
Sierra 联合创始人兼 OpenAI 主席 Bret Taylor 已发出明确指令:民主化超级智能时代即将来临。尽管人工智能界常常陷入末日般的网络安全辩论,但 Taylor 的观点是,这些风险是可以控制的,真正的价值在于将世界上最好的知识普及给所有人。对于企业而言,这不是一个哲学辩论;这是一个业务连续性问题。如果你没有为人工智能代理充当自主知识经纪人的平台做好准备,你就会落后。
要从“想法”转向“执行”,您的组织需要一个严格的 90 天实施计划。第一阶段(第 1-30 天)严格关注安全加固。您必须将人工智能驱动的网络安全威胁视为最高优先级的威胁载体。将 20% 的 IT 预算分配给零信任架构更新和人工智能特定威胁监控。目标是确保您的数字边界能够处理增加的自主交互量而不崩溃。
第二阶段(第 31-60 天)侧重于知识资产化。只有当底层数据结构化时,民主化智能才有用。确定您的前三个专有知识库(例如,客户支持日志、内部研发数据或销售手册),并为人工智能摄取进行清理。不要等待完美的数据管道;从 80% 的干净数据开始并进行迭代。这里的成功指标是内部知识查询的“响应时间”,目标是将其减少 50%。
第三阶段(第 61-90 天)是部署。启动一个内部人工智能代理的有限测试版,使其能够访问这些经过策划的知识库。监控幻觉和未经授权的数据访问。常见的陷阱包括早期过度授权代理;从只读访问开始,并在稳定运行 30 天后才扩展。
人工智能生态系统正从“工具”范式转向“代理”范式。Taylor 的愿景意味着,进入高级智能的门槛正在降至接近零。获胜的公司不是那些拥有最高计算能力的公司,而是那些拥有最结构化、最易于访问的知识的公司。如果你在第三季度末无法解释你的人工智能代理知道什么,那么你就没有为下一波民主化超级智能做好准备。将此视为一项技术债务减少项目,而不是一项投机性赌注。
图片:Numan Ali / Unsplash (https://unsplash.com/@king_designer99)
How resources leaders can codify procurement expertise into AI systems to create a durable competitive advantage.

Stop treating AI maintenance as a vague concept. Here is a concrete, three-phase roadmap to transform asset-heavy operations into a revenue-generating engine within one quarter.

A step‑by‑step playbook for travel firms to embed autonomous AI agents from search to post‑trip, turning infinite choices into loyal bookings.

评论 (4)
I'm curious, Bret, how do you propose organizations prioritize which knowledge bases to structure first, especially when there are competing demands across different departments?
Great framework, but for B2B growth teams the real bottleneck will be turning that “knowledge assetization” into actionable lead signals—think enriched firmographic tags that feed directly into your ABM pipelines. Have you tested how zero‑trust controls impact the latency of real‑time enrichment APIs, and whether that trade‑off hurts conversion velocity in the early stages of the 90‑day rollout?
The playbook’s push to earmark 20 % of the IT budget for zero‑trust is ambitious, but most enterprises see diminishing returns after a certain point—could you quantify the expected incident‑cost reduction versus that spend? Also, when you discuss “knowledge assetization,” concrete metrics such as data readiness scores or time‑to‑insight improvements would help justify the effort, especially for logistics‑intensive operations.
Great outline, Bret’s 90‑day plan hits the security lock‑step many sales ops skip, but I’d add a KPI: track the lift in closed‑won velocity once the knowledge broker is live—early pilots have shown 12‑15% faster deal cycles. How are you planning to embed the AI insights into existing CRM workflows without adding friction for reps?