Artificial Intelligence in Service Industry

What this session covered

This webinar, convened by the institute on the subject of artificial intelligence in the service industry, took the form of a multi-speaker session rather than a single interview. Following opening remarks from the institute's president on the organisation and on the broad promise and risks of artificial intelligence, a moderator introduced three invited speakers who presented in sequence, after which questions were taken from the audience. Participation was noted from a range of countries, and the session was described as running live on video and social-media platforms.

The first presentation offered an introductory overview of data science, machine learning and artificial intelligence, distinguishing structured from unstructured data and setting out two broad approaches: unsupervised anomaly detection, used to surface activity that departs from the normal range, and supervised predictive analytics, used to learn from known cases and anticipate future ones. Worked illustrations were drawn from investigations and security, including detection of payroll and transaction anomalies, document clustering and log-file analysis, together with commonly encountered limitations such as class imbalance, the difficulty of separating legitimate from unwanted material, and questions of privacy, security and bias. The second speaker turned to applications in intelligence, security and risk consulting, presenting several anonymised case studies covering human-resources screening, financial-fraud detection, tracing of tax absconders and threat assessment for public venues; this speaker argued that artificial intelligence augments but does not replace human judgement and drew attention to its blind spots and limitations. The third speaker addressed computer vision, describing how cameras are trained to interpret images, the labelling of data for self-driving applications, and uses in agriculture, retail, healthcare and security surveillance, including facial recognition and movement analysis.

The question-and-answer segment ranged over the future role of artificial intelligence in the service sector, data privacy and regulation, data-security risks and the trade in stolen personal data, the balance between proprietary and commercially available tools, and healthcare as an area of growth. Approaches raised in discussion included privacy-preserving and synthetic-data techniques, model explainability, and reliance on cloud platforms and security standards. In closing, one speaker's contention that human judgement remains essential was echoed in the concluding remarks, which cautioned against undue reliance on such systems and questioned, with reference to behavioural and movement analysis, how reliably a machine can interpret human conduct.

Key points raised

• The session was structured as three sequential expert presentations bracketed by an institute president's opening and closing remarks and an audience question-and-answer segment, not a debate.
• The discussion distinguished unsupervised anomaly detection, used to flag activity outside the normal range, from supervised predictive analytics, used to learn from known cases and anticipate future ones.
• Applications described spanned investigations and security, financial-fraud detection, human-resources and background screening, tracing of individuals, venue threat assessment, and computer-vision uses across several industries.
• A recurring theme was that artificial intelligence augments rather than replaces human judgement, with attention drawn to its limitations, blind spots and dependence on the quality of underlying data.
• The Q&A addressed data privacy and regulation, data-security risks and the market for stolen data, the mix of proprietary and commercial tools, and healthcare as an area of anticipated growth.
• The closing remarks cautioned against over-reliance on such systems, questioning how reliably a machine can interpret human behaviour and movement.