Drawing experience in artificial intelligence, enterprise data, financial systems, and supply-chain technology, the product leader is developing a practical framework for turning sophisticated models into systems that people can understand and use.
By Paresh Dave
Artificial intelligence is entering corporate operations at a remarkable pace, yet many of the systems attracting executive investment continue to face a deceptively simple obstacle: the people expected to use them do not always understand or trust what they produce. A model may perform impressively in a technical evaluation but struggle to influence decisions once it reaches a finance, merchandising, or supply-chain team. An analytics platform may process large volumes of information while employees continue relying on spreadsheets and manual reconciliation because the system does not explain where its conclusions came from. For Aakanksha Polepalli, an enterprise product leader whose experience spans artificial intelligence, predictive analytics, data platforms, financial technology, and supply-chain operations, this gap between technical performance and human adoption has become one of the defining problems of modern enterprise technology. Her work focuses on closing it by designing AI-enabled products around the decisions people must make, rather than around the capabilities a model can demonstrate.
“The bottleneck in enterprise AI adoption usually isn’t model performance—it’s trust and adoption,” Polepalli said in an interview. That conclusion emerged from a pattern she encountered repeatedly across large and operationally complex organizations. Businesses were accumulating significant amounts of financial, customer, inventory, supply-chain, and transactional data, yet important decisions remained constrained by disconnected systems, conflicting definitions, and limited visibility.
Designing AI Around Decisions
Polepalli’s approach begins before a model is selected or a technical roadmap is approved. She first asks which decision the organization is attempting to improve, who is responsible for making it, what information that person currently uses, and what would have to change for a new product to affect the outcome. This decision-centered method allows her to work backward toward the required data, architecture, user experience, governance controls, and performance measures. It also forces a distinction between introducing AI because it is available and applying it because it is appropriate.
This philosophy shaped Polepalli’s work on an AI- and machine-learning-powered forecasting platform developed for large-scale financial-planning and supply-chain operations. The system had to account for variables including pricing, promotions, seasonality, consumer behavior, inventory conditions, and operational change. Polepalli helped define the product strategy, forecasting use cases, data requirements, user workflows, success measures, and roadmap while coordinating the priorities of engineers, machine-learning specialists, data scientists, finance professionals, and operational stakeholders. Based on internal project performance metrics at the department and sales-channel planning level, the forecasting platform achieved 99.6% forecast accuracy, maintained a mean absolute percentage error (MAPE) below 1%, and improved year-over-year forecasting accuracy by 10%.
Those results demonstrated technical capability, but Polepalli considers the product’s explainability layer the more consequential aspect of the work. Business users did not simply need a forecast; they needed to understand why it had changed before incorporating it into a purchasing, planning, or operational decision. Polepalli introduced visual experiences that allowed users to examine forecast drivers and generative-AI summaries that translated complex analytical output into accessible business language. Rather than asking a planner to accept a prediction as a black-box conclusion, the product was designed to reveal the factors behind it at the moment the information was being evaluated.
“The harder win was getting users to trust and act on the forecasts, not just view them,” Polepalli said. “A model output has to become something a planner can actually build a decision around.” Her observation points to a broader weakness in how enterprise AI programs are frequently evaluated. Deployment is often treated as proof of success, while user behavior is measured later or not at all. Polepalli instead treats adoption as a product requirement that must influence architecture, interface design, explanation, and workflow integration from the beginning. A technically successful model that is routinely bypassed by its intended users, in her view, is not a successful enterprise product.
A Consistent Method Across Complex Systems
Polepalli has applied this decision-centered approach to a range of enterprise problems involving generative AI, analytics modernization, predictive assessment, cloud-based systems, financial workflows, and supply-chain infrastructure. Although the applications differ, the problems often share a common structure: data is distributed across multiple platforms, users depend on manual work to interpret it, and technology teams are asked to automate a process before the organization has agreed on what the information means.
In one enterprise analytics initiative, generative AI capabilities reduced the manual effort required to generate structured database queries by approximately 99%, based on internal project performance metrics, allowing analysts and business users to retrieve information more quickly and focus on interpreting insights rather than writing complex queries. The objective was not simply to demonstrate that natural-language technology could generate technical output. It was to allow analysts and business users to obtain useful information without repeatedly translating every question into a specialized query. In another machine-learning initiative, a predictive model for assessing claims severity was associated with a 4% reduction in severe claims and substantial annual savings. The system was designed within health-data handling requirements, requiring product decisions to account for business impact, governance, compliance, and responsible use.
Polepalli’s work has also involved modernizing enterprise-data architecture and the systems through which information moves. In one large technology environment, redesigning data paths across several enterprise sources reduced data-transmission time by approximately 96%. Related product improvements were associated with a 17% increase in customer-satisfaction measurements and an improvement of nearly 29% in application responsiveness. These outcomes reinforced Polepalli’s view that visible AI features cannot be separated from the infrastructure beneath them.
The same decision-centered approach extended into retail supply-chain and financial operations, where fragmented data across supplier, logistics, and enterprise systems can create significant operational and margin implications. Polepalli led product initiatives connecting enterprise resource planning, warehouse-management, order-management, and point-of-sale environments to improve inventory visibility, purchasing workflows, tariff and tax calculations, and the allocation of freight and landed costs across a diverse supplier network. Internal performance measurements associated this work with a 12% improvement in tariff and tax accuracy, an 18% improvement in profit-margin visibility, and a 10% reduction in freight, landed-cost, and transfer-cost variance. Additional improvements included stronger portal performance and a reduction of more than 90% in transaction timeouts. Across these initiatives, Polepalli applied a consistent product methodology: identifying the business decisions constrained by unreliable information, establishing accurate data flows across enterprise systems, and measuring whether the resulting products improved the organization’s ability to make timely operational and financial decisions.
Building an Applied Research Agenda
Polepalli has increasingly translated the lessons from her professional experience into an applied research agenda examining enterprise data, artificial intelligence, supply-chain systems, financial technology, and product management. Between 2020 and 2026, she developed a portfolio of 18 research abstracts that she is preparing for publication. The work has not yet been presented as a body of published or peer-reviewed scholarship, and Polepalli does not characterize it that way. Instead, the abstracts represent an effort to organize recurring enterprise problems into frameworks that can be evaluated, refined, and eventually shared with a broader professional audience.
The research addresses subjects including explainable AI, machine-learning forecasting, generative AI in business workflows, data-driven product prioritization, enterprise-platform modernization, financial decision systems, and responsible AI adoption. A central question connects many of these topics: what conditions must exist before an organization can depend on an intelligent system? Polepalli’s answer extends beyond model performance. The organization must have trustworthy data, clearly defined business metrics, ownership and governance mechanisms, an appropriate use case, and a product experience that explains the system’s output at the level required by the person making the decision.
This applied perspective distinguishes Polepalli’s research interests from work focused solely on foundational algorithms. She does not claim to have created a new machine-learning architecture or patented a proprietary model. Her contributions lie in product architecture, AI-adoption frameworks, explainability strategies, data requirements, and decision-support experiences that operationalize existing technology within large organizations. The distinction matters because many of the most visible failures in enterprise AI do not originate in the model itself.
The Organizational Challenge Behind the Technology
The greatest obstacles Polepalli has encountered have often been organizational rather than technical. Enterprise AI projects bring together teams that speak different professional languages and evaluate success according to different standards. Data scientists may focus on accuracy and statistical performance. Engineers may emphasize reliability, scalability, and system design. Finance leaders may prioritize cost, margin, control, and forecast quality. Operational users may care most about speed, clarity, and whether the recommendation reflects the realities of their work. Compliance teams may require traceability, documentation, and clearly defined responsibilities.
Polepalli leads across those boundaries through strategy, influence, and cross-functional coordination. Her roles have generally required her to direct product priorities without relying solely on formal reporting authority. She has worked with multidisciplinary teams of engineers, data specialists, designers, finance professionals, and operational leaders, translating business needs into technical requirements while also explaining technical constraints to nontechnical stakeholders.
Her philosophy is captured in a distinction she frequently makes between “doing things right” and “doing the right thing.” Execution quality remains essential, but Polepalli argues that product leadership carries a deeper responsibility: determining whether the organization is solving a problem that merits the investment. In the context of AI, that can mean resisting pressure to begin with a model, platform, or trend before the business need has been established. It can also mean concluding that automation is not appropriate when the data, process, or governance structure is not mature enough to support it.
“I refuse to treat AI adoption as a technology rollout,” Polepalli said. “I build explainability, trust, and workflow fit into the product from the start, and I measure success by whether people actually change their decisions because of what has been built—not just whether the model shipped.” That position reflects both a practical and a cautious view of enterprise innovation. It acknowledges the potential of advanced systems while placing responsibility on product leaders to ensure that the technology can be understood, challenged, and used appropriately.
Preparing Enterprises for What Comes Next
Polepalli’s current work focuses on enterprise data governance and AI readiness within a large, multi-brand retail environment. The initiative involves creating consistent metric definitions, strengthening data lineage, improving visibility across supply-chain operations, and sequencing the platform investments required for future AI and machine-learning adoption. Rather than treating AI readiness as a final technical assessment performed immediately before deployment, Polepalli approaches it as an organizational discipline combining governance, infrastructure, product strategy, operating processes, and user preparation.
An organization may be capable of running a machine-learning model while remaining unprepared to rely on it. When departments calculate the same measure differently, when data ownership is unclear, or when users cannot trace a recommendation to its source, expanding AI can amplify confusion rather than resolve it. Polepalli’s work attempts to address those conditions before new intelligent products are introduced at scale. The goal is to create an environment in which future systems can be governed consistently, connected to reliable information, and integrated into the decisions they are meant to support.
In an industry often captivated by what artificial intelligence can accomplish, Polepalli’s work is focused on what must happen before an organization should depend on it. Her projects and developing research agenda suggest that enterprise AI’s next stage will not be determined only by increasingly capable models. It will also depend on the ability of product leaders to connect those models to trustworthy data, explainable experiences, responsible governance, and measurable business decisions.
Her advice to emerging technology leaders is straightforward: “Get obsessed with adoption, not just accuracy.” A technically impressive platform may demonstrate what is possible, but its real value begins only when the people responsible for using it understand what it is telling them, recognize its limitations, and trust it enough to act. For Polepalli, that moment—not the model’s launch—is the true measure of enterprise intelligence.
