解决企业中人工智能自信出错的问题（由 PromptQL 赞助）.pdf

1、 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Tanmai GopalCEO and Co-founder,PromptQLFixing AIs Confidently Wrong problem in the enterpriseA I M 2 6 9-S/3 Whats our current T1 WIP inventory at Seven Oakswarehous

2、e?Pull up patient readmission rates for DRG 470this quarter.Run SKU analysisfor our private labelbeauty products.Will your AI on data be able to answer/4Even if your AI cant answer,will that stop it from being confidently wrong?/5Whats missing?Better models?/6“The reason humans are so useful is not

3、mainly their raw intelligence.Its their ability to build up context,interrogate their own failures,and pick up small improvements and efficiencies as they practice a task.”Dwarkesh Patel/7Whats missing?Better models?Ability to continually absorb tribal knowledge and improve.DATA TEAMTeaching loopWha

4、t does PromptQL do?pUSERUSERUSERUSERUSERUNSTRUCTURED DATASTRUCTURED DATAWEBAPIsSAASData analystteaches the AIData scientistteaches the AIData engineerteaches the AIData analystteaches the AIData scientistteaches the AIData engineerteaches the AIPromptQLMultiple data sourceswithout moving dataBullet-

5、proof data security,based on end-user identityService Desk10:1 100 :1USERSDATA TEAMPROMPTQLUSERS1 :1Service DeskUSERUSERUSERUSERUSERUNSTRUCTURED DATASTRUCTURED DATAWEBAPIsSAASData analysteg:BI,analysisData scientisteg:predictive modellingData engineereg:data-prepData analysteg:reportingData scientis

6、teg:sentiment analysisData engineereg:unstructured-structured dataThe only way to scale AI accuracy is if it can be trusted to:1.Know when its wrong2.Continuously improve upon its failuresKnows when its unsureAsks your expertsLocks in the learning for everyone