A Senior Manager of Technical Product Management for Alexa Smart Home at Amazon on building trust in AI-powered home experiences.
According to the 2025 Home Security Market Report, Smart home adoption in the United States continues to expand, with over 13 million new home security systems projected for installation in 2025 alone. Concurrently, introducing new state-level regulations governing IoT privacy and data usage underscores the increasing complexity of aligning consumer expectations with the operational realities of AI-driven home technologies.
While technical capabilities have advanced, user experiences often remain fragmented, opaque, or challenging to manage. As industry stakeholders work to deliver AI assistants that are not only multimodal and context-aware but also secure and intuitive, the challenge lies in achieving scale without compromising trust or usability.
To examine how these issues are being addressed in practice, we spoke with Mahak Rawal, Senior Manager of Technical Product Management for Alexa Smart Home at Amazon. Ms. Rawal oversees a portfolio that has reached millions of monthly users and has led the development of multimodal AI experiences, including the Echo Hub and computer vision–based features such as enabling Alexa to search and summarize videos from connected Ring cameras. Under her leadership, the platform has seen a 22% increase in customer satisfaction, a 15% year-over-year growth in connected device engagement, and a 20% improvement in feature performance driven by experimentation frameworks and analytics. Before her current role, she led large-scale digital health initiatives at UnitedHealth Group, improving patient outcomes by 15% and supporting various programs generating over $100 million in annual revenue. Ms. Rawal is a Senior Member of IEEE and serves as a judge for the Globee® Awards 2025 in Artificial Intelligence and Disruptive Technologies categories.
Mahak, multimodal AI systems that combine language, vision, and automation, are becoming standard in smart home platforms. What are the current limitations of these systems when it comes to ensuring real-time responsiveness and contextual accuracy in domestic environments?
Multimodal AI presents numerous opportunities in the smart home, but delivering seamless, real-time experiences across devices remains a challenge. Performance can vary depending on hardware, network quality, and the number of people sharing the space. Keeping context, who’s speaking, what’s happening, what matters right now- isn’t always straightforward. Especially in areas such as security or automation, timing and relevance are crucial. To make it work, the system needs to process information locally when possible, handle it efficiently, and remain responsive to users’ needs.
Your redesign of the Echo Hub interface for advanced users managing 50+ devices led to a 22% increase in satisfaction within months. What does this say about the current gap between consumer complexity and the design maturity of mainstream smart home systems?
The Echo Hub helped uncover a segment of users managing dozens of connected devices—routines involving lights, locks, thermostats, and cameras across multiple rooms. For these users, even small delays or missing device status create friction in daily interactions. The redesign focused on improving visual responsiveness, supporting a broader range of devices, and ensuring real-time accuracy. For example, enabling smoother integration with less common devices helps reduce setup complexity and user frustration. Feedback indicated that many existing platforms may not fully meet the demands of more advanced setups. Addressing this requires moving beyond generic interfaces toward more purpose-driven, responsive, personalized experiences.
You led executive reviews and oversaw the development of searching for Ring camera videos through Alexa, using CV to surface relevant footage. How did you address the growing concerns around passive surveillance and the normalization of always-on visual data in private spaces?
Features like event-based triggers and event-specific summaries, such as detecting a person or package, helped avoid the need for constant monitoring. Privacy and legal teams were involved early to shape a careful, use-driven design. Rather than normalizing surveillance, the focus was on giving users meaningful context without overwhelming them. The response demonstrated that when AI remains transparent and purposeful, it can support everyday needs while respecting boundaries within the home.
You led the development of Alexa’s security panel UX, resulting in a 10% increase in feature adoption. To what extent do current smart home platforms adequately reflect the diversity of user personas, from children to caregivers, when designing critical access functions?
Smart home security needs can differ a lot—some households prioritize control, while others just want things to work simply. In my experience, designing for everyday roles like parents, guests, or caregivers makes access feel more natural. When people can set different permissions without having to dig through complicated menus, the system becomes easier to trust. Today, many platforms still assume a single-user model, which doesn’t reflect how homes actually operate. Adapting to real-life dynamics—shared spaces, shifting routines—goes a long way in making security features both safer and more widely used.
You’ve consistently secured VP+ buy-in for high-risk initiatives like AI automation in security systems and CV-power home monitoring. What does that internal support reveal about how large tech companies now perceive accountability in consumer-facing AI?
Gaining alignment often comes down to showing how an idea delivers value while managing potential risks. Conversations with leadership focus not only on what AI can do, but also on how it behaves in the home, how people interact with it, and where concerns might arise. Involving customer privacy, product, and engineering voices early helps shape features that are ambitious yet grounded in reality. This approach fosters a shared understanding that advancing consumer AI also entails clear responsibility for how these technologies operate in everyday life. Accountability becomes part of the process, not just the outcome.
Your career spans healthcare and smart home tech, both sectors with strict privacy constraints. How transferable are responsible AI practices across industries, and where do you see the most friction between innovation and compliance?
Work in healthcare shapes a mindset where privacy and responsibility guide every product decision. That approach translates well to consumer AI, especially when designing experiences that deeply connect with everyday life. Friction tends to arise when innovation outpaces the ability of policies to keep up, which is common with emerging technologies such as computer vision or conversational systems. Involving legal and privacy perspectives early in the process makes it easier to navigate those gaps. When responsibility and transparency is part of the culture, it remains consistent across different industries, helping teams move forward with greater clarity.
You doubled enrollment and generated multimillion-dollar revenue through coaching programs at UnitedHealth Group. What changes made that scale possible?
One of the most rewarding projects was the complete redesign of our wellness coaching programs. We shifted away from a static recommendations model that wasn’t engaging members and used data to personalize offerings based on individual health conditions. That change more than doubled program enrollment year over year, generating millions of dollars of recurring revenue for the company. The new model not only increased participation but also improved overall health outcomes, because we were recommending the right programs to the right people at the right time.
The Quit for Life program you designed improved user control and was projected to reduce eligibility setup times by 80%. What was your approach to rethinking behavior change design?
With the Quit for Life program, we realized that rigid, time-based progression wasn’t resonating with users dealing with addiction. Over 40% reported feeling a lack of control. We redesigned the platform into a hybrid model that combines time-based and action-based progression. This empowered users to take ownership of their journey. It also enabled us to personalize recommendations and cross-sell other relevant health programs, such as weight loss and hypertension coaching. That redesign would help decrease healthcare eligibility setup times from 12 weeks to just two and became the foundation for several new programs across the organization.
As computer vision features become more embedded in consumer products, how should product leaders balance commercial viability with the ethical boundaries of domestic surveillance?
In my view, trust is what makes or breaks computer vision features in the home. People want tools that solve real problems, such as identifying who was at the door, without feeling watched all the time. That’s why it helps to build with clear intent from the start: make it optional, keep it local when possible, and explain what it’s doing. When the design reflects how people live, the technology feels more like support than surveillance. The goal is to do the right things in a way that people feel comfortable inviting into their space.
