McKinsey & Company — searchable YouTube transcripts & AI Q&A

What Does a Forward-Deployed Engineer Actually Do?

A forward-deployed engineer embeds with clients to tailor AI and software to real business problems, leaving behind working products rather than just recommendations. The role requires a blend of technical skills, business acumen, and interpersonal abilities, as success hinges on changing workflows and driving adoption. The speaker argues that with AI, the role is merging with that of a consultant, and the hardest part is often managing the human side of change.

Why Energy Demand Keeps Growing | McKinsey Global Energy Perspective 2026

The energy transition is actually an expansion of global energy demand, driven by unequal energy access and rising electricity use from AI and data centers. The speaker argues that while decarbonization progress is real, no scenario meets the Paris Agreement's 1.5°C goal. The Strait of Hormuz shock showed that resilience measures work but have limits, so executives should use scenarios to identify safe bets and optionalities.

How AI Could Help Revive Growth in the Food and Beverage Industry

McKinsey partners argue that food and beverage companies should focus AI investments on a single domain to scale benefits and fund growth, rather than running scattered pilots. They report that the sector lost nearly a trillion dollars in enterprise value due to slowing volume growth, and that consumers are trading both up and down, favoring private label and small brands. To win, companies need true product innovation, smart portfolio choices, and AI-driven growth, not just cost cutting.

Addressing AI Frontier Lab model risks and opportunities

Frontier AI is compressing the cyber attack lifecycle, making sophisticated attacks faster, cheaper, and more autonomous, which forces organizations to shift from a pyramid-shaped security team to a diamond of overseers who manage AI agents. The speaker argues that to stay ahead, companies must delegate decision-making lower in the organization, automate vulnerability remediation, and focus on securing non-human identities and data. They also emphasize that while quantum computing is a longer-term risk, organizations should start preparing now by improving data security and planning for post-quantum cryptography.

What Does AI Really Cost? Why Companies Are Spending More Than Expected

AI costs are rising faster than expected because providers are shifting from per-seat licensing to consumption-based pricing, and users are consuming more tokens than necessary, often choosing more powerful models than needed. The speaker argues that companies must track all AI spend sources, including indirect costs like cloud compute and embedded SaaS, and measure cost per successful output rather than just per token. A key example is a client whose usage and spend doubled in two months as users switched to a more advanced model without needing it, highlighting the need for governance and visibility.