted calculation copied into a worksheet.
CHAPTER 6 OF 18 · Explore worked examples
6. A practical prompt pattern for learners
A useful prompt should ask AI to become a thinking partner rather than a substitute author. Students can adapt a structure such as this: “I am investigating for [learning objective]. Do not give me a finished assignment. Give me three possible explanations, identify the assumptions behind each, tell me what evidence would help test them, and point out what you are uncertain about. Ask me questions that help me make my own decision.”
The student should then inspect the response. Are the explanations genuinely different? Are the assumptions visible? Does the suggested evidence exist? Are there claims that need checking? What question should be asked next? A prompt is not a guarantee of a reliable answer. It is a way of structuring an interaction that keeps the learner active.
Assessment should be aligned with the learning purpose. If the task is to practise a basic calculation, the teacher may require an unaided solution. If it is to investigate a real-world question with AI, the teacher should assess the quality of the question, the suitability of the evidence, the student’s ability to evaluate suggestions, the reasoning behind revisions and the clarity of the final communication.
- Initial thinking: what did the learner believe or predict before using AI?
- Question quality: is the problem specific, meaningful and investigable?
- AI judgement: did the student use, adapt or reject suggestions with reasons?
- Evidence: can important claims be traced to reliable sources or tests?
- Subject understanding: can the learner explain the relevant concepts independently?
- Original contribution: what did the student decide, create, measure or discover?
- Transfer: can the student apply the learning to a changed situation?
The distinction matters because success during assistance can differ from subsequent understanding. A 2025 high-school mathematics experiment found that unrestricted GPT-4 assistance improved practice performance while reducing performance on a subsequent unaided test; a teaching-oriented design largely avoided that loss. That finding supports checking the learner after assistance. It does not validate the particular rubric below, which is an illustrative eduKate design for teachers to adapt.
CHAPTER 8 OF 18 · Assess and support
8. Age-appropriate use and gradual responsibility
A Primary 1 child does not need an open-ended AI research project. The teacher might use AI privately to prepare alternative examples, while the child makes observations, asks questions and explains a pattern. In Primary 3 to 6, learners can begin to compare simple explanations with teacher guidance, use approved tools and record what they checked. Secondary students can manage more complex research logs, cross-check sources and defend decisions. Post-secondary learners can undertake longer investigations with clearer methods and ethical review.
The key is gradual release of responsibility. Young learners require more scaffolding, shorter tasks and closer adult oversight. Older students can take responsibility for a larger part of the inquiry while still working within clear privacy, safety and academic-integrity rules. Access to an AI tool should never be used as a reason to abandon direct teaching or social support.
CHAPTER 9 OF 18 · Assess and support
9. Safeguards that make coexistence trustworthy
Students should not upload personal data, medical records, financial information, private school reports or information about other children into unapproved tools. They should understand that AI may invent facts, references, calculations or explanations. High-stakes claims require stronger verification than low-risk brainstorming. Teachers should identify approved services, explain age restrictions, understand data handling and provide a non-AI route where necessary.
There should also be clear rules for attribution. Students should disclose meaningful AI assistance in the manner required by the school. Teachers should distinguish permitted brainstorming from prohibited substitution. A learner should not be punished for following an unclear rule. Schools need examples of acceptable and unacceptable use, and students need opportunities to practise the difference.
Access should be planned with the assessment. The teacher can provide a common approved tool during class or give everyone the same small packet of AI outputs to evaluate. Students can then demonstrate the target skill without needing a paid subscription or a personal account. The packet option still teaches AI judgement, although it does not assess live interaction with a tool. State that difference openly so an accessible alternative measures what it claims to measure.
CHAPTER 10 OF 18 · Introduce the approach
10. A six-week introduction for a school or tuition group
A school does not need to redesign every assignment at once. It can begin with one subject and one carefully selected task. In week one, teachers identify the learning outcome and establish the student’s starting point. In week two, students learn how to ask AI for explanations and alternatives. In week three, they practise source checking and error detection. In week four, they investigate an open question. In week five, they create and present a result. In week six, teachers review the evidence of learning and revise the assignment design.
The review should ask whether students understood more, whether teachers could assess the work fairly, whether the task was accessible and whether the time required was reasonable. It should also ask what failed. A new model should be improved through evidence rather than adopted because it sounds modern.
Build the review into the plan before teaching begins. Preserve a short baseline response, use a fresh individual problem after the activity, and return to the key idea after a delay. Ask separately whether the AI helped complete the product and whether the student can now explain more. A class becoming faster without stronger understanding is a reason to change the task. A class becoming more accurate but spending excessive time on documentation may need a simpler record. The six weeks are a suggested sequence, not a tested dosage or a guarantee of improvement.
Parents can ask children what they were trying to find out, what AI suggested and which suggestions they accepted or rejected. They can encourage children to explain ideas in their own words, verify important facts and admit when they do not know. Parents do not need to supervise every prompt or become experts in every application. They can help maintain the distinction between getting an answer and understanding an answer.
A useful family conversation might be: “What did you think before you asked AI? What did it show you that you had not considered? How did you check it? What do you now understand better?” This rewards intellectual development rather than the quantity of homework produced.
CHAPTER 12 OF 18 · Introduce the approach
12. How this fits the eduKate learning ecosystem
The assignment model can connect existing eduKate learning resources. Vocabulary supports precise questioning and explanation. Mathematics and MathOS support modelling, calculation and verification. ScienceOS supports observation, controlled investigation and evidence. English and VocabularyOS support reading, argument and communication. The Super Intelligence hub explains models, tools, retrieval, agents and the need for human checks. The Civilisation articles help students see why their chosen question matters beyond the classroom.
The Project Work, Research and Collaborative Problem-Solving guide is a natural companion because it already connects project work with research and collaboration. The How Singapore’s Education System Fits into Civilisation: Computing, AI and Digital Trust article adds information systems and digital trust. The Assessments, Examinations and Meaningful Learning Progress guide provides the assessment connection. The next step is to make each learning topic offer two routes: a foundation route for mastery and a frontier route for AI-extended inquiry.
CHAPTER 13 OF 18 · Investigate and review
13. A worked assignment brief: improve a familiar public experience
Consider a class investigating how clearly a visitor can find a community facility. Students choose an approved public location, such as a library or community centre, and define a practical question: what information does a first-time visitor need before arriving? The project should not involve collecting personal information about strangers or recording identifiable people. Students can use public maps, official opening information, published accessibility details and their own observations in a supervised visit.
Before AI use, each student writes a short prediction about the most confusing part of the journey. The group then asks AI to list possible information needs and identify questions they may have overlooked. They check those suggestions against official sources and direct observations. They may discover that opening hours are easy to find but step-free access information is less prominent, or that a map uses a label unfamiliar to a new visitor. These are findings to verify, not assumptions to announce.
The group creates an improved one-page guide or prototype. It explains which source supports each practical detail, what was changed and how the group checked readability. The students present the work and answer a changed-scenario question: what would need to change for a visitor arriving outside normal hours, or for someone who cannot use stairs? The assignment connects language, geography, design, digital literacy and civic awareness while giving AI a defined role in widening the inquiry.
A simple learning record can make AI use transparent without asking students to preserve every prompt or produce a large administrative document. It can contain five short entries: my starting idea; the question I decided to investigate; the AI suggestions I considered; the evidence I used to check them; and what I changed in my final explanation. For older learners, a sixth entry can identify limitations and the next experiment they would conduct.
The record should focus on consequential decisions. A student does not need to copy pages of chatbot conversation if those exchanges did not affect the work. They should identify a moment when AI changed the direction of the inquiry, a moment when they rejected a suggestion, and a moment when evidence changed their own view. This helps teachers see intellectual development without rewarding prompt volume.
Students can also keep a small vocabulary list of concepts they had to learn in order to understand the problem. This connects AI-assisted inquiry back to foundational education. A learner who uses an unfamiliar term without understanding it should be encouraged to investigate and explain it. Precision of language supports precision of thought.
CHAPTER 15 OF 18 · Investigate and review
15. A rubric that rewards intellectual contribution
A practical rubric should show students what intellectual contribution looks like. The following example totals 100 marks. It is an illustrative teaching proposal, not an MOE, SEAB or research-validated scale. Adjust the weights to the subject and tell students the criteria before they start. A mathematics fluency task, for example, would need a different balance from an extended inquiry.
| Criterion | Example weight | Evidence that earns credit |
|---|---|---|
| Question and scope | 15 | A clear question, relevant constraints and an explanation of what would count as an answer |
| Subject reasoning | 30 | Correct concepts and a defensible method that the learner can explain |
| Verification | 20 | Traceable sources, calculations or observations that actually test important claims |
| Revision and judgement | 20 | Reasons for accepting, adapting or rejecting a suggestion, including honest limits |
| Communication and responsibility | 15 | An understandable result, appropriate acknowledgement and a clear account of the learner’s decisions |
For each criterion, teachers can use four simple descriptions: clear and well-supported; mostly sound with a small gap; partly explained with an important unresolved gap; or not yet demonstrated. The descriptions should refer to the subject. In the percentage task, “verification” means checking the second base and the restoring calculation. In a history task, it means inspecting relevant source evidence. Counting links, prompts or paragraphs would reward activity rather than the intended capability.
Do not let the total hide a serious misconception. A polished explanation with an incorrect percentage base still needs further mathematics teaching. Keep a brief separate record of whether the learner can explain the core idea, retain it later and use it in the changed task. A strong group product is evidence of group achievement; individual understanding should be checked individually. A student’s oral fluency, accent or confidence should not replace evidence of the subject knowledge being assessed.
Teachers can also identify the permitted role of AI. For example, AI may be allowed for brainstorming but not for generating the student’s core calculation. It may be allowed to provide language feedback but not to write the final personal reflection. It may be required for comparing alternative explanations, but students must verify the sources independently. These distinctions should appear in the assignment brief before work begins.
Rubrics should not award marks merely because students used AI. Nor should they penalise students for using an approved tool in the way the teacher requested. The assessment should reward the learning objective and the quality of the learner’s decisions. This is a more defensible basis for academic integrity than attempting to identify machine-written sentences by appearance alone.
Errors should be treated as learning opportunities when the task is safe and the teacher has prepared suitable boundaries. If AI gives an incorrect factual statement, the student can trace it to an original source. If it makes an arithmetic mistake, the student can calculate independently. If it suggests an experiment with uncontrolled variables, the student can explain why the method cannot answer the question. If it invents a reference, the student can record that the source could not be found and remove the claim.
The correction process must be proportionate. A low-stakes example can be discussed in class. A consequential health, safety or legal question requires an authoritative source and qualified human guidance rather than a student experiment. Students should be taught that confidence and fluent language are not substitutes for verification. They should also learn to stop using a tool when the quality of its output is inadequate.
Teachers can make this a routine by occasionally supplying a deliberately flawed AI explanation and asking students to find the errors. The exercise should be clearly identified as a classroom activity, not an attempt to trick students during a graded assessment. It develops attention, subject knowledge and the confidence to challenge a machine-generated answer.
CHAPTER 17 OF 18 · Investigate and review
17. Different assignments require different kinds of independence
There is no single rule for AI use across all subjects and tasks. When a student is learning basic spelling, arithmetic facts or a new grammar structure, the teacher may want evidence of unaided recall. When the task is to explore a complex issue, AI may be an appropriate research assistant. When students are learning to code, they may compare generated code with their own version, but they must understand and test the program. When the task involves personal reflection, AI should not invent the student’s experiences or feelings.
Teachers should label tasks by purpose: independent demonstration, AI-supported practice, AI-assisted inquiry or AI-aware critique. These labels make expectations easier for students and parents to understand. A task may combine categories, but it should explain which part is independent and which part permits assistance. The goal is to make assessment credible while preparing students for the world in which AI is present.
Parents can support the process by asking what the child was expected to learn and how the tool was used. They do not need to check every prompt. They can ask the child to explain a key idea without looking at the screen, describe one thing AI got wrong and show what evidence changed the final answer. This encourages responsibility without turning family learning into surveillance.
CHAPTER 18 OF 18 · Investigate and review
18. Conclusion: give students work worth thinking about
A good AI-powered assignment does not make the learner passive. It gives the learner a larger problem, more possibilities to consider and greater responsibility for the result. Students still need practice, explanations, examinations and independent demonstrations. But they also need to use powerful tools to investigate questions that are not already answered in the worksheet.
The educational test should not be whether AI was present. It should be whether the learner’s understanding, judgement and ability to investigate have grown. When assignments are designed around that test, AI becomes a means of expanding intelligence rather than replacing the work through which intelligence develops.
Do not ask students merely to finish more work with AI. Give them work that becomes more meaningful because they can think further with AI.
Continue the AI, Education and Civilisation series
Follow the argument from human learning into the institutions that shape shared life. Each article pairs its own questions with practical examples and evidence limits.
Learning and human capability
- AI and the Paradigm Shift in Education | How Superintelligence Can Boost Human Intelligence
- AI-Powered Assignments | How Students Can Use AI to Expand Their Intelligence
- From the Centre to the Edge | How Teachers Can Teach Discovery with AI
- The Future of Education and Civilisation | How Human-AI Coexistence Expands Discovery
Culture and shared life
- AI and the Paradigm Shift in Culture | Human Creativity, Cultural Memory and New Frontiers
- AI and the Paradigm Shift in Society | From Passive Systems to Collective Intelligence
- AI and the Paradigm Shift in News | From Information Distribution to Verified Understanding
- AI and the Paradigm Shift in War | From Faster Conflict to Prevention, Accountability and Human Security
Knowledge and public institutions
- AI and the Paradigm Shift in Museums | From Preserving Objects to Expanding Human Discovery
- AI and the Paradigm Shift in History | From Recording the Past to Discovering Its Hidden Connections
- AI and the Paradigm Shift in Science | From Applying Knowledge to Discovering the Unknown
- AI and the Paradigm Shift in Government | From Administrative Automation to Public Intelligence
Economics, organisations and work
- AI and the Paradigm Shift in Economics | From Faster Output to Shared Capability and Opportunity
- AI and the Paradigm Shift in Business | From Faster Tasks to Organisations That Learn
- AI and the Paradigm Shift in Finance | From Faster Answers to Accountable Financial Judgement
- AI and the Paradigm Shift in Work | How People Build Expertise Alongside AI