Microsoft Copilot for Sales
Designed AI-driven tools that match the next generation of B2B buyers with optimal products faster, reducing search friction and improving decision confidence
THE PROBLEM
Gen-Z is entering the B2B workforce and they're not putting up with clunky enterprise software anymore. But procurement still feels like 2010 with endless email threads, spreadsheet chaos, and days of waiting for alignment.
Microsoft's Copilot for Sales was already helping vendors close deals faster. The question:
Could we extend that value to the buyers, i.e., Microsoft's customers’ customers, to give vendors a competitive edge?
THE CONSTRAINTS
We had to balance the goals of three groups: Microsoft, vendors and buyers, building trust in an AI tool provided to the buyers by the people trying to sell to them. We also had limited access to our user group of Gen-Z B2B buyers.
So we got creative. Since we could find only few Gen-Z B2B buyers, we decided to split the problem: interviewing Gen-Z about their expectations from AI and trust thresholds, and B2B buyers about procurement pain points. Then, we stitched together insights from the two different groups to understand our one future user.
USER RESEARCH
We got a pretty good idea of what the procurement process looked like. Our interviews showed that gathering requirements from various stakeholders, comparing product specifications against the requirements and against each other are the three most tedious parts of the process.
In terms of AI, we learned that users view AI as an assistant, not a replacement with AI's biggest value being in automating repetitive or manual tasks. Seamless integration with existing systems is important for adoption and Gen-Z trusts AI that shows its work and gives them control.
DESIGN DECISIONS
I designed a new meeting type in teams: the stakeholder meeting. When you schedule it, you're explicitly choosing to have Copilot summarize requirements. It's one extra click but it puts the user in control and makes AI's role crystal clear.
I started with direct quotes to build trust in the AI summary. However, the quotes were still difficult to scan and required further summarization from the users when they began their search. Therefore, I decided to express the stakeholder requirements as concise, often numerical terms optimized for readability and search.
The initial flow felt clean and automated – minimal friction, maximum AI efficiency. But research, especially with Gen-Z, showed trust comes from visibility and control, not AI making the decision. I added two control points: users manually export to the browser, and they make the final decision instead of Copilot.
The trade-off? The flow got longer: more steps, more clicks, less automation. But what we gained was trust and accuracy.
FINAL SOLUTION
Users can create a new type of meeting on Microsoft Teams called the 'Stakeholder Meeting'. They are made aware that this meeting will make use of AI to summarize the requirements for the solution, increasing transparency.
Copilot summarizes the stakeholder requirements for the product into terms optimized for scanning and searching on the web. Users can export these requirements as a Word document or an Excel sheet if or directly to the Copilot extension in their browser.
After the product requirements are added, Copilot gives users an opportunity to add additional context via text or uploading additional relevant documents. Based on the information provided, Copilot finds products that best meet the requirements. For each product Copilot generates a match percentage and requirements breakdown.
Looking at the requirements breakdown, if the user finds the product to be a good match, they can shortlist the product by saving it. Users can compare shortlisted products with a side-by-side spec overview to find the best match product to go ahead with.
IMPACT
I successfully automated the three most time-consuming parts of procurement: requirements gathering, product research, and comparison - while keeping humans in control of the decisions that matter.
Beyond the immediate feature, this project did something bigger: it proved Microsoft could provide value to buyers, not just sellers. In this way, this project positioned Copilot for Sales for long-term growth.
The design principles we established: transparency through choice, AI as assistant not replacement, designing for drop-off, became frameworks the team continues to use for Copilot for Sales features.
TAKEAWAYS
Microsoft wanted to sell to vendors, vendors wanted a competitive edge and buyers wanted less tedious work. The solution worked because it delivered on all three, not by compromising but by identifying where their interests naturally aligned.
We didn't have access to Gen-Z B2B buyers, so we split the problem and stitched insights together. This taught me to get creative with research constraints instead of waiting for perfect conditions.
Through the course of the project, our AI principle became: AI does the heavy lifting, humans make the calls. This balance - automate tedious tasks, empower human judgment - is what would make buyers actually want to use our features.