This article was based on the interview with Kelly Soligon, VP of Consumer Digital Direct Sales at Microsoft by Greg Kihlström, AI and MarTech keynote speaker for The Agile Brand with Greg Kihlström podcast. Listen to the original episode here:
For years, the promise of one-to-one personalization at scale has been the holy grail of digital marketing. We, as leaders, have sat through countless presentations and read innumerable whitepapers on the subject. We’ve invested in platforms, hired data scientists, and run pilot programs. Yet, for many enterprise organizations, personalization has often remained a “sidecar”–an interesting set of experiments running parallel to the core business, but never fully integrated into the engine itself. The sheer volume of data, customer signals, and potential touchpoints made true, dynamic personalization at the scale of a global enterprise feel more like a theoretical ambition than an operational reality.
But the ground is shifting beneath our feet. The rapid maturation of artificial intelligence is transforming this ambition into a tangible, strategic imperative. It’s no longer just about optimizing a landing page or serving a recommended product. It’s about fundamentally re-architecting the customer experience to be more intelligent, responsive, and, frankly, more human. At the helm of Microsoft’s global consumer e-commerce business, which sees over 10 billion annual visitors across 180 markets, Kelly Soligon is navigating this transformation on a scale most can only imagine. Her experience provides a clear-eyed view not of the hype, but of the practical application of AI to solve complex D2C challenges—from making sense of millions of customer feedback points to preparing for the coming wave of “agentic commerce.”
Making Personalization a Core Competency
The most significant shift in leveraging AI for the customer experience isn’t about a single tool or algorithm; it’s a change in mindset and organizational structure. Moving personalization from the periphery to the center of your strategy requires treating it not as a project with a start and end date, but as an essential, ongoing capability that everyone on the team is invested in. Soligon explains how her team made this crucial transition, moving beyond isolated tests to an embedded, always-on approach.
“I’ve been really proud of the progress we’ve made, we’ve moved from more of a… a experimentation sidecar, if you will, to having personalization embedded across our customer journey… I like to say it’s more of our core capability now versus something that a few people worked on. Everyone is invested in personalization and we’re seeing the results from that as well.”
This evolution from “sidecar” to “core capability” is a powerful model for any marketing leader. It signifies a move away from simply proving that personalization can work to operationalizing it so that it does work, consistently and at scale. The key, as Soligon notes, is leveraging AI to focus on what truly matters. Instead of trying to personalize everything, her team trains AI models to optimize for high-value engagements like “add to cart” or cart completion. This is a critical distinction. It’s not about personalizing for personalization’s sake; it’s about identifying the moments in the customer journey that have the most significant business impact and applying technology to improve the outcomes of those moments. This disciplined approach ensures that resources are focused where they can generate measurable results, making the business case for further investment clear and compelling.
Translating Customer Voice into Actionable Intelligence
Every enterprise leader is familiar with the challenge of Voice of the Customer (VoC) programs. We collect vast amounts of data—surveys, support transcripts, reviews—but turning that firehose of feedback into a prioritized action plan is another matter entirely. With over four million pieces of feedback annually, Microsoft’s challenge is one of extreme scale. The traditional approach of manual analysis is simply not feasible. This is where AI becomes an indispensable partner in understanding, and acting upon, customer sentiment.
“We get over 4 million pieces of feedback from customers annually… we’re using AI to help us identify high quality feedback and then quickly take action on it to improve our customer experience… Our AI model picks that up and puts it at the top of the list for our analysts to look into further. We further than married that with what we call basically high value engagement actions.”
Soligon’s methodology here offers a brilliant blueprint for any organization drowning in feedback. First, use AI to define and surface “high-quality feedback”—comments that are specific, actionable, and linked to a real experience. This immediately filters out the noise (e.g., a one-word comment like “slow”) and elevates the signal. Second, cross-reference that high-quality feedback with high-value business metrics, like add-to-cart rates or average order value. When a specific, actionable piece of feedback is directly linked to a critical business KPI, it jumps to the top of the priority list. This creates a direct line between what customers are saying and what the business needs to achieve, transforming the VoC program from a listening exercise into a strategic driver of both customer satisfaction and revenue.
A Pragmatic Framework for AI Agents
With the rise of generative AI, the clumsy chatbots of yesteryear are being replaced by sophisticated virtual assistants. Microsoft’s own virtual store assistant has already managed over 1.7 million conversations, helping customers with everything from product comparisons to order support. However, launching a customer-facing AI agent is not without its risks. We’ve all seen the headlines about bots going off the rails. Soligon provides a pragmatic, three-part framework for deploying these agents successfully and responsibly.
“I would look at three different elements when you’re creating an agent. Really being first upfront about what your use case is. Clear intent. Give the agent some guardrails… Second, I think you got to look at data quality and governance… And then finally, I would say the humans are still important. Like you can’t just set it and forget it.”
This framework is a masterclass in execution for any leader venturing into this space.
- Clear Intent and Guardrails: Before writing a single line of code, define the agent’s purpose. Is it for shopping assistance? Post-purchase support? Giving the AI a clear mandate and operational boundaries is the first line of defense against brand-damaging irrelevance or hallucinations.
- Data Quality and Governance: The timeless principle of “garbage in, garbage out” is amplified in the age of AI. The agent is only as good as the data it’s trained on. Establishing rigorous data governance is essential for ensuring accuracy, transparency, and privacy.
- Human Oversight: An AI agent is not a “set it and forget it” technology. Continuous human monitoring, checking, and optimization is critical. This “human-in-the-loop” ensures the agent’s performance aligns with brand standards and that it continues to improve over time. It also provides the wisdom to know when to hand off a conversation to a human for a better customer experience.
Preparing for the Era of Agentic Commerce
Perhaps the most profound shift on the horizon is the change in how consumers discover and purchase products. The traditional journey of opening a browser and searching on a brand or retailer site is being complemented, and in some cases replaced, by interactions with personal AI assistants like ChatGPT, Copilot, or Gemini. This new paradigm, which can be called “agentic commerce,” requires a new strategy for discoverability and transaction.
“I truly believe that shopping in the future is going to be agent driven or influenced to some extent… these AI agents… are great new traffic sources for us… for me, I’m really thinking about what do I need to do with APIs? What do I need to do with my data to make it discoverable? … just sitting back and waiting, I don’t think as an e-commerce leader, you can do that.”
This is a direct call to action for every e-commerce and marketing leader. The work we’ve done on SEO for search engines must now be re-imagined for discoverability by AI agents. This is a technical and strategic challenge that involves structuring product data, building robust APIs, and ensuring your brand’s information is easily accessible and interpretable by these new platforms. The goal is to enable seamless purchase experiences directly within the conversational stream, meeting customers where they are already spending their time. As Soligon points out, this isn’t a distant future to plan for; it’s happening now. Leaders who are not actively experimenting and building the necessary infrastructure risk being left behind as their customers begin to shop in entirely new ways.
The insights from Microsoft’s D2C journey provide a clear roadmap for navigating the complexities of modern marketing technology. The overarching theme is one of deliberate, strategic evolution. It’s about elevating personalization from a tactical experiment to a core business competency. It’s about using AI not just to listen to customers, but to understand them in a way that drives prioritized, impactful action. And it’s about looking beyond the current landscape to prepare for a future where commerce is increasingly conversational and agent-driven.
The pace of innovation is undeniably daunting, but the principles for success remain grounded in business fundamentals. It requires a clear vision, a culture of curiosity and experimentation, and a disciplined focus on delivering measurable value. For marketing leaders, the challenge is to harness the power of these incredible new technologies not just to optimize the present, but to architect a more intelligent, responsive, and ultimately more valuable future for the customers we serve. The time for waiting is over; the time for building is now.
