Skip to content
The Definitive Source To Digital Transformation

2025 April 2nd Update: Integrating AI Strategy Duality, Skill Inference, and Human-AI Collaboration Nuances

Today’s update integrates recent thinking on developing AI strategy, managing talent in the AI era, and understanding the practical challenges and nuances of human-AI collaboration. We introduce the concept of strategic duality for AI planning, drawing parallels to Kahneman’s “thinking fast and slow.” New talent management approaches include using AI for skills inference and alternative career progression models like the “career lattice.” The update also incorporates research findings on the risks associated with novice users training others on Generative AI and the differing risk mitigation tactics employed by novice versus expert AI users, emphasizing the importance of system design and expert oversight.

Furthermore, specific case studies are added, such as Johnson & Johnson’s AI-driven skills inference initiative and Accenture’s work using AI to predict GenAI’s impact on job roles, providing concrete examples of these evolving practices. The importance of leadership in managing systemic AI risks and embedding responsible AI practices is also reinforced.

New

Section titled “New”
  • AI Strategy Duality (Fast/Slow Thinking): Introduces the concept of applying a dual approach to AI strategy development, mirroring Kahneman’s “thinking fast and slow.” This involves using rapid, focused experiments (“fast thinking”) to generate insights that inform more deliberate, long-term strategic planning (“slow thinking”), balancing agility with sustainability. Ch 2, 4
  • AI for Skills Inference: Details the emerging practice of using AI (specifically LLMs) to analyze diverse employee data sources (HRIS, LMS, recruiting databases, project platforms) to quantify skill proficiency, identify gaps, and inform strategic workforce planning. Includes the Johnson & Johnson case study demonstrating this approach. Ch 12, 1
  • “Career Lattice” Progression Model: Presents an alternative to the traditional “career ladder,” suggesting interconnected roles that allow for lateral moves, transitions into emerging roles (like AI prompt engineering), or even voluntary step-backs, offering more flexible career pathways aligned with dynamic skill needs. Ch 12, 3
  • Novice vs. Expert AI Risk Mitigation: Incorporates research findings highlighting that novice users of Generative AI tend to recommend risk mitigation tactics focused on changing human routines or project-level fixes, while experts prioritize systemic design improvements, use case selection, and rigorous testing. This underscores the need for expert guidance in AI deployment. Ch 12, 2, Ch 16, 8
  • AI Cognitive Offloading Risk: Explicitly adds the risk of “cognitive offloading” where users excessively trust AI outputs, especially when using tools like GenAI outside their optimal capabilities, leading to potential errors. Ch 16, 3

Updated

Section titled “Updated”
  • Establishing Digital Leadership Capabilities: Updated to emphasize that AI risk mitigation strategies must address systemic factors (system design, firm policies, ecosystem interactions) and not remain solely at the project or individual routine level. Ch 3, 4
  • Demystifying AI (Precision vs. Speed): Added a nuance contrasting “legacy machine learning” (often better for precision tasks like individual prediction) with Generative AI (often better for speed in areas like content creation). Ch 5, 5
  • Generative AI Limitations (Jagged Frontier & Task Performance): Incorporated research highlighting the “jagged technological frontier” of GenAI, showing performance significantly improves on tasks inside the frontier but degrades significantly on tasks outside it. Also noted the finding that lower-skilled workers may see larger performance gains within the frontier. Ch 5, 8
  • Augmenting Human Capabilities (Human x Machine & Interaction Styles): Reframed augmentation as “humans multiplied by machines” and introduced different human-AI interaction styles: “Cyborg behavior” (intertwined effort) and “Centaur behavior” (strategic switching between human/AI execution). Ch 5, 11
  • Strategic AI Implementation (Subproblem Decomposition): Added the technique of breaking down larger business challenges into subproblems to better identify appropriate AI techniques. Ch 5, 13
  • Building AI Capabilities (Technical Decision-Making & Partners): Emphasized the need for technical decision-making skills alongside technical execution skills, and the role of third-party data partners. Ch 5, 15
  • Integrating IA/AI (Role Reconfiguration & Wrappers): Added the importance of potential role reconfiguration and considering interface design (“wrappers”) to mitigate AI risks during integration. Ch 5, 17
  • Upskilling and Reskilling (Risks of Informal Training): Included the specific risk associated with relying on informal peer training for GenAI, as novices may propagate suboptimal risk mitigation tactics. Also emphasized empowering workers in what, when, and how they learn and apply skills. Ch 12, 2, Ch 12, 3
  • Impact of Automation/AI (AI for Workforce Planning): Added the Accenture example of using AI to analyze task/skill data to predict GenAI’s impact and inform workforce redesign. Ch 12, 4
  • Embedding Digital Norms (AI-Specific Practices): Incorporated specific practices for embedding AI use, including structured onboarding, peer training programs (with recognition), role reconfiguration discussions, and reinforcing accountability separate from AI attribution. Ch 13, 8
  • AI Imperfections Test (Human Oversight): Added a tenth point emphasizing the need for continuous human cognitive effort and expert judgment to validate AI outputs and determine appropriate use cases, given AI’s limitations (e.g., the “jagged technological frontier”). Ch 14, 2
  • AI in Sales (Propensity Models & Look-alikes): Enhanced the description of intelligent lead qualification to include propensity-to-buy models and the use of ‘look-alike’ modeling based on external data signals. Ch 20, 3
  • Sector Case Studies (New AI Examples):
    • Healthcare: Added Pfizer and Takeda case studies using GenAI/LLMs to accelerate knowledge transfer (R&D to manufacturing) and clinical trial design, respectively. Ch 23, 4
    • Retail: Added Dick’s Sporting Goods using GenAI combined with customer data for rapid personalized email campaign creation. Ch 23, 6
    • Telecoms: Added Comcast using AI on customer/call data to improve real-time agent responses and identify churn risk. Ch 23, 11
    • Transportation/Logistics: Added CMA CGM using GenAI on transactional data for dynamic pricing advice. Ch 23, 15
    • Insurance: Added CogniSure using GenAI to extract information from varied customer submission formats (PDFs, emails) to speed up quoting. Ch 23, 17
    • Internal KM/Productivity: Added McKinsey & Co. using GenAI developed with MIT Sloan students to automate document labeling for knowledge management. Ch 21, 5

Removed

Section titled “Removed”
  • No specific content sections were removed in this update; changes primarily involved adding new concepts and integrating AI-related nuances and examples into existing chapters.