• Frame an AI proposal as a specific, measurable business problem.
• Distinguish AI-suitable tasks from problems better addressed through process improvement, rules, or conventional automation.
• Compare predictive and generative AI in terms of use cases, evidence, evaluation, and risk.
• Assess feasibility across data, business value, technical fit, organisational readiness, workflow, and governance.
• Identify recurring failure patterns and redesign proposed uses with appropriate controls and human oversight.
• Apply a selection checklist, decision matrix, and Problem Selection Canvas to candidate projects.

Module 1: The AI adoption-value gap (adoption versus measurable value, why pilots stop before production, RAND’s management-rooted failure factors, and the technology-first trap)

Module 2: Which problems are suitable for AI? (five characteristics of an AI-suitable problem, when to use rules, six problem classes, predictive versus generative AI, four filtering questions, and the role of context)

Module 3: Feasibility assessment (the five-dimensional framework, data readiness, baselines and business impact, technical feasibility, organisational readiness, workflow integration, risk, compliance, and total cost)

Module 4: Case lessons and the selection guide (Air Canada, McDonald’s and Klarna, seven failure patterns, corrective design, the start-prepare-postpone-do not use AI matrix, the ten-question checklist, the Problem Selection Canvas, and a 30-day executive pre-assessment). Closing, three key messages, BSC AI Factory support, references and conclusion

Total duration: Approximately 69 minutes

Dr. Duygu ÇAKIR

Dr. Duygu Çakır is an academic in the Department of Software Engineering at Bahçeşehir University. She has extensive experience in academic research, software engineering education, and research and development, focusing on artificial intelligence, computer vision, automated machine learning, AI governance, and bias mitigation. She has contributed to and managed government-supported and industry-collaborative projects and has taught programming, data structures, computer graphics, data science, and responsible AI at undergraduate and graduate levels. Dr. Çakır holds a national patent on mobile eye tracking and received an IEEE Best Paper Award in 2021. Her current work focuses on the responsible and value-oriented adoption of AI in organisations. Alongside her academic work, she actively contributes to advancing AI governance practices for SMEs.

Organisations increasingly experiment with artificial intelligence, yet adoption does not automatically translate into measurable value. This course equips managers and decision-makers with a practical framework for determining whether a proposed business problem is genuinely suitable for AI and whether the organisation is ready to proceed. It treats AI primarily as a management and problem-selection decision, rather than as an algorithm or software-purchasing exercise. Participants examine the AI adoption-value gap; compare predictive AI, generative AI, rule-based software, conventional automation, and process improvement; and assess candidate initiatives across data readiness, business impact, technical feasibility, organisational preparedness, workflow integration, human oversight, and risk. Real-world cases illustrate recurring failure patterns such as technology-first projects, poorly defined outcomes, pilot limbo, weak baselines, and probabilistic tools placed in deterministic tasks. The course concludes with a ten-question selection checklist, a Problem Selection Canvas, a four-way decision matrix – start, prepare, postpone, or do not use AI – and a 30-day executive pre-assessment for making disciplined project decisions.

Participants should have a basic understanding of AI concepts and familiarity with organisational processes, project management, or managerial decision-making. Programming experience is not required.

None required.

An internet-connected computer or tablet, a current web browser, and audio-playback capability. No specialised software or programming environment is required.