The trillion-dollar question

Plus, McKinsey strategy with AI, Gemini 2.0 expansion, and more.

Lewis Walker
February 09, 2025

WELCOME, EXECUTIVES AND PROFESSIONALS.

McKinsey reveals its latest take on strategy with AI, while scaling best practices and models continue to evolve. But, as leaders pursue value, are they prepared to address the trillion-dollar question?

Since the previous edition, we've reviewed hundreds of the latest insights on best practices, case studies, and innovation. Here’s the top 1%...

In today’s edition:

McKinsey: How AI is transforming strategy.
Deloitte: Improving gen AI data and model quality.
Google expands Gemini 2.0 family production use.
The trillion-dollar question.
Transformation and technology in the news.
Career opportunities & events.

Read time: 4 minutes.

BEST PRACTICE INSIGHT

McKinsey: How AI is transforming strategy

Image source: McKinsey & Company

Brief: McKinsey’s article explores five roles for AI in transforming strategy: researcher, interpreter, thought partner, simulator, and communicator. These roles play a part in various phases of strategy development, from design to execution.

Breakdown:

Researcher. AI summarizes and connects data across sources, such as identifying potential M&A targets by spotting under-the-radar assets.
Interpreter. AI transforms data into insights and tracks trends to advance goals, for instance by conducting "growth scans" or monitoring competitor signals.
Thought partner. AI acts as a brainstorming partner, generating ideas, challenging strategies, and uncovering blind spots to refine decision-making.
Simulator. AI enhances scenario analysis by modeling the impact of macroeconomic economics, competitor moves, and stakeholder reactions.
Communicator. AI tailors narratives for different audiences and formats, from briefs to podcasts, to help drive action.
McKinsey expands on these roles in the context of a Southeast Asian bank pursuing expansion and outlines steps to level up your strategy with AI.

Why it’s important: McKinsey believes AI can’t, and won’t, replace human logic and interpretation in complex domains like strategy. However, it can provide faster, more objective answers to improve decision-making. By improving efficiency and allowing room for creativity, AI enables bold moves needed to beat the market.

BEST PRACTICE INSIGHT

Deloitte: Improving gen AI data and model quality

Image source: Deloitte

Brief: Deloitte's article investigates how tech leaders: data architects, engineers, scientists, and security leaders, can build AI and data environments to tackle gen AI data integrity and model accuracy challenges as programs scale.

Breakdown:

The research draws on a literature review, specialist interviews, and analysis of survey data from Deloitte’s Q3 and Q4 2024 State of Generative AI surveys.
Enterprises struggle without a clear strategy and data architecture that cuts across data types and modalities while aligning with regulations.
Legacy data environments aren’t designed for probabilistic gen AI. Transform data efficiently to help mitigate increased training and retraining costs.
Leverage automated review approaches, enhanced chunking, and advanced retrieval methods to address RAG engineering challenges.
Hallucinations that impair trust can be mitigated with human oversight, and for instance, by constraining a prompt’s scope to predefined parameters.

Why it’s important: The article offers more detail but stresses there’s no one-size-fits-all solution. A clear data strategy, optimized training data integrations, and continuous tuning are key to quality gen AI solutions. Neglecting these risks financial loss, reputational damage, and scaling issues.

INNOVATION INSIGHT

Google expands Gemini 2.0 family production use

Image source: Google

Brief: Google announced new and updated AI models in its Gemini 2.0 family, including the highly anticipated Pro Experimental, the cost-efficient Flash and Flash-Lite, and Flash Thinking, a model designed for advanced reasoning.

Breakdown:

Google expanded Gemini 2.0 access for developers and enterprises. These models are now in the Gemini API via Google AI Studio and Vertex AI.
Gemini 2.0 Flash is now generally available, offering higher rate limits, stronger performance, and simplified pricing.
Gemini 2.0 Flash-Lite, Google's new variant is its most cost-efficient model, and now available in public preview.
Gemini 2.0 Pro Experimental, Google's top model for coding and complex tasks, is now available with a 2M token context window.
These releases add to the recently launched Gemini 2.0 Flash Thinking Experimental, Google’s Flash variant that shows its reasoning before answering.

Why it’s important: Google’s Gemini 2.0 releases strengthen its range of models. Google Cloud’s revenue grew 30% YoY last quarter, as enterprises leverage its integrated Gemini-Vertex AI offer. But with rivals like OpenAI and Microsoft continuously raising the bar, the competition remains intense.

ACCELERATOR

The trillion-dollar question

Brief: Cognizant’s 25-page report addresses the trillion-dollar question: what to do with the workforce in the age of gen AI. Analyzing 1,000 professions, it provides leaders with an initial framework to rethink roles and realign workforce strategies.

Breakdown:

Cognizant forecasts gen AI will improve US productivity by $1 trillion annually by 2032, impacting 90% of jobs, with 52% significantly affected.
Gen AI will reshape the talent pyramid by narrowing the base (entry-level roles), which are more automatable, while expanding higher levels.
Roles that remain fundamentally unchanged by AI, such as physical highly specialized work like electricians, will experience minimal impact.
Those augmented by AI, such as executives and lawyers, will be impacted moderately. They should upskill in new ways of working with gen AI tools.
Others transformed by AI, such as programmers and financial analysts, will experience significant reskilling and evolution in their roles.
Fully automated roles, like data entry and many customer service positions, will be reassigned or phased out entirely.

Why it’s important: Enhanced, augmented, transformed, and fully automated roles exist across enterprises. Talent management decisions today are key for preparing individuals, shaping the workforce to balance automation with human expertise, and ensuring long-term success in the market.

Bain explored five key questions every CEO should consider for winning with AI. These focus on speed, the future of industry, AI’s role in strengthening differentiation, building tech foundations and making it happen.

IBM shared an article on agentic architecture, covering how it works, the difference between agentic and non-agentic, types of agentic architectures, key use cases, benefits, and choosing the right frameworks.

Booz Allen published an article on the rise of agentic AI, exploring its practical applications and two in-depth use cases. It also addresses the balance between autonomy, functionality, and control.

Microsoft updated its repository with 50 new AI transformation stories, bringing the total to over 300. New customer examples include Lloyds Banking Group, Air India, UBS, BMW, McKinsey & Company, and more.

Operand is building an AI to "kill McKinsey". The startup helps businesses make better decisions by deeply analyzing their data across all sources, something they argue mainstream consultants can’t do at scale.

GitHub announced major Copilot updates, including an agent mode for self-correcting code, ‘Vision’ for image-to-code generation, and an upcoming autonomous coding agent.

OpenAI demoed an automated sales agent in Tokyo, showcasing its ability to handle enterprise lead qualification and meeting scheduling.

ByteDance researchers unveiled OmniHuman-1, an AI that generates ultra-realistic deepfake videos from a single image and audio input.

Stanford released a 1-hour introduction to agentic language models (LMs), covering common limitations and patterns, such as reflection, planning, tool use, and iterative LM usage.

Hugging Face's 17-page paper argues against developing fully autonomous AI agents, citing the increased risks to people as system autonomy grows.