Examples of AI: Benefits, Challenges and Risks
What the domain covers
This domain tests your ability to recognise real AI applications and to separate benefits, implementation challenges and potential risks. It also checks whether you can identify ethical, legal and social issues and propose proportionate mitigations.
Core concepts
- Benefits: speed and scale, 24/7 consistency, pattern detection in large data, personalisation, freeing humans from dangerous or repetitive work, and better forecasting.
- Challenges: data quality and availability, integration with legacy systems, skills and change management, cost, model maintenance, explainability, and measuring return on investment.
- Risks: bias and discrimination, privacy intrusion, security and adversarial attacks, unsafe errors, opacity and unclear accountability, automation bias, job displacement, concentration of power, and environmental cost.
- Distinguish the terms: a challenge is a difficulty in adopting or operating AI; a risk is a potential harm if the system fails or is misused. Benefits and risks can coexist in the same application.
Worked example
A recruitment tool ranks CVs. Benefits include faster shortlisting and consistency. A challenge is that historical hiring data under-represents some groups and the model is hard to explain. A risk is discriminatory rejection, legal challenge and reputational damage. Reasoning: the harm arises from biased training data plus automated decision-making at scale; mitigation includes diverse data, bias testing, human review, audit trails and candidate feedback. If bias causes actual discriminatory outcomes, call it a risk, not merely a challenge.
Traps that cost marks
- Not every automation is AI; rule-based systems may not learn from data.
- High accuracy does not mean fair, safe or lawful.
- Explainability and accuracy can trade off; a black-box model may be accurate but hard to challenge.
- Do not treat job replacement as only a benefit (productivity) or only a risk (displacement); context matters.
- Generative AI can hallucinate; that is a risk, not a benefit.