Impact Areas
Four domains where systems meet people directly.
Governance, education, agriculture and emerging technology differ in almost every practical respect — users, institutions, infrastructure, regulation. What they share is that the consequences of a poorly designed system fall on people who did not choose it and may not be able to work around it.
Each domain below is examined through the same four questions: what the real-world problem is, what technology could plausibly do about it, what the system must account for, and what outcomes a well-designed system might reasonably support.
Domain 01
Governance & Public Systems
Public systems carry legal authority and affect people who often have no alternative provider. Technology here must strengthen institutional accountability, not bypass it.
A The real-world problem
Citizens interact with public bodies through a patchwork of offices, forms, phone lines and portals. Requests are re-entered by hand, records are split across departments, and people cannot easily find out where their case stands. Inside the institution, staff spend significant time chasing information rather than making decisions, and the absence of a consistent record makes it difficult to show how an outcome was reached — or to learn from patterns across many cases.
B Potential technological intervention
Well-designed digital service delivery begins with the citizen-facing workflow: a clear, accessible way to submit a request with the information and evidence needed to act on it. Behind it, administrative case management provides structured queues, assignment, deadlines and a record of every action. Traceability links each outcome to the people and evidence involved. AI may assist with categorisation, summarisation or duplicate detection, but decisions remain with authorised officers.
C System considerations
- Respect established authority: software models the institution’s procedures; it does not redefine them.
- Role-based access aligned to legal remit and jurisdiction, enforced server-side.
- Data protection obligations for personal and sensitive information.
- Accessibility for users with disabilities, low digital literacy or limited connectivity.
- Immutable audit trails sufficient for internal review and external scrutiny.
- Information quality: validation at the point of entry to reduce downstream rework.
D Possible outcomes
- Citizens can see the status of their request without having to chase it.
- Officers receive complete, structured cases instead of fragmented correspondence.
- Supervisors can identify ageing cases and workload imbalances.
- Decision processes become reviewable after the fact.
Potential outcomes of well-designed systems; not reported results of field deployments.
Domain 02
Education & Learning
Learning depends on accurate information delivered in a form the learner can understand. Errors in educational content are absorbed as knowledge, which makes accuracy non-negotiable.
A The real-world problem
Learners face both scarcity and excess: high-quality, level-appropriate material may be hard to find, while the wider internet offers abundant content of uncertain accuracy. Generative AI has intensified this problem. It can produce fluent explanations instantly, but those explanations may contain errors that a learner is, by definition, poorly placed to detect. Educators, meanwhile, have limited visibility into where learners struggle and limited time to provide individual feedback.
B Potential technological intervention
Structured knowledge access organises approved material by topic, level and prerequisite, so learners can find what is relevant to them. Digital learning interfaces support workflows such as guided practice, revision and self-assessment. Feedback mechanisms help learners understand mistakes and help educators see patterns. AI-assisted learning can be valuable when it is grounded in reviewed sources, clearly labelled as AI-generated, and bounded to tasks where an error is recoverable.
C System considerations
- Accessibility across devices, abilities and reading levels.
- Language support, including content in learners’ first languages.
- Information accuracy: AI-generated material must be grounded in and traceable to reviewed sources.
- Educator oversight of what the system presents and how it explains.
- Learner data privacy, particularly for minors.
- Avoiding designs that reward engagement over understanding.
D Possible outcomes
- Learners locate appropriate material more quickly.
- Feedback becomes more immediate and specific.
- Educators gain visibility into common misconceptions.
- AI assistance supports learning without silently introducing errors.
Potential outcomes of well-designed systems; not reported results of field deployments.
Domain 03
Agriculture & Rural Innovation
Rural contexts expose every assumption a digital system makes — about bandwidth, devices, literacy, language and cost. Designs that work here tend to be robust everywhere.
A The real-world problem
Agricultural decisions — what to plant, when to treat a crop, where to sell — depend on information that is time-sensitive and highly local. That information is frequently unavailable at the moment it is needed, or available only in a language or format the user cannot easily use. Rural service delivery is further complicated by distance, intermittent connectivity and fragmented record-keeping, so coordination between producers, advisers and service providers depends heavily on informal channels.
B Potential technological intervention
Digital tools can improve agricultural information access by delivering relevant, locally validated guidance in regional languages and simple formats, including voice where text is a barrier. Operational records — inputs, activities, yields — can support better planning and access to services. Information coordination tools can connect producers with advisers and service providers. Decision support may combine local data with expert-reviewed rules, presenting recommendations with their basis and uncertainty.
C System considerations
- Variable connectivity: offline-first design with reliable synchronisation.
- Regional languages and non-text interaction such as voice and imagery.
- Affordability of devices, data and the service itself.
- Usability for people with varied literacy and digital experience.
- Local validation: guidance must be verified against local conditions by qualified experts.
- Data ownership and consent for records about land, produce and income.
D Possible outcomes
- Relevant guidance reaches users when decisions are being made.
- Records support planning and access to services.
- Coordination between producers and advisers becomes more dependable.
- Recommendations carry their basis, enabling informed judgement.
Potential outcomes of well-designed systems; not reported results of field deployments.
Domain 04
Future Technology
Exploring new technology responsibly means asking not only whether something can be built, but whether it should be, at what cost, and how its effects will be measured.
A The real-world problem
Emerging technologies — more capable intelligent applications, advanced interfaces, connected devices and new automation patterns — are frequently promoted ahead of evidence. Organisations face pressure to adopt them before feasibility, operating cost, safety implications or user acceptance are understood. The result can be expensive pilots that never transition to dependable service, or deployments that create new risks without delivering the expected value.
B Potential technological intervention
Responsible exploration treats new technology as a hypothesis to be tested against a defined problem. This involves building focused prototypes of intelligent applications or advanced interfaces, examining how connected systems and automation could change an operating model, and evaluating each against explicit criteria. The aim is to produce reliable evidence for a decision — adopt, adapt, defer or reject — rather than a demonstration designed to impress.
C System considerations
- Feasibility under real conditions rather than controlled demonstrations.
- Measurable evaluation criteria and baselines defined in advance.
- Safety, including failure modes and potential for misuse.
- Full cost of ownership, including operation and maintenance.
- Dependence on specific vendors, models or platforms.
- Real-world constraints of the people and institutions who would use the system.
D Possible outcomes
- Technology decisions grounded in evidence rather than momentum.
- Early identification of risks and hidden costs.
- Clear transition plans for concepts that merit further investment.
- Documented negative results that prevent repeated mistakes.
Potential outcomes of well-designed systems; not reported results of field deployments.
Cross-domain constraints at a glance
The same technical component — a form, a classifier, a notification — must be designed differently depending on the domain. This matrix summarises the constraints that most often change architectural decisions. It is a starting point for assessment, not a substitute for it.
| Factor | Governance | Education | Agriculture | Future Tech |
|---|---|---|---|---|
| Data sensitivity | High | High (minors) | Medium | Varies |
| Connectivity assumption | Mixed | Mixed | Low / intermittent | Context-specific |
| Language diversity | High | High | Very high | Varies |
| Cost of an AI error | High | High | High | Must be assessed |
| Institutional authority | Statutory | Educational bodies | Advisory services | Sponsoring organisation |
| Primary human safeguard | Authorised officer | Educator | Local expert | Evaluation gate |
Working on a problem in one of these domains?
We are interested in conversations with institutions, practitioners and researchers who understand these contexts first-hand. Domain knowledge is the most important input to a system that works.