Remember those days when finding the right answer in a sprawling knowledge base felt like an archaeological dig? For many support teams, especially in large tech companies, that's been the reality. Mountains of troubleshooting guides, countless articles, and an ever-growing list of new features meant that finding accurate, up-to-date information could be a real challenge. It wasn't uncommon for crucial documents to become outdated, leaving support engineers scrambling.
It's fascinating how quickly things can change, isn't it? Just a couple of years ago, the idea of using AI to sift through this digital labyrinth was more of a futuristic dream. But then, the landscape shifted dramatically. At Microsoft, for instance, a team within their Modern Solutions & Support (MSS) group had been exploring semantic search – a way to make their vast repositories more accessible. They envisioned a single point where engineers could ask a question and get the most relevant answers.
Then came the generative AI revolution, and suddenly, the possibilities exploded. The announcement of tools like ChatGPT felt like a seismic event, accelerating their efforts. "It was fascinating how fast things were changing," recalls DJ Ball, a senior escalation engineer on the MSS Supportability team. The pace of development, from GPT 2.0 to 4.0 almost overnight, presented both a challenge and a massive opportunity.
This team, recognizing the potential, quickly pivoted. They secured access to Microsoft Azure OpenAI and discovered an internal tool that was remarkably similar to what they were building. This serendipitous alignment made joining forces much smoother. "People thought we planned it, which we didn’t, but for once we felt we were ahead of the game," shared Sam Larson, a senior supportability PM. This collaboration really kicked their AI-based solution into high gear.
Since no one had quite tackled enterprise content with this new ChatGPT technology before, it was a journey of discovery. They learned by doing, experimenting, and providing constant feedback to the engineering team. This iterative process helped shape what eventually became Microsoft Azure AI Studio, a platform designed to simplify the integration of external data into the Azure OpenAI Service. This development allowed them to create a private chat workspace, aptly named Modern Work GPT (MWGPT).
Initially, they focused on curating content specifically for the Teams product, feeding it into the large language model. By leveraging Azure Cognitive Search to break down complex documentation into smaller, manageable chunks, they could effectively test the results with subject matter experts across the Teams support division. The scope has since broadened significantly, now encompassing all Modern Work Technology solutions – that's over 300,000 pieces of content for 34 different products! Along the way, they've gained invaluable insights into content curation, prompt engineering, and the inner workings of LLMs.
It quickly became clear that more hands were needed to rigorously test and refine this powerful tool. Today, the work continues, with ongoing efforts to update its capabilities and ensure unwavering accuracy. Mayte Cubino, now an AI strategy and programs director, was an early volunteer. As an engineer at heart, she was captivated by the buzz surrounding ChatGPT's potential to transform customer support.
From a support delivery perspective, the implications are profound. AI-driven tools can act as an intelligent assistant, not just for finding information, but for understanding context, suggesting solutions, and even drafting responses. This frees up human support engineers to focus on more complex, nuanced issues that require empathy and critical thinking. It's about augmenting human capabilities, not replacing them, leading to faster resolutions, more consistent answers, and ultimately, happier customers. The future of support is looking a lot smarter, and a lot more human, thanks to AI.
