Block's AI Agents Graduate From Chat to Collaborative Workspaces

Block Inc. demonstrated a shift in how artificial intelligence integrates into enterprise work at its Builderlab Launch Event 2026 in San Francisco. Rather than confining AI agents to individual chat windows, the company has moved them into collaborative workspaces where they operate as identifiable team members with full audit trails. The approach builds on Buzz, Block's open-source workspace designed to accommodate both humans and AI agents working in tandem. Block developed these capabilities internally to operate as an intelligence company and is now making them available to other organizations through an open-source model.

The evolution of agent technology hinges less on individual agent capability and more on how effectively multiple agents can work together, access organizational context and remain within security parameters. Block intends to distribute this model through open-source code, architecture that remains independent of specific AI models and cloud infrastructure services. This expansion occurs as Block's core business achieved its third consecutive guidance increase.

If you're talking about making the whole organization more productive, you need to break down these distinctions between what the software does and what the people do.

Paul Gillin, enterprise editor of SiliconANGLE Media Inc.

Alison Kosik and Paul Gillin conducted interviews with industry leaders at the event, exploring Block's open-source strategy, agent identity systems, the transition from manual prompting to automated overnight tasks and the security frameworks that enable enterprises to rely on autonomous systems.

Four Key Takeaways

1. Open-source architecture transforms Block's internal agent platform into a commercial offering

Builderlab provides developers with highly adaptable agents built on an open-source plug-in framework. The launch drew enthusiasm from application developers in attendance, according to Gillin. Agents within Buzz possess distinct names, visual avatars and defined personalities, shifting attention away from juggling multiple software-as-a-service tools toward the actual work being performed. This capability could pose challenges to traditional SaaS vendors. The business model adheres to the established open-source pattern: distribute the software at no cost while generating revenue from the convenience of a managed cloud service. Block maintained limited public discussion of monetization strategy or Cash App, which accounts for 60% of its revenue. Gillin characterized this restraint as strategic, allowing developers to construct integrations and discover applications independently.

2. Buzz transforms AI agents into tagged team members within shared environments

Conventional AI tools operate on a single-user basis, creating a constraint when findings must reach a broader team, explained Lucas Isaza, product lead of Buzz at Block. Within Buzz, agents possess individual identities, can be mentioned like human colleagues and maintain a complete record of their operations. Plug-ins introduced following the platform's July launch enable users to customize the interface. Non-technical roles, particularly account managers designing custom dashboards, have emerged as the primary users of the companion apps platform. Organizations have three deployment options: hosting the open-source code themselves, accessing a free tier or paying for cloud-based agents, AI models and expanded storage capacity, Isaza noted.

3. Cloud-based agents enable enterprises to delegate work overnight rather than manage prompts

Builderlab functions as an interface layer positioned above AI models, enabling Buzz to incorporate agents from any provider and upgrade to emerging frontier models as they become available, according to Brad Axen, head of AI capabilities at Block. Cloud-based agents operate in centralized environments that offer superior security compared to distributed laptop deployments. Teams can assign substantial tasks to agents for completion overnight and subsequently review results. Rather than relying on soft guardrails, agents operate within secure foundations that prevent them from obtaining authorization for restricted actions. Certain teams have accelerated software releases by 50% to 75%. Axen anticipates agents will require significantly less human oversight within the coming year because they already participate in ongoing conversations.

4. Agent identity and data governance form the foundation of Block's security approach

Securing autonomous agents does not require abandoning established security principles such as least privilege access, environment isolation and egress filtering, explained James Nettesheim, chief information security officer and head of enterprise technology at Block. Builderlab assigns each agent a complete identity controlled by a designated human, executes it within hardened cloud virtual machines and logs all actions in a cryptographically signed ledger resistant to tampering. Block is developing information flow control capabilities that attach security policies to sensitive data, preventing unauthorized transmission. Nettesheim noted that AI already assists Block in code review and security alert categorization. He added that agents enable security teams to operate at machine speed, matching the pace of potential threats.