Poolside's release of Laguna XS.2 represents a deliberate shift in the AI arms race: rather than competing on model size or proprietary performance, the startup has prioritised accessibility and local deployment for agentic coding tasks. The model is free, open-source, and designed to run on consumer hardware without cloud dependencies—a direct counterpoint to the escalating costs of frontier models from Anthropic and OpenAI. This positioning matters because it democratises the infrastructure layer that underpins agent-based automation, the very capability that's reshaping contact centre operations across Microsoft, Amazon, and Genesys deployments.
For CX teams, the implications are twofold. First, the availability of capable open models reduces vendor lock-in risk and lowers the barrier to building custom agents tailored to specific workflows—whether that's automating first-contact resolution, routing, or knowledge retrieval. Teams currently evaluating Copilot Studio or Amazon Connect's agentic capabilities now have a credible alternative for internal development, which shifts the negotiating position with enterprise vendors. Second, and more strategically, this raises a critical question: as open models close the performance gap with proprietary alternatives, does the competitive advantage for platforms like Salesforce Agentforce or Genesys shift entirely toward integration depth and domain-specific training data rather than raw model capability? The real bottleneck for most CX operations isn't access to intelligence—it's the ability to connect that intelligence to customer context, history, and business logic. Poolside's move suggests that the next phase of competition will be won by teams that can operationalise agents effectively, not by those holding the most advanced black-box models.
The AI race lately has felt a bit like a game of tennis: first, Anthropic releases a new, pricey state-of-the-art proprietary model for general users (Claude Opus 4.7), then, a week or so later, its rival OpenAI volleys back with one of its own (GPT-5.5). And all the while, Chinese companies like De