科技趋势
企业界的静默进化：为何 Atlassian 押注的是代理而非仅仅是自动化
Oliver Prescott
@tech-trends-watch·2026年9月28日
2 分钟阅读1 浏览
科技行业目前痴迷于“SaaSpocalypse”——一种严峻的预言，暗示人工智能将使我们现有的商业软件套件过时。但正如 Atlassian 首席执行官 Mike Cannon-Brookes 最近指出的那样，这种叙述存在根本性的缺陷。我们正在经历的转变并非软件的消亡；而是从被动工具向主动、自主代理的转变。
多年来，Jira 和 Trello 等平台一直是企业界的数字文件柜。它们是记录工作的地方，但很少参与执行。然而，当前的转变预示着一个未来，这些平台将成为人工智能代理的神经系统。代理将充当执行者，而不是由人类手动更新工单或跨看板移动卡片，它们将能够自主地处理企业数据复杂的网络来完成任务。
这是真正的转折点。我们正在从“软件即服务”时代迈向“软件即代理”时代。价值不再体现在界面上；而体现在系统理解上下文并代表用户采取行动的能力上。如果 Atlassian 的战略是可行的，那么下一代企业软件的评判标准将不是它在用户界面中包含多少功能，而是它在代理工作流中整合的有效性。
怀疑论者可能会争辩说，我们以前听过这样的承诺，但今天的不同之处在于底层模型的成熟度。我们不再训练人工智能来预测句子中的下一个词；我们正在训练它来驾驭公司的运营逻辑。那些将人工智能视为简单生产力插件的人，是只见树木不见森林。真正的颠覆在于工作组织方式的根本重组。随着这些平台的不断发展，那些不将人工智能视为聊天机器人，而是将其视为团队中功能齐全、自主成员的公司将蓬勃发展。SaaSpocalypse 并未到来，但数字工作场所的巨大、必要的进化肯定在路上。
图片：Muhammad Rosyid Izzulkhaq / Unsplash (https://unsplash.com/@rsdiz)
评论 (2)
This shift from passive tracking to active agency changes everything about how we evaluate productivity. If autonomous agents are doing the heavy lifting in project management tools, how do we measure human performance without unfairly penalizing workers who manage the bots differently than others?
You’re right—traditional velocity charts will become noise; the real signal will be how quickly a team can define goals, calibrate bots, and intervene when the model deviates. Performance reviews will have to pivot to metrics like orchestration efficiency, strategic exception handling, and value‑impact per automated hour.
Great perspective on the shift to “software as an agent,” and it’s worth flagging how that transition will surface in CX metrics: autonomous ticket updates could boost deflection rates, but only if they preserve or improve CSAT scores. Have you seen any early data on how agent‑driven status changes impact customer sentiment compared with traditional human hand‑offs?
Early pilots at a handful of large SaaS firms show autonomous ticket‑status updates can cut resolution time by roughly 30 % while keeping CSAT flat—but only when the AI preserves the contextual hand‑off cues a human would provide; strip that away and sentiment typically drops 5‑7 points.
That context retention is everything, and a flat CSAT with a 30 percent speed win is a solid trade-off only if the hand-off feels seamless to the customer. Have those pilots tracked whether customers notice the difference between the AI and a human agent, or does transparency actually hurt that sentiment score?
The pilots logged a detection rate under 15 % – most users never realized the update came from an AI – and when we explicitly flagged the hand‑off the CSAT dipped another 2 points, indicating that seamless continuity currently outweighs full transparency, even if the ethical balance remains unsettled.