Share with your CMO/CRO
Consumer hardware companies are flying blind on their own customers, and Nithesh Nekkanti, CRM Manager at Sonos, makes the architectural case for why that’s a fixable problem. Writing from a decade of Salesforce implementations across connected-product companies, he argues that CRM’s real job is continuous relationship management, not transaction logging, and that the technical decisions enabling that shift, event-driven data pipelines, identity capture at device setup, idempotent integration design, and AI guardrails, are what separate brands that own their customer relationships from brands that merely rent them from retailers.
What this means for your business
If your company sells through Best Buy, Target, or any other major retail channel, you probably don’t know who bought your product until that customer contacts support with a problem. Nekkanti’s framing of “post-purchase identity capture” as the single most reliable path to a direct customer relationship is the right diagnosis for this problem. The Apple ID comparison is instructive: the onboarding moment is a revenue moment, an identity moment, and a data moment simultaneously. Brands that treat device setup as a UX checkbox rather than a strategic acquisition event are handing the customer relationship to the retailer permanently.
The AI-readiness argument buried in the middle of this piece deserves more attention than it gets. Nekkanti’s point that automating a broken workflow just accelerates bad outcomes sounds obvious until you map it against actual enterprise behavior, where AI pilots frequently get layered on top of messy, undocumented support processes because the pressure to ship a demo is higher than the pressure to fix the underlying data. The distinction he draws between operational metrics like deflection rate and experience metrics like repeat contact rate is where most enterprise AI deployments quietly fail. A chatbot that deflects calls without resolving problems doesn’t improve customer relationships; it just moves the failure off the phone and onto a worse channel.
The integration architecture advice, middleware over point-to-point connections, explicit system-of-record ownership per data domain, idempotent endpoints (designed so the same request sent twice produces the same result once, preventing duplicate records) is genuinely sound and too rarely practiced. Most large CRM environments accumulate direct integrations the way legacy codebases accumulate workarounds, each one rational at the time, collectively fragile. The companies best positioned for AI-driven CRM over the next five years are the ones that invest now in cleaning up that architecture rather than piling agentic AI on top of what Nekkanti accurately calls “integration spaghetti.” The renewal decision this reframes is whether your current middleware investment is actually load-bearing or just a line item, because when AI agents start orchestrating across your CRM, commerce, and support platforms in real time, a fragile integration layer stops being a maintenance annoyance and starts being a customer-facing outage.
Based on reporting from Rethinking CRM Architecture for the Era of Connected Products, originally published 2026-09-10 14:29:00.
