Chapter 7. AI in Business, Public Services, and the Economy

7.1 Business functions

AI may support customer service, marketing, operations, finance, human resources, supply chains, software development, risk analysis, and management reporting. Examples include classifying customer requests, drafting product descriptions, forecasting demand, detecting unusual transactions, assisting quality inspection, and preparing first-pass summaries of documents.

The correct starting point is not 'Where can we add AI?' but 'Which problem is important enough to solve?' Identify the current process, its cost and quality, the people affected, and the likely cause of the problem. Then compare AI with process redesign, training, automation, and conventional software.

7.2 Education and research

Institutions can use AI for tutoring support, accessibility, administrative workflows, feedback drafts, and research assistance. Risks include inaccurate explanations, inequitable access, privacy concerns, over-reliance, and unclear authorship. The institution should define learning outcomes and make sure the technology supports them rather than undermines them.

7.3 Healthcare and social care

Potential uses include medical-image support, administrative documentation, patient communication, research, and workflow coordination. Clinical systems require domain-specific validation, privacy protections, human oversight, and compliance with applicable regulations. AI output should not be treated as a personal diagnosis or a substitute for qualified healthcare professionals.

7.4 Public administration

Public agencies may use AI to help classify correspondence, translate information, detect anomalies, summarise policy material, and improve access to services. The stakes can be high because public decisions affect rights and access to essential support. Public-sector use needs transparency, meaningful human review, accessible challenge mechanisms, and strong accountability. The OECD’s 2025 analysis stresses governance, quality data, skills, infrastructure, safeguards, and public engagement as enablers of trustworthy government AI (OECD, 2025).

7.5 Small and medium-sized enterprises

Small organisations can begin with lower-risk tasks such as drafting standard communications, summarising public information, preparing internal training, or categorising non-sensitive enquiries. They should avoid purchasing multiple overlapping tools before identifying a clear use case. A pilot should have an owner, a budget, a baseline, a success measure, and an exit plan.

AI can create dependency on a vendor. Before adopting a service, examine data export, service availability, pricing changes, integration, support, accessibility, and what happens if the provider changes or discontinues a feature.

7.6 Socio-economic effects

AI may improve productivity and create new services, but gains will not be distributed equally. Workers may experience task changes, new skill requirements, displacement pressures, and different forms of monitoring. Organisations and governments should invest in reskilling, social dialogue, inclusive access, and evidence-based transition plans.

AI also has physical and environmental costs, including data-centre energy and water demands, hardware supply chains, and electronic waste. A responsible assessment considers the full lifecycle rather than treating digital services as impact-free.