Unit 2: Design Thinking with AI and its Relevance in Business

Design Thinking is a human-centred, iterative problem-solving approach; when fused with Artificial Intelligence (AI), it becomes a data-augmented method for scaling empathy, generating options, and testing solutions faster than manual methods allow.

Foundational properties this unit relies on:

• The five-stage cycle: Empathise → Define → Ideate → Prototype → Test — a non-linear loop where insights feed back to earlier stages.
• Human-centredness: Solutions begin with real user needs, not technology capability; AI amplifies but never replaces this orientation.
• AI as augmentation: Machine learning, natural language processing (NLP), and generative models act as tools within the cycle, not substitutes for the designer.
• Iteration under uncertainty: Both disciplines treat early failure as cheap learning; AI shortens the feedback loop.
• Divergence and convergence: Design Thinking alternates between generating many options (diverge) and narrowing to a choice (converge); AI accelerates both.

II. How AI Tools Support the Innovation Life Cycle

Mapping AI capabilities onto each stage of the design cycle.

AI contributes a distinct capability at every phase of the innovation life cycle rather than at a single point.

• Empathise (understanding users): Sentiment analysis and NLP process thousands of reviews, support tickets, and social posts to surface unmet needs at scale.
  • Concrete anchor: clustering 50,000 app-store reviews by topic reveals that "slow login" dominates complaints — a signal no single interview would give.
• Define (framing the problem): Clustering algorithms (e.g., k-means) group raw observations into personas and problem statements.
• Ideate (generating solutions): Generative AI (large language models, image generators) produces many concept variations on demand, widening the divergence phase.
• Prototype (building fast): AI code assistants and text-to-design tools convert a described concept into a wireframe or working mock-up in minutes.
• Test (validating): Predictive models and A/B analysis forecast which prototype will perform, and simulate user reactions before launch.

A. Speed and Scale Advantage

• Compression of cycle time: tasks that took weeks (survey coding, competitor scans) complete in hours.
• Volume of data handled: AI reads unstructured data — text, images, audio — that humans cannot review exhaustively.

B. Limitations Within the Cycle

• Garbage-in problem: biased or thin data yields misleading personas.
• Loss of nuance: AI detects patterns but misses cultural or emotional context a human observer catches.

III. Design Thinking with AI and its Impact for Innovation in Realms of Business & Life

How the fusion reshapes value creation across commercial and personal domains.

Combining the two disciplines shifts organisations from intuition-led to evidence-augmented innovation, affecting both enterprises and everyday life.

A. Impact in the Realm of Business

• Personalisation at scale: AI tailors offerings per customer while Design Thinking ensures the tailoring serves a genuine need — e.g., recommendation engines framed around user "jobs to be done".
• Faster time-to-market: rapid AI prototyping shortens the concept-to-launch gap, lowering the cost of experimentation.
• Data-driven empathy: decisions rest on observed behaviour (click paths, purchase logs) rather than assumption alone.
• Risk reduction: predictive testing filters weak ideas before capital is committed.

B. Impact in the Realm of Life

• Healthcare: AI-assisted diagnostics combined with patient-journey mapping design more humane care pathways.
• Education: adaptive learning platforms redesign lessons around each learner's pace.
• Accessibility: speech-to-text and image-recognition tools, designed empathetically, extend services to differently-abled users.
• Everyday decision-making: AI assistants applied through a human-centred lens reduce friction in tasks like budgeting, scheduling, and navigation.

IV. Illustrative Examples Across Domains

Applying the AI-augmented cycle to five representative sectors.

Each domain shows a stage of the cycle where AI plus Design Thinking creates concrete value.

A. Banking

• Empathy through data: transaction analysis identifies customers struggling with cash flow.
• Solution designed around the need: AI-driven fraud alerts and chatbot advisors resolve queries instantly.
  • Concrete anchor: a chatbot handling balance and transfer queries frees branch staff for complex advisory work, framed by mapping the customer's "moment of anxiety".

B. FMCG (Fast-Moving Consumer Goods)

• Trend detection: AI scans social media to spot emerging flavour or ingredient preferences.
• Rapid concept testing: generative tools mock up packaging variants tested virtually before printing.
  • Concrete anchor: predicting demand for a plant-based variant from search-trend data prevents overstock.

C. Retail

• Personalised journeys: recommendation engines suggest products based on browsing behaviour.
• Layout and inventory design: computer vision analyses in-store movement to redesign shelf placement around actual shopper paths.

D. Food & Beverage

• Menu and recipe innovation: AI suggests ingredient combinations from flavour-pairing datasets, then chefs prototype and taste-test.
• Demand forecasting: predictive models reduce food waste by matching preparation to forecast footfall.

E. Sales & Distribution

• Lead prioritisation: AI scores prospects by conversion likelihood so effort targets the highest-value customer segments.
• Route and supply optimisation: algorithms plan delivery routes, redesigned around the field agent's real constraints.
  • Concrete anchor: optimising distribution routes cuts fuel cost and delivery time while a service-blueprint keeps the driver's workload realistic.

V. Ethics and Human Values in AI Innovation

Guardrails that keep AI-augmented design trustworthy and humane.

Because AI decisions affect real people, embedding ethics into the design process is a core requirement, not an afterthought.

A. Core Ethical Principles

• Fairness: models must not discriminate on race, gender, or income — biased training data produces biased outcomes.
  • Concrete anchor: a loan-approval model trained on historically skewed data may reject qualified applicants from under-served groups.
• Transparency: users deserve to know when they interact with AI and how decisions are reached (explainability).
• Accountability: an identifiable human or organisation must own the consequences of an AI system.
• Privacy: personal data must be collected with consent and protected — aligned with regulations such as data-protection law.

B. Human Values in the Design Process

• Human-in-the-loop: critical decisions retain human oversight rather than full automation.
• Inclusivity: diverse teams and test users reduce blind spots in what AI is trained on.
• Beneficence over profit: the guiding test is whether the innovation genuinely improves the user's life.

C. Risks and Their Mitigation

1. Algorithmic bias: mitigated by auditing datasets and testing outputs across demographic groups.
2. Over-automation: mitigated by preserving human judgement at high-stakes points such as medical or financial decisions.

VI. Introduction to Prompt Engineering — Basics of AI

How to instruct generative AI effectively within the design workflow.

Prompt engineering is the practice of crafting inputs that guide a generative AI model toward useful, accurate outputs.

A. Basics of AI

• Artificial Intelligence: systems that perform tasks normally requiring human intelligence — learning, reasoning, perception.
• Machine Learning (ML): a subset where systems learn patterns from data rather than explicit rules.
• Generative AI: models that create new content (text, image, code), typically built on large language models (LLMs).
• Large Language Model: a model trained on vast text to predict the next token, enabling human-like responses.
• Token: the unit of text an LLM processes — roughly a word fragment; models have token limits per request.

B. What a Prompt Is

• Definition: the instruction or query given to an AI model to produce a desired response.
• Components of a strong prompt:
  • Role: who the AI should act as ("You are a market researcher").
  • Task: the specific instruction ("summarise these reviews in five themes").
  • Context: background the model needs ("the product is a budget smartphone").
  • Format: the required output shape ("as a bulleted list").

C. Prompting Techniques

• Zero-shot prompting: asking a task with no example — quick but less reliable for complex tasks.
• Few-shot prompting: supplying a few examples to steer the format and style.
• Chain-of-thought prompting: asking the model to reason step by step, improving accuracy on multi-step problems.

Weak:   "Give feedback on this idea."
Strong: "You are a UX researcher. Evaluate this feature idea for a
         banking app used by first-time savers. List three strengths
         and three risks as bullet points."

D. Role in the Design Cycle

• Empathise: prompt an LLM to synthesise interview transcripts into themes.
• Ideate: prompt for twenty distinct concept directions to widen divergence.
• Prototype: prompt for draft copy, code, or wireframe descriptions.
• Limitation — hallucination: LLMs can produce confident but false output, so every AI response must be verified against real data before use.