GenAI travel chatbot delivers 90% accuracy for major Australian airline

Project Summary

Avec partnered with a major Australian airline to reduce internal support overhead and enhance the employee experience using Generative AI. By embedding a smart travel policy chatbot within Microsoft Teams, the solution provided instant, accurate answers to common staff travel queries, eliminating manual triage and reducing wait times. Built using Microsoft Copilot Studio and Azure AI Search, the proof-of-concept demonstrated that enterprise-grade GenAI can meaningfully reduce workload, increase satisfaction, and scale with minimal maintenance. The project also validated a clear path to broader rollout, setting a benchmark for AI-powered self-service tools in large organisations.

This sounds familiar

The challenge

The airline’s internal Team Travel unit was inundated with routine queries about staff travel entitlements, policy rules, and edge case scenarios. These enquiries, received via email and social platforms, required manual responses, often involving time-consuming searches through complex policy documents. This process was repetitive, inefficient, and unsustainable at scale.

Previous chatbot initiatives had failed to gain traction due to:

  • High administrative overhead for training models using traditional machine learning methods.
  • Poor-quality responses that eroded trust in the solution.
  • Limited capacity within business units to maintain and update the chatbot systems.

The solution

The Avec team was engaged to design and deliver a proof-of-concept solution that would validate the feasibility of using Generative AI to automate internal staff queries at scale. In response, Avec implemented a conversational AI chatbot embedded within Microsoft Teams, with the following architecture:

  • Azure OpenAI GPT-4o accessed via Copilot Studio, using Retrieval-Augmented Generation (RAG) to answer queries directly from source policy documents.
  • Azure AI Search to semantically index policy documents in PDF and DOCX formats.
  • SharePoint as the central policy repository, managed by the client.
  • Power Automate to escalate unresolved queries via email.
  • Full chatbot orchestration through Copilot Studio, including persona design, logging, analytics, and query flow management.

Key features of this solution demonstrated:

  • Instant, one-click launch from within Teams, with 24/7 availability.
  • Context-aware conversations, retaining thread memory across up to 10 prompts.
  • AI-powered exception handling and escalation mechanisms.
  • Custom chatbot persona tailored to the airline’s brand identity.
  • Full traceability with zero data retention outside the airline’s Microsoft environment.
  • Near-zero maintenance, with new documents auto-indexed upon upload.
  • A clear roadmap for scaling and production rollout.

The difference

Avec’s solution significantly diverged from traditional chatbot implementations by leveraging Generative AI capabilities through Microsoft’s Copilot Studio. This allowed the team to deliver a high-quality conversational experience without the need for extensive model training or constant system maintenance.

Additionally, the solution:

  • Required no training on internal data, avoiding complex and resource-heavy preparation.
  • Used semantic search to deliver accurate, explainable answers from indexed documents.
  • Integrated seamlessly into Microsoft Teams, a platform already embedded in daily staff workflows.
  • Provided a clear technical path to production deployment and scaling.

The results

Through the solution provided, the airline saw measurable improvements in both performance and user experience during the proof-of-concept phase. These included:

  • Over 90% accuracy in answering all test questions.
  • 93% question completion rate, exceeding client expectations.
  • 21 out of 28 success criteria fully met, with the remaining 6 considered feasible in future iterations.
  • Response times under 5 seconds, ensuring strong user engagement and usability.

Furthermore, this project proved that GenAI, when thoughtfully configured through Microsoft Copilot Studio and Azure AI Search, can serve as a robust internal support tool. It not only reduces operational overhead but also significantly enhances employee experience.

Avec now uses this successful PoC as a scalable blueprint for deploying similar AI-powered solutions across industries.

< Back to case studies

Sound familiar?

If this sounds like a solution that would enhance your business then please don’t hesitate to get in touch below.