Cepheid by Danaher: AI-Powered Migration Support Agent

Introduction

Every major migration comes with a predictable aftermath: an avalanche of support tickets. "How do I access this?" "Where did my files go?" "What are the new security rules?" IT teams spend weeks firefighting the same questions from hundreds of users. Help desk queues overflow. Users get frustrated waiting for answers. And the very people who should be optimizing the new environment are instead buried in repetitive support work. There's always been an assumption that this is just the cost of change—until it isn't.

Client Background

Cepheid, a Danaher company and global leader in molecular diagnostics, had just completed a massive multinational Teams migration and data governance transformation with our team (see related case study). This wasn't a simple technology upgrade—it represented a fundamental change in how thousands of employees across EMEA, Americas, and other regions would collaborate, access data, and follow new governance protocols.

We knew from experience that even the most successful migration creates a surge in user support needs. New interfaces, changed workflows, updated security policies, and unfamiliar governance rules inevitably generate questions. With Cepheid's scale—thousands of users across multiple continents and time zones—the impending support burden threatened to overwhelm their IT help desk and delay user adoption of the new environment.

Having established deep trust through our previous work, Cepheid engaged us to explore an innovative approach: could we use AI to proactively deflect support tickets before they ever reached IT, while simultaneously empowering users with instant, 24/7 access to answers?

The Challenge

Cepheid faced the classic post-migration support dilemma, amplified by organizational scale and complexity:

Anticipated Support Surge:

• Thousands of users transitioning to new Teams environment simultaneously
• Predictable spike in "how-to" questions following go-live
• IT help desk resources finite and already allocated to critical issues
• Risk of support backlog creating frustration and slowing adoption

Knowledge Accessibility Gap:

• Comprehensive training materials created but scattered across documents
• Users uncertain where to find specific answers to their questions
• No single entry point for self-service support
• Training documentation not searchable in intuitive, conversational way

Time Zone & Availability Constraints:

• Global operations meant users needed support across all time zones
• IT help desk operating during limited hours
• Users facing urgent questions outside support hours left without answers
• Delayed responses impacting productivity and migration success

Repetitive Question Burden:

• Many support tickets asking identical or similar questions
• IT staff spending time answering questions already documented
• Knowledge transfer inefficient—same answer given repeatedly to different users
• High-value IT resources diverted from complex issues to routine queries

Governance Complexity:

• New environment had specific, custom governance rules
• Security protocols different from previous environment
• Access request processes changed
• Users needed immediate clarity on compliance requirements

Adoption Risk:

• User frustration with new environment could undermine migration success
• Lack of immediate answers might push users toward workarounds or non-compliance
• Poor support experience could damage perception of new platform
• Slow question resolution delaying productivity gains from migration

Goals

Cepheid sought a proof-of-concept solution that could validate a fundamentally different approach to migration support:

Primary Objectives:

Ticket Deflection:

• Reduce anticipated support ticket volume by enabling user self-service
• Provide instant answers to common "how-to" questions
• Free IT help desk to focus on complex, high-value issues
• Prevent support queue bottleneck that could delay adoption

User Empowerment:

• Give users ability to find answers independently, on-demand
• Eliminate waiting time for routine questions
• Build user confidence in new environment through accessible guidance
• Create positive migration experience driving adoption

24/7 Availability:

• Provide support access across all time zones without staffing increases
• Enable users to get help outside IT help desk hours
• Ensure no user blocked waiting for business hours to resolve simple questions

Knowledge Accessibility:

• Create intuitive, conversational interface to training materials
• Make comprehensive documentation instantly searchable
• Connect users to exact relevant sections of training guides
• Transform static documentation into interactive support resource

Custom Training:

• Train AI specifically on Cepheid's governance and security rules
• Ensure answers reflected actual organizational policies, not generic guidance
• Maintain accuracy and relevance for Cepheid's specific environment
• Support custom access request and compliance processes

Proof-of-Concept Validation:

• Demonstrate technical feasibility within rapid timeline (1 month)
• Validate user acceptance of AI-powered support
• Measure impact on ticket volume and user satisfaction
• Build business case for full production deployment

Strategy & Approach

Our strategy balanced rapid proof-of-concept delivery with validation of production viability:

Design Philosophy: Intelligent Triage, Not Full Automation

Rather than attempting to automate all support functions, we positioned the AI agent as intelligent first-line triage:

• Handle common, well-documented questions instantly
• Guide users to specific training materials for detailed explanations
• Recognize when questions require human IT intervention
• Reduce ticket volume without eliminating human support entirely

Knowledge Foundation First

We recognized that AI quality depends entirely on training data quality:

• Centralized all training materials into structured SharePoint library
• Organized documentation for AI accessibility and human navigation
• Created "single source of truth" for migration guidance
• Ensured documentation comprehensive before AI training

Custom Training on Organizational Context

Generic AI couldn't answer Cepheid-specific questions accurately:

• Trained agent specifically on Cepheid's governance rules
• Included organization-specific security protocols
• Documented custom access request processes
• Embedded understanding of Cepheid's Teams environment structure

Conversational Interface Design

Users needed to ask questions naturally, not learn query syntax:

• Designed conversational AI understanding varied question phrasings
• Enabled follow-up questions for clarification
• Provided both quick answers and links to detailed documentation
• Created intuitive user experience requiring no training

Rapid Proof-of-Concept Approach

One-month timeline demanded focused scope and efficient execution:

• Prioritized most common anticipated question categories
• Built core functionality first, demonstrated extensibility
• Used proven AI platforms rather than custom development
• Focused on validating concept, not building production-scale system

Execution

Duration: Less than 1 month (proof-of-concept)

Phase 1: Knowledge Organization & Documentation (Week 1)

Training Material Consolidation:

• Created central SharePoint document library as knowledge repository
• Migrated all end-user training materials into structured organization
• Categorized content by topic (governance, security, access, workflows, etc.)
• Ensured documentation comprehensive and current

Content Structuring:

• Organized materials for both AI consumption and human navigation
• Created clear navigation hierarchy
• Standardized formatting for AI parsing
• Indexed key concepts and questions each document addressed

Gap Analysis:

• Identified common questions not yet documented
• Created supplementary guidance for anticipated gaps
• Validated coverage of all critical migration topics
• Prepared FAQ section addressing most common expected queries

Phase 2: AI Agent Development (Weeks 2-3)

Platform Selection & Configuration:

• Selected appropriate AI/chatbot platform supporting custom training
• Configured integration with SharePoint document library
• Established secure authentication and access controls
• Set up conversation logging for analysis

Custom Training:

• Trained AI agent on Cepheid's specific governance rules and policies
• Loaded security protocol documentation
• Included access request processes and approval workflows
• Embedded understanding of new Teams environment structure

Answer Logic Development:

• Programmed agent to provide direct answers for simple queries
• Configured intelligent linking to detailed documentation sections
• Implemented fallback to "contact IT" for complex or unclear questions
• Created response templates maintaining helpful, professional tone

Integration Development:

• Connected AI agent to SharePoint training library
• Implemented dynamic content retrieval from documentation
• Enabled agent to reference and link specific document sections
• Ensured real-time access to updated materials

Phase 3: Testing & Demonstration (Week 4)

Internal Testing:

• Conducted extensive testing with common question scenarios
• Validated answer accuracy against actual policies
• Tested edge cases and unusual question phrasings
• Refined training based on gaps discovered

User Experience Testing:

• Gathered feedback from pilot user group
• Assessed conversational flow and answer helpfulness
• Validated documentation links worked correctly
• Measured time-to-answer for common questions

Performance Metrics:

• Tracked question types and frequency
• Measured answer accuracy rate
• Recorded user satisfaction ratings
• Analyzed conversation logs for improvement opportunities

Stakeholder Demonstration:

• Presented working proof-of-concept to Cepheid leadership
• Demonstrated handling of common support scenarios
• Showed integration with training materials
• Outlined potential impact on ticket volume and user experience

Key Deliverables

1. AI Support Agent (Proof-of-Concept)

• Conversational AI interface accessible to all users
• Custom-trained on Cepheid governance, security, and migration guidance
• Integrated with SharePoint training library
• 24/7 availability across all time zones

2. Centralized Training Library (SharePoint)

• Consolidated repository of all migration training materials
• Structured organization supporting both AI and human access
• Comprehensive coverage of governance, security, access, and workflows
• Maintainable foundation for ongoing updates

3. Answer Coverage

• Instant responses to common questions:
  • "How do I use the new Teams environment?"
  • "How do I request access to a site?"
  • "What are the security protocols I need to follow?"
  • "Where did my files go after migration?"
  • "How do I find documents in the new structure?"
  • "What governance rules do I need to know?"
  • Plus dozens of variations and related queries

4. Intelligent Documentation Routing

• Direct answers for simple questions
• Guided navigation to specific training document sections for complex topics
• Context-aware linking ensuring users reached relevant content
• Reduced information overload by surfacing only applicable guidance

5. Analytics & Insights

• Question tracking revealing common user concerns
• Answer accuracy metrics validating AI performance
• User satisfaction measurements
• Data supporting business case for full deployment

Results & Outcomes

Proof-of-Concept Success:

• Rapid Delivery: Functional AI support agent delivered in under 1 month
• Technical Validation: Demonstrated AI could accurately answer Cepheid-specific questions
• User Acceptance: Positive feedback from pilot user testing
• Business Case Proven: Clear demonstration of potential ticket deflection value

Projected Impact (Based on POC Testing):

• 50-70% Ticket Deflection Potential: Majority of common "how-to" questions answerable by AI
• Instant Response Time: Users received answers in seconds vs. hours waiting for IT
• 24/7 Availability: Support accessible across all time zones without staffing increases
• IT Efficiency Gain: Help desk could focus on complex issues requiring human expertise

Knowledge Management Achievement:

• Single Source of Truth Created: All training materials centralized and accessible
• Improved Documentation Discoverability: Users could find relevant guidance conversationally
• Maintainable Foundation: Structure supported ongoing content updates and additions
• Dual-Purpose Library: Served both AI agent and users seeking direct documentation access

Strategic Insights:

• AI Training Quality Validated: Custom training on organizational context proved feasible and effective
• User Experience Confirmation: Conversational interface well-received by users
• Integration Success: SharePoint library integration enabled dynamic, current answers
• Scalability Demonstrated: Architecture could support full production deployment

Innovation Leadership:

• Early AI Adoption: Demonstrated forward-thinking approach to post-migration support
• User-Centric Design: Prioritized user empowerment over traditional support models
• Proactive Problem-Solving: Anticipated and addressed support surge before it occurred
• Strategic Technology Use: Applied AI to high-value, high-volume use case

Key Takeaways

Proactive Support Beats Reactive Firefighting: Every migration team plans for training and communication, but few plan for the inevitable support surge. By building the AI agent before go-live, we transformed anticipated problem into managed advantage. The best time to reduce ticket volume is before the first ticket arrives.

AI Quality Depends on Knowledge Organization: The AI agent was only as good as the training materials we fed it. Consolidating and structuring documentation wasn't a prerequisite for AI—it was the foundation of AI success. Organizations that want AI-powered support must first invest in knowledge management. The cleanup work creates value even if AI is never implemented.

Custom Training Transforms Generic Tools: General-purpose AI chatbots provide generic answers. By training specifically on Cepheid's governance rules, security protocols, and environment structure, we created a support agent that reflected organizational reality, not internet consensus. Custom training is the difference between "interesting demo" and "actually useful tool."

Proof-of-Concept Should Validate Business Model, Not Perfect Technology: We delivered in under a month because we focused on proving ticket deflection value and user acceptance, not building production-scale perfection. The POC succeeded because it answered the strategic question: "Will this approach work?" Everything else could be refined in production development. Fast validation beats slow perfection when testing new approaches.

‍