AI Pulse · AI Trends Pulse

Google just released two new Gemini models, 3.6 Flash and 3.5 Flash-Lite, and the interesting bit is not the version bump. It’s what they’re built for. These aren’t general-purpose chat models. Google explicitly positioned them for high-volume agentic tasks and subagent workflows, the kind of work where one AI calls another AI to break down and execute steps inside a larger process.

That positioning matters because it signals where the real infrastructure spend is going. Companies aren’t just asking AI to answer questions anymore. They’re building systems where agents orchestrate other agents, one handling data retrieval, another validation, another formatting, all chained together. Flash-Lite in particular suggests Google expects people to run thousands of these calls in parallel, cheap and fast, without worrying about token budgets blowing out.

What this means for operators

If you’re running any kind of repeatable workflow, document processing, lead enrichment, compliance checks, this is the architecture you should be thinking about. Not one big model doing everything, but a coordinated set of smaller, cheaper calls that hand off context and results. That’s exactly the kind of orchestration we build into the Omni Command Centre, where you define the workflow once and let subagents handle the repetitive execution.

The models picked up traction across AI Twitter within hours, which tells you the developer community was already waiting for this. They’ve been hacking together multi-agent systems on older models, and now Google’s given them purpose-built tools. If your competitors are paying attention, they’re already testing how to automate parts of their operations you’re still doing manually. The question isn’t whether this approach works. It’s whether you’re building it before someone else does.