Grainger's AI & ML Lead On Navigating AI Transformation Through Organizational Redesign

The tech industry's intense focus on AI models and platforms is missing the point. Real transformation succeeds or fails long before deployment, determined by how organizations restructure roles, decision rights, and incentives. Without that organizational blueprint, even the most advanced AI strategy is little more than an expensive experiment.

Hardi Gokani, Director of Product Management for AI/ML at Grainger, is a Fortune 500 growth catalyst whose work sits at the intersection of AI product strategy and organizational transformation. At Grainger, she enabled over $398M in revenue through AI-driven product initiatives and scaled a computer vision application from five to 1,000 users in four months. Earlier, at CVS Health, she protected $35M in annual revenue during an enterprise-wide move to the cloud. She believes the industry's focus on technology is misplaced, and the real work of AI transformation is deeply human.

"AI transformation doesn't succeed or fail at the point of deployment. It succeeds or fails much earlier, when companies decide who owns AI-driven decisions, how roles change, and what leaders are accountable for," said Gokani. That stance cuts against the instinct to treat AI adoption as a procurement decision. The organizational blueprint has to come first. "If your roadmap doesn't explicitly address roles, decision rights, and incentives, it's not a strategy, it's just a science experiment," she added. Without that foundation, even well-resourced AI initiatives tend to fail before maturity.

Gokani identified two organizational pitfalls that consistently keep AI stuck in pilot mode. The first is a talent imbalance that most companies create without realizing it. The second is a leadership fluency gap that turns even well-designed AI initiatives into sources of frustration.

• The demo trap: The first pattern of failure is often rooted in how organizations staff their AI initiatives. Companies unintentionally over-invest in "builders," data scientists and engineers, while underfunding the "translators": the business leaders, product leaders, and domain experts who connect AI to real problems. The resulting imbalance has a predictable outcome. "If 90% of your AI talent sits in model development, you're optimizing for demos, not for business decisions. You have to make sure you have a good balance when you structure your talent for an AI initiative," said Gokani.

• Mind the human gap: The fluency problem lives at the top of the organization. Executives routinely sponsor AI but struggle to evaluate its probabilistic nature, which leads to inflated expectations and team burnout. At the same time, AI actively reshapes power and identity at work, stoking fear among employees that their experience will be made irrelevant. She noted that ignoring that fear will guarantee resistance rather than adoption. The companies seeing real impact address that anxiety directly, by being explicit about where AI will and will not be used and by rewarding good human judgment alongside automation. Measurement is where that commitment becomes visible. "In addition to model accuracy, you need to measure things like: How many decisions did you improve? What time did you save for each user? What adoption persisted after the initial launch? If humans are not in your metrics, value won't be either," she said. Netflix offered her clearest example of what a genuine AI-native operating model looks like: completely and intentionally invisible, with recommendation and personalization algorithms integrated into every product, content, and marketing decision the business makes.