Table of Contents

• This Week

• Business Use Cases

• Things to Pay Attention to

This Week

Bottom Line Up Front

• OpenAI has released GPT-5.2, its latest model, which aims to "unlock even more economic value for people" according to Fidji Simo, CEO of Applications.

• GPT-5.2 is described as the company's "best model yet for everyday professional use", capable of tasks such as creating spreadsheets, building presentations, and writing code.

• The model is also designed to be more useful for AI agents' workflows, with the goal of making ChatGPT a personalized assistant, and features improvements such as reduced "hallucinations" in its Thinking model.

Business Use Cases

Bottom Line Up Front

• CEOs expect AI spending to continue rising through 2026, despite difficulties in measuring clear returns on investment.

• The main barriers to stronger AI returns include the challenge of scaling pilots, infrastructure costs, and issues with data quality, system links, and regulatory requirements.

• To achieve sustainable returns, companies are resetting their expectations, prioritising fewer use cases, and focusing on integrating AI into everyday operations with clearer ownership, governance, and realistic timelines.

Why is this important

• Salesforce CEO Marc Benioff has initiated a discussion on enterprise AI pricing, shifting from usage-based to seat-based licensing due to customer demand for financial predictability.

• The AI market is expected to reach $450 billion by 2035, but vendors face a paradox: if AI delivers productivity gains, enterprises may need fewer licenses, conflicting with traditional per-user economics.

• Salesforce's adoption of seat-based pricing with credits, caps, and "fair use" language offers a compromise, providing certainty for enterprises while protecting vendors from spiraling costs.

Why is this important

• Hallym University Sacred Heart Hospital in South Korea has deployed AI brain imaging analysis software from JLK to support emergency stroke diagnosis.

• Apollo Hospitals in India has implemented an AI-powered scan system to accelerate acute stroke care, reducing scan-to-interpretation time from 30 minutes to 7 minutes.

• Other hospitals, including Yashoda Hospital and Bangkok Hospital, have also introduced AI-enabled solutions, such as an AI lung clinic and a remote heart monitoring device, to improve patient care.

Bottom Line Up Front

• The percentage of U.S. employees using AI at work has increased from 40% to 45% between Q2 and Q3 2025, with frequent use growing from 19% to 23%.

• Employees in knowledge-based jobs, such as technology or professional services, are more likely to use AI than those in frontline positions, with 76% of employees in technology or information systems using AI at least a few times a year.

• Gallup research shows that broader AI adoption among employees is strongly associated with having greater managerial support for AI and strategic integration of AI in their role, with leaders like Ryan Pendell and Andy Kemp contributing to the research.

Bottom Line Up Front

• AI is no longer an advantage, but a necessity for gyms to stay competitive, with those using it experiencing higher conversion rates, lower cancellation rates, and better member spend.

• According to Dheeraj Koneru, Co-Founder of Zenoti, AI is taking over repetitive tasks such as follow-ups, pattern recognition, and automated engagement, allowing gyms to scale without adding headcount.

• Zenoti, a platform built on AI from the ground up, is helping gyms operate with greater consistency, clarity, and long-term stability, providing instant answers to key business questions and enabling operators to focus on building experiences and connections with members.

Things to Pay Attention to

Bottom Line Up Front

• Many tech leaders, including OpenAI CEO Sam Altman and Google CEO Sundar Pichai, believe that the current hype around AI is outpacing reality and that the industry is experiencing a bubble.

• The bubble is being inflated by companies raising enormous sums of money and seeing unprecedented valuations, with much of the funding going towards building massive data centers.

• If the bubble bursts, it could happen when overfunded startups can't turn a profit or grow into their lofty valuations, with potential ripple effects being profound, especially for companies like OpenAI and Anthropic that have yet to turn a profit.

Bottom Line Up Front

• Researchers have been exploring the ability of language models to reason about language itself, with some arguing that AI models are not capable of analyzing language in a sophisticated way.

• A recent paper by Gašper Beguš, Maksymilian Dąbkowski, and Ryan Rhodes challenged this view, finding that one large language model was able to analyze language in a way similar to a human expert, including diagramming sentences and resolving ambiguous meanings.

• This finding is significant, as it challenges our understanding of what AI can do and has important implications for evaluating the abilities of language models to reason like humans, according to experts like Tom McCoy.

Bottom Line Up Front

• Scientists at the Icahn School of Medicine at Mount Sinai have developed a new AI tool called V2P (Variant to Phenotype) that can predict how specific DNA mutations translate into disease.

• V2P uses advanced machine learning to connect genetic variants with their likely phenotypic outcomes, allowing for faster diagnoses and supporting the search for new treatments for complex and rare conditions.

• The tool has the potential to guide drug discovery and move the field closer to precision medicine, where therapies are chosen to match an individual's genetic profile, according to researchers such as Dr. Yuval Itan and Dr. Avner Schlessinger.