AI Summary
Generated by AI. Be sure to check for accuracy.
Meeting notes:
- Microsoft Support's AI Transformation Journey: Steve Kerr and Arielle Lupo detailed Microsoft's evolution from a fragmented support system to a unified, AI-powered support organization, highlighting the consolidation onto Dynamics, the introduction of intelligent routing, and the deployment of generative and agentic AI features to enhance both customer and support engineer experiences.
- Support System Consolidation: Steve explained that prior to 2020, Microsoft Support operated with 16 different CRM systems and 500 tools, leading to inefficiencies such as manual case transfers. In 2020, the team migrated all support operations to Dynamics, centralizing data and streamlining processes, which enabled universal skilling and easier case transfers.
- Intelligent Routing Implementation: Steve described the deployment of an AI-powered intent agent for call routing, which interprets customer issues in 62 languages and routes them to the appropriate team. This system reduced misroutes by 20% by training on categorized call transcripts and now handles 91% of calls through the IVR, significantly improving efficiency.
- Generative Copilot Features: The team introduced generative AI features such as case summarization and email generation to reduce repetitive tasks for support engineers. These features provide instant case overviews and draft empathetic, up-to-date communications, allowing engineers to focus more on problem-solving rather than administrative updates.
- Agentic AI Capabilities: Steve highlighted the latest advancements, including autonomous agents that can create knowledge articles, perform quality reviews, and automate administrative tasks. These agents continuously learn from interactions and are integrated into the support workflow to further reduce manual effort and improve service quality.
- Enhancing Support Engineer Productivity with Copilot: Arielle Lupo demonstrated how Copilot's generative and agentic features within Dynamics 365 streamline support engineer workflows, from case summarization and Answer Assist to automated knowledge management and administrative task handling.
- Case Summarization Utility: Arielle showed how the case summary feature in Dynamics 365 automatically compiles all relevant case information—such as chat transcripts, emails, and internal notes—into a concise, structured summary. This aids in rapid onboarding during case transfers and improves documentation quality, serving as a reliable baseline for case notes.
- Answer Assist for Troubleshooting: Arielle explained the Answer Assist tool, which allows support engineers to query Copilot for solutions based on internal and customer-facing documentation. The tool generates insights, provides source references, and consolidates information from multiple articles, reducing the time spent searching for answers and improving response quality.
- Automated Knowledge Article Creation: Arielle demonstrated the Knowledge Management Agent, which detects when a case lacks existing documentation and auto-generates a draft knowledge article based on the case resolution. This draft is then reviewed and approved by a secondary approver before being published, ensuring quality and relevance in the knowledge base.
- Administrative Task Automation: Steve described how agentic features automate routine administrative tasks, such as sending follow-up emails after 48 hours of customer inactivity and auto-filling closing notes with case details. These automations save time for support engineers and ensure consistency in customer communications.
- Continuous Improvement Through Feedback and Training: Arielle Lupo outlined the strategies used to improve Copilot's effectiveness, including large-scale knowledge base cleanup, targeted training for support engineers, and a robust feedback loop leveraging direct input from experienced staff.
- Knowledge Base Cleanup: Arielle described an eight-month initiative to reduce the knowledge base from one million to 400,000 articles by retiring outdated content, consolidating articles, and reformatting for AI readiness. This cleanup was critical to improving the accuracy and relevance of AI-generated responses.
- Support Engineer Training: The team developed training modules and events, such as the 'Be an Expert Core Edition' and Spark Weeks, to teach support engineers effective prompting techniques and build confidence in using Copilot. These efforts helped transition engineers from using keyword searches to more conversational, effective interactions with AI tools.
- Feedback Loop Integration: Arielle emphasized the importance of the thumbs up/down feedback mechanism, which collects actionable feedback from support engineers. This feedback is triaged by knowledge custodians—rotational roles filled by senior engineers—who ensure that model improvements and documentation updates are implemented promptly, often within 48 hours.
- Impact on Customer Satisfaction and Operational Metrics: Steve Kerr presented data showing significant improvements in customer satisfaction, operational efficiency, and cost savings resulting from the adoption of AI-powered support tools, including reductions in handle time, increased case throughput, and substantial cost savings.
- Customer Satisfaction Gains: Steve reported a 12% average increase in customer satisfaction scores (CSAT), with survey targets rising from 4.6 to 4.8 out of 5. Faster problem resolution and improved support experiences contributed to these gains.
- Efficiency and Cost Savings: The adoption of AI tools led to a 12–16% reduction in average handle time for chat cases, a 9–12% increase in case volume handled, and a 13% reduction in peer assistance required. Headcount was reduced from 45,000 to 35,000 support engineers, resulting in $500 million in annual support cost savings.
- Automation of High-Volume Tasks: AI automation achieved a 90% automation rate for Microsoft Account Recovery Form reviews, reducing human review from 600,000 to 60,000 cases per month and saving over $1 million annually. Average customer wait times for account recovery dropped from 12 hours to 31 seconds.
- Best Practices for Introducing AI in Contact Centers: In response to Jeffrey's question, Steve and Arielle Lupo recommended starting with case summarization as an initial AI feature due to its simplicity and impact, and advised involving support engineers in the adoption process and feedback loops for continuous improvement.