Redesign the CPG and Retail Business Around AI (Do Not Retrofit It)

Introduction

Opus Advisory Board member Subra Goparaju recently explored why agentic commerce is emerging as a strategic priority for CPG and retail leaders; this second part examines what it will take to govern.

The hard question for CPG and retail leaders is how to redesign the business so AI can create durable value across merchandising, supply chain, pricing, collections, and customer engagement. That is the shift Subra, a veteran of enabling enterprise transformation at HanesBrands and beyond, focuses on in this conversation. He argues that companies need to examine where enterprise value resides and which data foundations matter most. He believes in adopting governance practices that enable AI to scale without undermining trust. For leaders managing fragmented technology estates and channel complexity, that insight is critical.

Subra’s view will resonate with retail and CPG leaders because tough market economics are forcing them to find new ways to boost productivity. His perspective links business pressures to concrete day-to-day operating decisions. His prescription is simple: first, leverage AI to govern a few critical data objects, then redesign workflows rather than layering AI on top of older ones. Finally, his prescription for AI deployment is to focus on domains that can measurably improve margins or revenue.

Question. For CPG and retail enterprises, what does better governance for agentic AI look like in practice?

Subra: Risk-based clarity is the first step to better governance. Organizations are struggling to find practical ways to classify AI use cases by risk and govern them accordingly. Companies need a methodology that can distinguish between low-, medium-, and high-risk use cases, rather than treating every application the same or leaving teams to improvise. They also need a more reliable data foundation, with tighter control over the layers that contain sensitive information and over data objects that matter most to the transaction journey. The idea need not be to govern everything at once. The ideal approach is to identify the five or six data objects that shape most of the journey and govern those first—then build confidence from there. This approach reflects a broader shift in which companies are reducing complexity to move faster and build AI readiness.

Question. How can companies balance effective governance with the need to keep innovating?

Subra: The balance comes from prioritization, not from trying to do everything at once. Many companies began with high-ROI use cases, but too often those efforts remained fragmented and failed to translate into enterprise value. The better question for leaders to ask is, “Which two or three domains (marketing/merchandizing, supply chain, sourcing, operations, sales, brand, customer engagement, financial performance, field operations, technology platforms, etc.) can genuinely move the needle on margins, revenue, or operating leverage?” Once those domains are identified, governance becomes easier to design because it is tied to a real transformation agenda rather than scattered experimentation. This is where the distinction between retrofit and redesign becomes important. If companies continue to layer AI onto disconnected workflows, both innovation and control will suffer. But if they redesign a few critical domains end-to-end, they can move faster while preserving the ability to innovate.

 Question. Where are the biggest opportunities to productize AI in CPG and retail?

 Subra: The strongest opportunities are at the customer edge and at the operating core. Companies are already seeing gains on the front end from hyper-personalization, loyalty optimization, intuitive search, and smoother checkout experiences. On the operational side, the opportunity is even larger. Order orchestration, dynamic fulfillment, returns routing, demand sensing, collections, and dynamic pricing all lend themselves to AI interventions. They involve high-frequency decisions, and the financial impact is measurable.

Question. If a company has experimented with AI in silos, how should it broaden adoption across the business?

 Subra: Broadening AI starts with solving a business problem end-to-end. Solving only parts of the process does not help. Many organizations have customer service bots, copilots, or isolated pilots, but these do not constitute transformation. The right approach is to identify two or three connected domains where value compounds across the workflow—for example, demand sensing, supply planning, logistics, order orchestration, and store execution. From there, the task is to re-architect the workflow around how the business creates value, rather than forcing AI into legacy structures. There is no universal blueprint. Every company is different. It has different processes and capability gaps. It has different business goals. But the core principle remains the same: Look at how domains are run with brutal honesty, then identify those where AI can improve outcomes, and redesign them with AI at the core.

– – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – –

Subra Goparaju is an advisory board member at Opus Technologies. Over a career spanning more than 30 years, he has led large-scale enterprise transformations across major consumer-facing organizations, with experience in enterprise architecture, cybersecurity, data, and digital modernization. He has helped global enterprises manage complex business and technology shifts. The blend of operational depth and strategic perspective he brings is invaluable to CPG and retail companies seeking to move beyond fragmented AI pilots. His views are especially relevant as payment networks build the foundations for agentic commerce and more intelligent, low-friction customer journeys.