Block and GSK pioneer AI agents to enhance workflows and drug discovery

August 26, 2025 — Global enterprises Block and GlaxoSmithKline (GSK) are at the forefront of integrating AI agents into their operations, demonstrating early success in financial services and drug discovery. While the hype around AI agents is at its peak, these companies are proving tangible ROI by aligning AI with existing workflows rather than overhauling processes.

Block’s AI Agent Framework: Goose

Block, parent company of Square and Cash App, has developed Goose, an open-source AI agent framework initially designed for software engineering. Since its January launch, Goose has been adopted by 4,000 engineers, doubling monthly. The platform automates 90% of code generation, saving engineers 10 hours weekly.

Key features of Goose include:

• Real-time code generation and debugging
• Integration with Slack, email, and other tools
• Autonomous file handling and dependency management

Brad Axen, Block’s Tech Lead for AI, emphasized the importance of human-centric design: "We want users to feel like they’re working with one colleague, not a swarm of bots." Goose operates on Anthropic’s Model Context Protocol (MCP), enabling seamless tool integration.

GSK’s Multi-Agent Approach to Drug Discovery

GSK is applying multi-agent systems to accelerate pharmaceutical research, particularly in genomics and proteomics. Kim Branson, SVP of AI/ML at GSK, highlighted the challenge of lack of ground truth in drug discovery, requiring rigorous testing and domain-specific models.

GSK’s strategy includes:

• Building custom LLMs for epigenomics
• Cross-checking results with parallel agent runs
• Developing internal benchmarks for reliability

Branson noted: "We hunt for problematic outcomes—that’s where we learn the most."

Key Takeaways

• Process alignment is critical: AI agents must fit into existing workflows.
• Human expertise remains vital: Especially in regulated fields like finance and pharma.
• Open-source and standards drive adoption, as seen with Block’s Goose and MCP.

For more on AI scaling challenges, join VentureBeat’s salon.