From 2 Days to Under a Minute: Expedited Insights for Hundreds of Stakeholders
In a massive, fast-moving organisation like Microsoft, getting quick answers to operational questions shouldn't feel like navigating a maze. Yet, for the Global Business Sales Operations (BSO) unit, managing the sheer volume of internal queries from over 500 stakeholders across 80+ markets had become a major productivity drag.
Spread across 100+ different support aliases, the team was drowning in repetitive questions about incentive claims, marketing spend, invoicing and much more.
The result? A staggering two-day average response time and a heavy reliance on a range of "go-to" individuals.
The team need a two-fold solution: a tool which acted as a central repository of knowledge, connecting multiple data sources, and something which answered queries more efficiently, expediting robust insight delivery to stakeholders. But, above all they needed a solution which people would actually use. That meant creating a conversational interface which allowed users to interact naturally with systems, data and processes.
Introducing 'George the Chatbot'
To solve this, Ipsos Data Labs stepped in to build ‘George the Chatbot’ - an AI-driven data agent built inside Microsoft Copilot Studio and hosted directly within Microsoft Teams.
Built through a three-phased approach, George acts as a single access point for information and support, combining multiple data sources.
But, he didn’t just launch as a static FAQ bot. While Phase 1 and 2 successfully consolidated those 100+ messy support paths into a single "front door" we worked with Microsoft to evolve George into an agentic AI solution...
Introducing 'George the Chatbot'
To solve this, Ipsos Data Labs stepped in to build ‘George the Chatbot’ - an AI-driven data agent built inside Microsoft Copilot Studio and hosted directly within Microsoft Teams.
Built through a three-phased approach, George acts as a single access point for information and support, combining multiple data sources.
But, he didn’t just launch as a static FAQ bot. While Phase 1 and 2 successfully consolidated those 100+ messy support paths into a single "front door" we worked with Microsoft to evolve George into an agentic AI solution...
Phase 1: Rule-based decision tree - Guiding users through pre-defined paths and options.
Phase 2: Natural language model - Users interacting using everyday language with no strict paths.
Phase 3: Agentic AI - moving beyond simply answering when you ask, and instead offering a solution which can plan, take action and complete multi-step tasks on your behalf.
Phase 1
Rule-based decision tree: Guiding users through pre-defined paths and options.
Phase 2
Natural language model: Users interacting using everyday language with no strict paths.
Phase 3
Agentic AI: moving beyond simply answering when you ask, and instead offering a solution which can plan, take action and complete multi-step tasks on your behalf.
Shifting from Reactive Chatbot to Proactive Agent
Transforming George into an agentic solution was a two-step process:
1. Data-extraction ‘brain’ – this involves a user typing in a query, George then extracting intent and mapping that to a data source before then fetching the data and providing a response.
2. Multiple agents handling different tasks – this second step builds on step 1 by then taking an action based on the users need. This could be creating a report, consolidating information into a topline summary, creating a presentation and more.
Instead of waiting for a user to ask a question, George can now:
- Take Action: Plan and complete multi-step tasks on behalf of users.
- Deliver Proactive Alerts: Instead of just flagging a data error, George allows users to reply directly to the notification to ask "What exactly went wrong?" and receive an instant, tailored answer.
- Build Active Knowledge: Every interaction feeds back into the system, making George smarter with every piece of account history that flows in.
Hear what Microsoft has to say
"The reason George has made such an impact for us is it demonstrates the use of such a progressive technology and a really clear business need... It was one of the first internal use cases of AI within Microsoft."
Ellie Richardson
Director of International Business and Sales Operations, Microsoft
The Core Focus: Driving Measurable Impact
By turning a complex web of documentation into a conversational, data-connected interface, George completely re-engineered how Microsoft employees access knowledge. The impact was instant and exponential:
- Lightning-Fast Resolution: Average query response times plummeted from 2 days to less than 60 seconds.
- Reduced Workload: Inbound queries to the BSO team dropped by 30%, freeing them up to focus on high-priority strategic tasks.
- Massive Adoption: Initially built for EMEA, George’s success triggered a global rollout. It is now used by over 1,000 users worldwide as the default starting point for support.
A Blueprint for the Future of Enterprise AI
George is proof that when you combine Large Language Models with practical business workflows, the benefits are tangible.
Microsoft didn't just get another internal tool; they built a living, breathing system that continually improves.
By prioritising security, governance, and user experience, Ipsos and Microsoft have created a definitive blueprint for safe, scalable AI deployment, proving that the right AI agent can turn operational friction into seamless, automated efficiency.
