As subscriptions, connected devices, AI, and post-purchase services reshape consumer technology, customer relationship management (CRM) has become a critical enterprise platform for customer experience, operational excellence, and long-term business growth. Nithesh C. Nekkanti discusses the architectural decisions, governance strategies, and customer identity frameworks that help organizations create more valuable long-term customer relationships.
Nithesh C. Nekkanti is Manager, CRM Applications at Sonos, where he leads Salesforce development teams responsible for AI-enhanced agent productivity, platform governance, and enterprise order management. During more than a decade designing CRM solutions for consumer technology organizations, he has worked across customer identity, enterprise architecture, AI-enabled service platforms, large-scale Salesforce implementations, and the operational systems that support customers long after the initial sale. His experience spans both the technical design of enterprise platforms and the organizational challenges of delivering CRM solutions that continue evolving as products, services, and customer expectations change.
CRM platforms were originally designed for a business environment where success was measured primarily by acquiring customers and recording transactions. Connected products, subscription services, digital support, and artificial intelligence have fundamentally changed that model, making long-term customer relationships central to both operational excellence and business growth. Sustaining those relationships requires a different approach to customer identity, enterprise integration, and CRM architecture, particularly as business decisions increasingly depend on accurate, connected customer data. In this interview, Nekkanti discusses the engineering judgment, technical leadership, and design philosophies shaping the next generation of enterprise CRM.
ELLEN WARREN: CRM has traditionally been viewed as a platform for sales and marketing. From your perspective, what role should it play as consumer technology companies become increasingly dependent on long-term customer relationships?
NITHESH C. NEKKANTI: At its core, CRM is about relationship management, not simply supporting sales and marketing activities. CRM should serve as a primary platform for customer engagement by enabling companies to maintain an ongoing relationship with the customer throughout the product lifecycle. Instead of treating CRM merely as a static system of record that identifies who the customer is, organizations should view it as a system of engagement that actively strengthens the customer experience and long-term relationship.
EW: Your technical contributions span customer identity, AI-enabled service operations, platform governance, and enterprise order management, areas that are often managed independently within large organizations. What common architectural principles have emerged from working across those different aspects of enterprise CRM?
NCN: Having worked across various domains within enterprise CRM, I’ve found that the primary challenge in building an effective system of engagement is establishing a unified 360-degree view of the customer. Delivering a comprehensive “Customer 360” requires aggregating data from diverse sources, including e-commerce, customer support, social interactions, and transactional systems, so teams have the context needed to serve customers effectively. The foundational architectural principle that makes this possible at scale is event-driven architecture. By enabling diverse systems to communicate asynchronously via trigger events, we can ingest, sync, and process high-volume customer data seamlessly across the enterprise.
EW: Customer identity becomes increasingly difficult to manage when products are sold through retailers instead of directly by manufacturers. What engineering choices have the greatest influence on creating a complete and reliable customer record across multiple systems and customer touchpoints?
NCN: Retailers rarely share raw point of sale (POS) data because they consider the customer relationship to be theirs, not the manufacturer’s. There are exceptions, such as automotive dealerships bound by contracts to share buyer details with Original Equipment Manufacturers (OEMs), or platforms like Walmart using secure data clean rooms to share aggregated analytics without revealing Personally Identifiable Information (PII). However, for most consumer hardware brands, the most effective engineering strategy is focusing on post-purchase identity capture. Companies like Apple don’t rely on retail POS data; instead, they capture customer identity during the initial product setup and registration phase (Apple ID). Treating device onboarding and product registration as a seamless, high-value identity event early in the ownership lifecycle is the single most reliable way for OEMs to establish a direct customer relationship.
EW: Artificial intelligence is rapidly becoming part of enterprise customer service, but successful AI deployments depend on much more than selecting the right models. What foundational work within CRM architecture needs to be in place before organizations can expect AI to deliver meaningful operational value?
NCN: AI output is only as good as the data you feed it, so data hygiene is the first major hurdle. You have to look at data in two ways. First is your institutional knowledge, which includes all the documentation and rules built up over the years. This knowledge data needs to be structured so an AI agent can actually understand context before taking action. Second is live customer data. Unlike knowledge data, operational details like order status change constantly throughout the day, so the system needs clean, real-time access.
Beyond data quality, you have to look at your existing workflows and clean them up before introducing AI. Automating a broken process just speeds up bad outcomes, so streamlining workflows helps you remove redundant steps and pinpoint areas where manual human errors happen most often. From there, you need clear guardrails so the AI knows its boundaries, specifically when it can act independently and when it needs to stop and ask for human approval. Finally, you need a closed feedback loop. AI won’t work flawlessly from day one, so having a mechanism to log mistakes, apply corrections, and retrain the model is essential if you want it to improve and scale over time.
EW: Large CRM environments often connect customer service, commerce, warranty management, order fulfillment, and other enterprise systems. Where do integration challenges most commonly emerge, and what design principles have proven most effective for maintaining reliable data across those environments?
NCN: Integration challenges usually boil down to three main issues: identity fragmentation, data conflicts, and point-to-point integration complexity.
While most companies now use unique customer IDs, the real problem is data duplication within individual systems. For instance, a CRM might have a single clean record for a customer, while the support platform has five duplicate entries for that same person. When your e-commerce system tries to sync data, it doesn’t know which record to pick unless you have a universal anchor and clear deduplication rules in place.
Another major challenge is data conflict between systems. A customer service system and an order management system might hold the exact same order record, but with different shipping addresses because the customer called support to update it. To handle this, you have to establish clear domain ownership so every system knows which platform is the authoritative source of truth for specific data attributes.
Lastly, as these systems integrations proliferate, you end up with what I call “integration spaghetti,” where dozens of direct point-to-point integrations create a fragile network that breaks as you scale.
To address these challenges, we rely on a few core design principles: First, use a robust integration middleware to centralize communication rather than building direct point-to-point connections. Second, define explicit system-of-record ownership for every data domain so there is no ambiguity about which system holds the truth. Third, leverage data virtualization whenever possible. You don’t need to replicate every piece of data into every system. Fetching data on demand through virtualization keeps systems light and prevents data from drifting out of sync. Finally, design all integration endpoints to be idempotent. System timeouts and retry mechanisms frequently cause the same payload to be sent more than once. Idempotent design ensures that redundant payloads or retried requests are safely processed without creating duplicate records or unexpected side effects in the destination system.
EW: Every enhancement, integration, and customization adds complexity to an enterprise CRM platform. How do you establish governance practices that encourage innovation while keeping the platform maintainable over the long term?
NCN: Governance shouldn’t be a bureaucratic bottleneck that slows teams down at the end of a release cycle. Instead, it needs to be built directly into the software development lifecycle from day one.
That means treating technical debt management as a continuous part of developing new features rather than an afterthought. It also means enforcement of automated quality gates in the CI/CD pipeline, such as automated code analysis and test coverage thresholds, so quality issues are caught immediately during development.
Finally, governance works best when you enforce domain-driven ownership. Empowering domain teams with clear architectural standards and local accountability creates far better outcomes than relying on a centralized, bureaucratic review board to approve changes right before deployment.
EW: Several of your projects have focused on improving customer service operations through AI, workflow engineering, and platform design, including measurable improvements in agent productivity and case handling. When you’re evaluating a new initiative, how do you distinguish between technology that simply automates work and technology that genuinely improves the customer experience?
NCN: The clear tell is which metrics actually move. I categorize metrics into two distinct buckets: operational metrics and experience metrics. To genuinely elevate customer experience, experience metrics must move alongside operational ones.
If operational metrics like deflection rate and average handle time improve, but experience metrics such as Customer Satisfaction Score (CSAT) and repeat contact rate stay flat or degrade, that is a classic sign of technology that simply automates work without helping the customer.
A great example of this is AI chatbots. A chatbot can easily boost deflection numbers, but that does not mean customers are happier. A self-service initiative truly passes the experience test when customers voluntarily choose to use the chatbot over calling phone support because it offers a faster, genuinely better resolution to their problem.
EW: You’ve also led initiatives to standardize development practices and platform governance across enterprise development teams. Why are engineering standards and technical discipline just as important to long-term CRM success as introducing new technologies?
NCN: Engineering standards give you a predictable foundation so you can scale without making the platform fragile. Without that technical discipline, a codebase eventually becomes so brittle that teams are afraid to touch it or make future changes. Having consistent standards also makes onboarding new talent much faster, because developers can jump into the code and understand the patterns right away.
EW: Your responsibilities have expanded from developing CRM solutions to leading engineering teams, establishing governance standards, and shaping enterprise platform strategy. How has moving from implementation into technical leadership changed the way you evaluate enterprise technology decisions?
NCN: Moving into leadership shifted my mindset from “How do I build this?” to “How does this scale?” and “How do we measure its success?”. Today, I evaluate technology decisions by balancing a solid technical architecture with real business impact and long-term maintainability.
EW: Connected products, artificial intelligence, subscription services, and digital customer experiences continue expanding the responsibilities of enterprise CRM. Which capabilities do you believe will become indispensable as organizations manage customer relationships over the next decade?
NCN: Real-time identity resolution will become indispensable, along with proactive support and the ability to understand customer intent instantly. I also expect that CRM platforms will increasingly go headless, exposing clean API surfaces that allow AI agents to take action dynamically. Organizations with a flexible CRM architecture will be better positioned to incorporate emerging technologies without disrupting existing business operations.
