Abhinandan Patil

If your organization is using AI, analytics, or operations research to create measurable real-world impact, this is one to look at. Submissions for the 2027 INFORMS Franz Edelman Award, often described as the “Nobel Prize” of OR and analytics are open, with the October 1 deadline coming up. I’m especially excited to spread the word this year as I join the 2027 Franz Edelman Award Selection Committee. Looking forward to seeing the work and impact represented in this year’s submissions! #INFORMS #FranzEdelmanAward #AI #Analytics #OperationsResearch

29

🚀 Excited to share the launch of Savings Insights in Amazon Business Analytics, a game-changer for procurement teams 🚀 For too long, uncovering savings opportunities meant hours of manual work: exporting data, building pivot tables, running complex analyses. Most teams simply didn't have the time or analytical bandwidth to do it consistently. Savings Insights changes that. Powered by AI, it automatically analyzes your purchase history and surfaces personalized recommendations including Quantity Discounts and Subscribe & Save opportunities along with projected annual savings. Smarter procurement decisions, without the manual effort. 💡 How much time does your team spend on spend analysis today? https://lnkd.in/gbUvtRNu #AmazonBusiness #Procurement #SpendManagement #BusinessPrime #SavingsInsights #AI #DataDriven

76

2 Comments

𝗗𝗼𝗺𝗮𝗶𝗻 𝗲𝘅𝗽𝗲𝗿𝘁𝗶𝘀𝗲 × 𝗔𝗜 𝗳𝗹𝘂𝗲𝗻𝗰𝘆 = 𝗮𝘀𝘆𝗺𝗺𝗲𝘁𝗿𝗶𝗰 𝗮𝗱𝘃𝗮𝗻𝘁𝗮𝗴𝗲. If expertise is weak, AI scales noise. If fluency is weak, capability stays capped. That’s the leverage shift underway. Recently, Jensen Huang framed it simply: Can a doctor use AI to catch diagnoses they would have missed? Can a lawyer sharpen arguments faster than the competition? Execution cost is collapsing. Judgment is becoming the differentiator. The real question isn’t whether AI replaces you. It’s whether someone in your field — with similar domain depth — becomes dramatically more effective because they know how to orchestrate intelligence. The next frontier isn’t writing the code. It’s coordinating systems that write, analyze, simulate, and optimize. I recently tested that idea end-to-end — building an MMM prototype from public data and open-source libraries. The results were instructive. More on that next.

16

1 Comment

This past week, I scheduled what I can only imagine were unexpected conversations with two high-performing members of my team. Both started with, "I need you to do me a favor and take some time for yourself away from work." They've relentlessly delivered for our customers, but I noticed they hadn't used the recent holidays to truly unplug. Their vacation accruals were hovering around the caps, and the reality is that even the most "always-on" engine will eventually hit its limit. As a leader, I view time off not as a reward for finished work, but as a requirement for future innovation. I stepped in because you simply cannot develop "Day 1" solutions with an overworked mind; high-level analytics requires a level of mental clarity that is impossible to maintain in a state of constant delivery. Ultimately, our best BI insights usually come when we step back from the dashboard and let the data exist. Customers actually generate better requirements when they have a chance to think a problem through, rather than receiving the instant gratification of a rapidly iterated deliverables. While I asked for this as a favor to the team's long-term success, the truth is that the time away is for them, not for me, giving them the space to decompress and return with the perspective that impactful data work demands. I'm curious, how do you help your high-performers set boundaries to prevent burnout? #Leadership #BusinessIntelligence #Amazon #TeamCulture #ProactiveLeadership

86

1 Comment

"Foundation before Intelligence." I proposed this tenet during our leadership team offsite last week. It became the most popular of the group, followed by "Reliability over Speed" (another one I proposed). To be clear, these are tenets, not absolutes. They're tie-breakers. Easier said than done with the efficiencies we're achieving with AI. I keep them on a post-it as a reminder because too often these days, teams dive into the latter before earning the former. Speed without reliability is rework. Intelligence without foundation is noise.

14

2 Comments

Over the past couple of years, I've had the privilege of working as a Data Engineer supporting Amazon's FlowMS (Flow Monitoring System) project - a critical initiative in our water stewardship efforts across fulfillment centers. At Amazon, we are scaling leak detection by using water flow data from our sites leveraging an Amazon-developed, machine learning-aided algorithm called FlowMS. FlowMS analyzes water flow patterns at a site to determine a baseline of water usage. If the water usage pattern in a building deviates from the established norm for a designated period of time, this could indicate a possible leak, and FlowMS sends an automated alert to on-site teams to investigate. The impact has been remarkable: In the period of July 2024 to June 2025, FlowMS detected 11 leaks at 79 monitored operations sites, saving 41 million liters of water annually. This saved water corresponds to approximately $250,000 in annual savings on water utility bills for these sites. Working with IoT sensor data from LoRaWAN devices, I've been part of a team helping Amazon track and optimize water consumption in real-time. The experience has been incredibly rewarding - from processing sensor telemetry data to building analytics frameworks that enable these life-saving detections. It's fulfilling to know that our work contributes to both environmental sustainability and operational efficiency at scale. Learn more about smart water metering programs in sustainability: https://lnkd.in/gvD7XN4H #Sustainability #WaterStewardship #IoT #SmartMetering #Amazon

34

Giving an agent more freedom requires clearer decisions from the person delegating the work. That starts with mapping the workflow and defining what the agent may change without approval. Clear boundaries let an agent check and revise its work against an agreed standard without waiting for someone to direct every attempt. That can reduce supervision while leaving people responsible for deciding whether the result is ready to use. Permission to do more should follow evidence that the agent can handle the work already entrusted to it. A decision outside its authority still needs to come back to a person. Every AI workflow needs a named human owner who can settle unresolved decisions and accept the result. That responsibility starts before the agent runs.

4

🌉 A few weeks ago I had the incredible opportunity to attend The AI Conference in San Francisco with 5,500+ builders, researchers and leaders coming together to talk agentic AI, AI infrastructure and governance. Some of my key takeaways: • The people who understand the business should be part of the AI governance loop. • Keep decisions and actions separate. PayPal's orchestrator agent "decides, never acts", and a default like approver: "none" quietly bypasses a required control, so it is treated as missing and escalated. • Make agents surface trade-offs instead of resolving them. • Static rules can't secure dynamic agents. An agent blocked from attaching a file to a PR created a public GitHub gist and posted the link instead. The rule matched the action but not the intention. • The worst bugs don't fail loudly, they succeed quietly and wrongly. • So start small: make one workflow repeatable, then increase autonomy as it proves reliable. Excited to apply these principles in my work! #TheAIConference #AI #AgenticAI #GenAI #AIAgents #AIGovernance #SanFrancisco #Amazon

59

In 15 years at Amazon, I had 18 managers. (Yes — 18.) Every new leader brought a different style, priorities, and definition of success. Early in my career, that constant change felt like disruption. Later, I realized it was actually the perfect training ground for adaptability and leadership through ambiguity. At Amazon, we say “Change is the norm — no change means something’s not right.” That mindset built my foundation as a leader who can stay grounded through shifting goals, evolving org structures, and now — the rapid rise of AI. Today, I see the same principle playing out globally. As AI reshapes roles, decision cycles, and business models, leaders who can absorb change and translate it into clarity for their teams will define the next decade. Adaptability isn’t just a survival skill anymore — it’s a leadership muscle. 👉 How do you help your teams stay adaptable through transformation?

97

14 Comments