The Empathic Algorithm — Leveraging AI for Next-Level CX

Robert Dew · First edition 2023

When she asked him to dance AI was shy — cover illustration — When she asked him to dance AI was shy.

That opening line sets the tone for everything that follows. This book explores not just the application of AI to automation but delves into the softer aspect of 'empathy' — often overlooked yet vital for impactful customer experience.

Note: This book was written in 2023 when ChatGPT 4 was state of the art. The AI landscape has moved considerably since — which, in many ways, makes the book's core argument about empathy and human-centred design even more urgent.

As the world evolves beyond the profound impacts of the COVID-19 pandemic, the transformative power of technology has come to the forefront. Whether you're an entrepreneur adapting to shifts in customer behaviour or a seasoned executive reimagining business strategy, this book is a tool to leverage AI for a deeper connection with your customers.

This is a comprehensive guide for businesses navigating the new normal, where digitization has ceased to be an option and become a necessity. It offers insights into the mind of the customer and the evolution of CX in the AI era — from leveraging AI for cost reduction to creating revenue growth.

Resources

Download the Book (PDF) PDF — coming soon
Watch the Presentation VIDEO
Listen to the Audio AUDIO

Chapters

Introduction to a Fast Changing World 001

The opening chapter sets up the book's core wager: the gap between human-delivered and machine-delivered customer experience has historically been massive — humans warm and expensive, automation cheap and cold — and large language models are about to collapse it. The prior decade is framed as a wholesale corporate retreat from human contact in pursuit of cost savings, with chatbots positioned as the impoverished substitute. The "personal chef vs Febo vending machine" food-quality ladder makes the perceived-quality argument visceral. The McKinsey claim that B2B buyers are now happy to spend $500K+ via self-serve is challenged on methodological grounds. Recommendation engines (Netflix, Spotify, Stitch Fix) are positioned as the "first-order solution" — useful but unable to feel anything. Sephora's hybrid model is identified as the prior best practice. The pivot: large language models can simulate empathy and theory of mind, which reframes the strategic question from "humans or software?" to "how to teach the software faster?" The Lemonade/Maya insurance example anchors what next-gen looks like in practice.

AI couldn't believe another AI rejected his credit card.
Theory of Mind for Mind Reading 002

Theory of Mind — the ability to attribute mental states to oneself and others — is pulled out of developmental psychology and yoked to customer experience design. Goal inference, behavioural interpretation, contextual reading and social-norm inference are walked through. Mentalising is explained as the narrative-construction process by which others are predicted. The Mehrabian 55/38/7 result is invoked but explicitly de-generalised. Homo Economicus is dissected as a poor Theory of Mind model, and Affectiva and Replika are presented as evidence that machines can now simulate Theory of Mind well enough to influence behaviour. The Sebastien Bubeck GPT-4 Theory of Mind test is reproduced and treated as a watershed. The chapter closes by introducing four banking personas and demonstrating ChatGPT-4 generating credible Human-Centred Design touchpoints for each — including a creative cross-domain transfer (a luxury-car test drive as a banking touchpoint) when prompted with a Synectics-style technique.

AI listened as his therapy patient complained about other AIs.
CX Personalization 003

The framework chapter. The Simple / Satisfying / Surprising hierarchy is introduced and mapped onto CAPFUL (Certainty, Aspiration, Prestige, Fulfilment, Uncommon, Love). Warby Parker is dissected to show that excellent CX can be systematised without personalisation. Simple CX is illustrated by friction-killers: CLEAR airport biometrics, autofill, Grammarly and Harry's Razors. Satisfying CX covers Studio Pilates' three-touchpoint personalisation alongside a critique of IVR systems. Surprising CX covers VR experiences (Australian Wildlife Conservancy, Thomas Cook, Marriott), Disney Magic Bands, and Ticketmaster's Verified Fan AI. CX metrics — Customer Effort Score, Customer Satisfaction and Net Promoter Score — are distinguished and argued to be causally linkable only via AI. Varian's Law (mass customisation as the synthesis of low cost and personalisation) frames AI as the engine for democratising luxury.

As a youngster, AI thought it was funny to steal the TV remote control.
Digitization for Cost Reduction 004

A chapter on cost reduction structured around seven theoretical bases — economies of scale, scope, policy choices, low-cost inputs, externalities, learning curves and technology — with worked examples from Walmart, McDonald's, Apple, Toyota and Adidas Speedfactory. Open-source AI is positioned as a positive externality. The bulk of the chapter is a parade of last-generation chatbot deployments (Bizbike, iFood, Belfius, Rawbank, Proximus, Salud Digna), all delivering 40–80% deflection of routine queries. The jobs debate presents Brookings, World Economic Forum and PwC orthodoxy: automation displaces but creates more than it destroys. The closer flips the lens — a 1% Customer Satisfaction improvement cuts cost-of-sales by 0.53%.

AI was sad it couldn't buy everything on special at the market.
Revenue Growth with Empathic AI 005

The conflict between customer centricity and profitability is set up, then next-generation AI is argued to dissolve it. Three opening cases demonstrate cost-out projects with revenue-up surprises: Shenzhen airport, Virgin Australia's Wi-Fi entertainment as a weight-saving stunt that lifted on-time perception, and an energy firm reshoring its call centre for a 10% market-share gain. The substantive piece walks the full customer journey — Awareness → Consideration → Decision → Purchase → Use → Support → Repeat → Refer — with last-gen vs next-gen AI use cases at each stage. Cases include Neiman Marcus's Snap.Find.Shop, KFC × Baidu facial-recognition-driven menu suggestion, Chorus.ai vs Gong, Owkin's federated learning for cancer biomarkers, and AbsolutData's "digital twin of the customer."

AI loved to beat the humans at poker.
Finding AI Opportunities for CX 006

Pareto Analysis and Theory of Constraints are framed as the two methods for locating high-value AI deployments. The Pareto section makes the empirically valuable point that real distributions are usually 50/20 to 73/20, not 80/20 — and that Pareto is fractal, so the second order is roughly 64/4 and the third order roughly 51/0.8. AI's strength is performing this nested decomposition across far more variables than a human can. The Theory of Constraints section presents Drum-Buffer-Rope with industrial cases and then escalates to Mafia Offers: Hyundai's recession buy-back, Domino's 30-minute guarantee, Xerox's per-page model. Cogito (real-time empathy coaching at MetLife) is presented as the AI Mafia Offer enabler, and Babylon Health as the AI-native Mafia Offer. The chapter closes with a 10-step roadmap combining nested Pareto, Drum-Buffer-Rope and AI.

AI was bored waiting to get service from the customer support AI.
The New Sources of Competitive Advantage 007

The strategy chapter. Mintzberg's 5 Ps frame the opening, then the Resource-Based View and Dynamic Capabilities View are committed to as the working pair. The Resource-Based View is given the PROFIT taxonomy (Physical, Reputation, Organisational, Financial, Individual, Technology) and the EDGE test of advantage (Extra, Differentiated, Growth, Enduring). The pivotal move is the open-source problem: most current AI deployments confer no sustainable advantage because TensorFlow, PyTorch and rapid open-source publication mean any model is replicable. Three sustainable foundations for AI advantage are identified: (1) data differentiation — Amazon Prime's $1,400 vs eBay's $70 per-customer spend as the proprietary-data scale-play; (2) customisation and integration — GrammarlyGO as the test case; (3) domain expertise — Builder.ai and Amgen × Syntegon's AI visual inspection of pharma syringes.

AI was determined to win at all costs.
Trust in AI is Complex 008

The trust chapter is structured around four problem domains: Privacy (Cambridge Analytica; Clearview AI's GDPR fine), Effectiveness (NEC NeoFace's 96% misidentification rate in UK trials; Kmart and Bunnings suspending facial recognition), Transparency (Volkswagen's emissions-cheat AI), and Implications (Google's Project Maven withdrawal). Trust theory is built from Doney & Cannon's competence/benevolence/integrity model. The AI Trust Cohorts are the most actionable part: those who mistrust because AI disadvantages them; those who mistrust because they don't understand it; those unaware of AI's deployment; and those who trust because AI advantages them. Each cohort receives a tailored adoption strategy. The chapter closes with three critical warnings: don't pass AI off as human, don't exploit "prime vulnerability moments," and account for the unemployment externality.

AI felt very strongly about keeping data private.
Future Outlook 009

The forecasting chapter, opened with an honest disclaimer about the foolishness of prediction. A Google-search-derived methodology plots 23 CX trends against their AI support, producing five "waves." A methodological digression follows: the book itself was written as a Varian's Rule case study — author plus ChatGPT plus WebChatGPT plug-in over approximately 4 weeks vs 12–18 months for prior books, with the LLM doing roughly 80% of the wordsmithing and none of the conceptual content. Hallucinations are explained via the episodic-memory analogy. Seven future scenarios are presented in increasing order of speculation: more powerful hardware (Graphcore IPUs), quantum computing, healthcare savings of 1% of GDP, smarter chatbots, ubiquitous chatbots, more human interfaces (Soul Machines digital humans), and humans-as-cyborgs (Neuralink BCIs).

AI just couldn't see it coming.
Conclusion 010

A compressed walk-through of all nine prior chapters with the strategic implications spelled out for specific roles. The COO is told to bring the Head of Customer into the conversation. CX professionals are told to shift from quantity to quality — briefing and curating AI rather than producing. Boards and executive leadership teams are told to chase 5x–20x gains rather than WACC plus a margin. CIOs are identified as the owners of "the unique layer of AI functionality the firm overlays to analyse privileged customer data." The closing argument: the strategic asset is no longer the AI itself — it is the proprietary data, the customised integration, and the domain expertise layered on top. Firms that try to exploit customers will draw regulatory reaction; firms that build trust will be the ones who turn the empathic algorithm into durable advantage.

Fortunes changed after AI took over everything.

About the Author

Robert Dew is a Managing Partner at CapFeather Global, a boutique consulting firm specialising in customer strategy. His work sits at the intersection of behavioural science, artificial intelligence, and customer experience design.

This book was written in collaboration with ChatGPT 4.0 — itself an experiment in human-AI partnership, and a proof of concept for the very ideas explored within.

Published by CapFeather Global, Melbourne, Australia. ISBN: 978-0-6458539-1-9 (eBook).

Download the Book (PDF — coming soon) Watch the Presentation Listen to the Audio