The AAI101 Generative AI Tools and Workflows ECA for the July 2026 semester at SUSS carries 70% of the course marks. It combines three applied scenarios, a property agency AI workflow, an airline satisfaction data analysis and an insurance Telegram chatbot, with research documentation and an individual reflection.

Question 1: Integrated AI Workflow Design (Urban Nest Realty)

Urban Nest Realty is a 35-employee Singapore property agency whose marketing, client services and training teams work separately. It wants one AI-powered system linking lead acquisition, enquiry handling and viewing scheduling, and agent training and buyer education, within a budget of $300 per month. Integrate concepts from Lessons 1 to 4.

Question 1a: Workflow Integration Map (10 marks)

• Visual workflow diagram from lead generation to enquiry handling and viewing to post-transaction follow-up, with decision points and feedback loops (4 marks).
• AI touchpoints: the tool used, what it does and what triggers it (3 marks).
• Data flow explanation between the three functions (3 marks).

Question 1b: AI Tool Selection and Integration (8 marks)

• At least one tool each for marketing content, enquiry and viewing automation, and training and client education, with name, function and monthly cost (4 marks).
• How the tools integrate and what data format they need (2 marks).
• Cost breakdown within $300 per month, including free-tier limits (2 marks).

Question 1c: Content and Process Consistency (7 marks)

• Brand voice guide of up to 150 words with three personality traits and “We say / We don’t say” examples (3 marks).
• A 5 to 7 item quality control checklist covering factual accuracy, brand alignment and property advertising rules (2 marks).
• At least two feedback mechanisms that feed back into the workflow (2 marks).

Question 2: Data Analysis and Visualisation Strategy (SkyMetric Analytics)

SkyMetric Analytics must explain to Pacifica Airlines why passenger satisfaction has plateaued, using the Airline Passenger Satisfaction dataset on Kaggle (train.csv, about 103,900 rows). Students must decide how to treat 0 ratings (service not applicable) and the roughly 310 missing Arrival Delay values. The audiences are the In-flight Experience, Customer Loyalty and Marketing, and Ground Operations teams.

Question 2a: Stakeholder Analysis (8 marks)

• 2 to 3 business questions per stakeholder answerable from the data (3 marks).
• The decisions each stakeholder will make (3 marks).
• The dataset columns needed for each (2 marks).

Question 2b: Visualisation Design and Justification (12 marks)

• At least one real visualisation per stakeholder in Looker Studio, Datawrapper or Flourish, with screenshots (6 marks).
• Chart type justification (3 marks) and audience adaptation (3 marks).

Question 2c: Insights and Recommendations (5 marks)

Top three data-based insights (3 marks) and 2 to 3 recommendations naming the action, the responsible stakeholder and the expected impact (2 marks).

Question 3: Customer Service Bot Development (SafeGuard Insurance)

SafeGuard Insurance handles over 45,000 interactions a month across motor, home and travel insurance. Queries split into claim status (35%), policy information (25%), document submission (20%), billing (12%) and complex disputed claims and legal or regulatory escalations (8%), which must go to a human claims specialist. The bot runs on Telegram, using Lesson 6 concepts.

Question 3a: Conversation Flow Design (10 marks)

• Welcome flow with greeting and menu (2 marks).
• Decision trees for at least three query types with branching logic (5 marks).
• Escalation protocol with automatic detection, trigger phrases such as “denied claim”, “appeal”, “lawyer” or “ombudsman”, and clear handoff communication (3 marks).

Question 3b: Response Templates and Brand Voice (10 marks)

• Two ChatGPT templates using the COOL framework, a claim status update and an escalation handoff, with variable placeholders (6 marks).
• Tone guidelines for distressed customers (2 marks).
• Safety disclaimers on binding decisions, emergency services and sensitive data collection (2 marks).

Question 3c: Integration and Data Management (5 marks)

System integration map linking Telegram, Make.com, ChatGPT and Google Sheets (2 marks), and a Google Sheets logging design of at least five columns that names the data-protection regulation applied and explains its requirements (3 marks).

Question 4: Research Process Documentation (12 marks)

• 4a: evaluate at least five sources consulted (3 marks) and describe 2 to 3 research challenges and how they were overcome (3 marks).
• 4b: at least two screenshots of AI prompts and outputs (3 marks) and one prompt iteration example showing the original prompt, the problem, the revised prompt and the improved output (3 marks).

Question 5: Individual Reflection (13 marks)

• 5a: 300 to 400 word reflection on how at least three lessons connected (4 marks) and the skills developed (4 marks).
• 5b: 150 to 200 word statement on workplace relevance (3 marks) and ethical considerations in implementing AI (2 marks).