扩大销售卓越：Databricks 如何利用自己的技术培训 GTM 团队.pdf

1、Scaling Sales ExcellenceHow Databricks Uses Its Own Tech to Train GTM TeamsSergio Ballesteros Solanas|Senior Solutions Architect DatabricksIvy NgDatabricks Sales OffsiteWorkshop for 30 people:Technical discoveryPositioning of technologiesDatabricks featuresGroups?LLMDatabricks on Databricks:Talk Wit

2、h Ian!ContextLLM(Llama 3.1-Foundational Model Serving)Chatbot100%built using Databricks!Masked conversations from IanPrompt engineeringDiscover all pain pointsHandle objectionsConvince IanDatabricks APJ GTM Org400ICs5 RegionsProblem Statement“Were Scaling Fast and Facing Real Challenges”1.Were growi

3、ng from a lean team to 400 ICs across 5 regions.2.That scale brings complexity different markets,different languages,different needs.3.Our ICs need sharper discovery skills and stronger objection handling consistently4.Human based training is not scalable:No 24/7 coaching,no real-time feedback,and l

4、imited multilingual resources.1.Traditional training(videos,reading material),is not enoughThe Goal?To deliver scalable,always-on,multilingual enablement so every GTM person gets the support they need,exactly when they need it.Databricks APJ GTM OrgUsing an LLM to scale the training“You dont lose yo

5、ur job to AI;you lose your job to people who knows how to use AI.”Llama 3.1 didnt handle well Korean and JapaneseThere Were More Challenges Along the WayThere Were More Challenges Along the WayIan wanted a free dinnerPrompt enrichmentArchitecture of Talk With Ian13Databricks AppsGPT model(Azure Open

6、AI)Llama(Databricks)EvaluationAI GatewayAI Agent FrameworkDashboardAI GuardrailsAI PlaygroundQueriesQueriesReads from/Writes toTracking and loggingWrites toEvaluation metricsWrites toReads fromQueriesQueriesQueriesInjectsInjects14Model ServingModel ServingLogging:Usage tracking&Payload loggingCompli