AI assistants shouldn't just wait for you to ask questions.

On July 23, Feishu aily — the Agent product inside Feishu — announced a major upgrade with three core capabilities: proactive work, team-shared agents, and multi-agent collaboration. This marks enterprise office agents' evolution from “Q&A chatbox” to “digital colleague that proactively gets things done within your authorization.”

From Passive Q&A to Proactive Execution

Past enterprise AI assistants shared one trait: passivity. You ask a question, it gives an answer; you tell it to write a doc, it writes a doc. But if you don't reach out, it never reaches out. That's not a real assistant — that's a fancy search engine.

The new aily's biggest breakthrough is proactivity. Within user authorization, aily can proactively follow up on work based on Feishu messages, docs, calendars, meetings, tasks, and bitables. For example: auto-extracting action items and assigning owners after meetings; proactively reminding as project deadlines approach; detecting cross-team bottlenecks and aligning information; automatically creating tasks from customer requests and tracking them to closure.

This sounds simple but requires solving hard problems behind the scenes: context understanding (knowing what matters vs. what's noise), permission boundaries (never acting beyond authorization), notification frequency (not becoming a pop-up spammer), and operational traceability (everything the agent does must be transparent and auditable). As an all-in-one collaboration platform, Feishu has a natural advantage here — all work data (messages, docs, calendars, tasks) lives in one system, so the agent doesn't need to stitch together context across disconnected tools.

Team-Shared Agents and Multi-Agent Collaboration

The other two capabilities deserve attention too. Team-shared agents mean a team can jointly create, train, and use a shared agent — for example, HR training a “recruiting assistant” that understands the company's interview process and evaluation criteria, or sales training a “lead follow-up assistant” that auto-triages and follows up on leads. These agents are no longer personal tools; they're shared digital employees for teams.

Multi-agent collaboration goes one step further. A complex business process often requires multiple roles — taking a product requirement from proposal to launch might involve a PM agent breaking down requirements, a designer agent generating prototypes, an engineer agent estimating effort, and a PMO agent scheduling and tracking. The new aily supports multiple agents dividing work and collaborating, like a virtual project team.

Feishu aily's upgrade forms an interesting contrast with OpenAI Presence, announced just yesterday: Presence focuses on external-facing customer service and high-risk workflows, while aily focuses on internal collaboration and knowledge work; Presence is a heavyweight product deployed by OpenAI Forward Deployed Engineers, while aily is a lightweight tool embedded directly in Feishu, ready out of the box. Different directions, but both point to one trend: AI agents are shifting from “personal productivity tools” to “organizational infrastructure.” DingTalk Wukong, WeCom AI Suite, and WPS AI are all in the same race, and competition in enterprise collaboration agents will intensify through H2 2026.

See you tomorrow.