Future of AI | Leaders on Turning Innovation into Action

Key Concepts

• AI Growth Acceleration: How Artificial Intelligence is driving significant growth and improving margins for businesses.
• Enterprise AI Adoption: Strategies and challenges companies face when implementing AI in their daily workflows.
• Pharmaceutical R&D: The impact of AI on accelerating drug discovery and development timelines.
• AI Security and Sovereignty: Concerns and solutions related to data privacy and control when using AI tools.
• Multi-LLM Strategy: The approach of utilizing multiple Large Language Models (LLMs) for diverse applications.
• AI Agent Building: The development of AI agents to enhance productivity and automate tasks.
• LLM Commoditization: The potential for LLMs to become a widely available, standardized technology.
• "Tech for Tech's Sake" Avoidance: The principle of implementing technology with a clear business purpose and measurable outcomes.
• AI Metrics for Success: Key performance indicators used to evaluate the effectiveness of AI implementations.

GetYourGuide: AI as a Growth Accelerator

Johannes from GetYourGuide highlights the company's significant growth, announcing they are on the cusp of one billion Euros in revenue. They have booked over 10 million experiences in Q3, representing a 30% year-over-year growth. GetYourGuide has also achieved profitability for the first time in its history over the last 12 months, being EBITA profitable and producing cash flows.

AI has been a key factor in this acceleration, particularly in managing the complexity of their business. GetYourGuide works with 35,000 local supplier partners, ranging from major attractions like the Louvre and Eiffel Tower to local artisans. AI has enabled them to:

• Aggregate content from these diverse suppliers.
• Help merchants describe their offerings properly in 40 different languages.
• Ensure consumers can understand the differences between various experiences in a scalable way.

Johannes emphasizes a multi-pronged approach to AI, focusing on enhancing internal productivity and, more importantly, improving the consumer experience by making it easier to find the right travel experience.

Enterprise AI: Seamless Integration and Cost-Benefit Analysis

Eléonore discusses the enterprise perspective on AI implementation. She notes that while many business leaders want to implement AI, the key is inserting it into individual contributors' day-to-day workflows rather than forcing adoption. The focus is on understanding painful daily tasks for teams, particularly supply chain teams, and automating them to enhance productivity. The goal is seamless integration without adding friction.

Regarding pricing, Eléonore explains that their current model charges per mission completed. They aim to transition to usage-based pricing, as success is measured by adoption, accuracy, productivity gains, and big business outcomes. She mentions a potential customer, a leader from Berkeley, who sees value only when AI is truly integrated into their daily workflow.

Pharmaceutical Industry: Accelerating R&D with AI

Shobie from GSK shares insights from transitioning from Silicon Valley tech to the pharmaceutical industry. She contrasts the iterative, controllable nature of tech ("like carpenters") with the longer, nature-dependent cycles of pharma ("like gardeners"), where bringing a medicine to market takes 10-12 years and billions of dollars.

For GSK, R&D productivity is paramount. With a 9 out of 10 failure rate for ideas, AI's ability to:

• Digitally imagine the lifecycle of disease.
• Understand disease progression.
• Intervene at the right stages.

is seen as having a massive impact. The primary priority is shortening the timeline for drug development. Equally exciting is AI's potential to identify new scientific hypotheses by processing data differently than humans. A crucial application is understanding which patient will benefit from which medicine, allowing for more targeted clinical trials. These capabilities were not possible before AI and machine learning.

AI Investment: Red Flags and Strategic Considerations

Matt discusses red flags for companies implementing AI, particularly in the context of an investment in Nexos. A major concern is companies being too eager to enable teams with AI toolkits without taking stock of the information being shared. This includes understanding what data is leaving the organization, how it's being secured, and whether processes are efficient.

Nexos is presented as a platform that helps companies securely adopt AI while monitoring information sharing. Matt highlights the concept of "enterprise sovereignty," the desire to keep certain data within company walls. He likens the current AI adoption phase to the early days of mobile, where the initial productivity boost was followed by the need for mobile security and efficient enterprise use. The focus is shifting towards using AI securely and managing it efficiently.

Multi-LLM Strategy and Agent Building

Eléonore explains that their multi-LLM strategy is driven by customer demand. They work with models from OpenAI, Anthropic, and Gemini. Since much of their work involves numerical data, compute happens on their platform (Pigment), with LLMs serving as the interface. They are mindful of not going too fast and are considering other LLMs, but prioritize ensuring their built agents are functional. This mirrors a multi-cloud strategy, focusing on the end product.

Shobie echoes the sentiment of not being attached to a single model, emphasizing the ability to find the right model for the right problem. They are moving towards agent building platforms that allow for evaluation of quality and cost. While curious about LLM innovation, their focus remains on the application standpoint for creating vaccines and medicines. They have evaluated models like Mistral but found them not differentiated enough for their specific needs. The key is tailoring the solution to specific problems, not just industry.

Johannes views LLMs as potentially becoming a commodity, with differentiation coming from the value proposition towards the customer and superior data sets. He notes GetYourGuide's significant growth in brand and app traffic over the last two to three years, with close to 50% of transaction volumes now coming through their app.

Matt agrees that LLMs will likely become commoditized, similar to cloud infrastructure. He believes successful AI companies will build applications on top of LLMs, making LLM providers like AWS, OpenAI, and Anthropic highly valuable due to their data troves. He acknowledges the risk of failures in these large providers impacting the wider internet, similar to cloud outages.

"Don't Do Tech for Tech's Sake"

Shobie emphasizes the principle of "don't do tech for tech's sake." In the pharmaceutical industry, the core purpose is to integrate technologies into workflows like discovery, drug development, and manufacturing. If work isn't reimagined, the coolness of the tech is irrelevant. The focus is on serving patients and measuring success by tangible outcomes, such as patients living longer. This principle grounds the industry in understanding what technology truly serves. They are reimagining planning with AI and addressing capability deficits to make technology work for enterprises.

AI Adoption Metrics and Business Outcomes

When measuring success, Shobie states they are relentless in chasing outcome metrics:

1. Does it help us make money or move medicine faster through the pipeline?
2. Learning loops: Establishing effective feedback between AI and humans.
3. Fluency metrics: How many people are learning to use the AI tools.

Johannes prioritizes business outcome metrics, viewing AI as a tool to achieve them. For GetYourGuide, the main metric is revenue growth.

Matt looks for direct links between AI tools and revenue growth, customer delight, and troubleshooting success. He highlights companies with significant revenue growth due to opening AI to new users and software creation. He also notes companies focusing on business outcomes and the ROI for larger enterprises. For companies like Anthropic, adoption is a key metric. He also stresses the importance of accuracy and governance.

AI Bubble and European AI

In a rapid-fire round:

• Buying AI companies solely because they are European: The consensus is to buy the best technology, regardless of origin.
• Are we in an AI bubble?
  • Eléonore: Partly yes, it will likely burst with some destruction and valuation adjustments.
  • Matt: No, it's a revolution with larger funding. We've barely scratched the surface. While there will be wasted money, greatness will emerge.

The panel concludes with thanks, acknowledging the significant impact and ongoing evolution of AI.