The Robot Report Podcast · 12 Jun 2026 · 1h 13m
Akash Gupta, CEO and co-founder of GreyOrange, explains the company's strategic pivot from manufacturing AMRs to becoming a pure-play software provider for warehouse orchestration. He details how GreyMatter uses AI and machine learning to optimize fulfillment outcomes rather than just robotic uptime, integrating with various hardware vendors through a certification program. The discussion covers the development of digital twins, predictive analytics, and the use of large language models to enable intent-based warehouse management.
Where this conversation describes AI being used to build or deliver a product. Part of the AI in practice tracker.
Engineering & simulationAI coding tools90-95% of test cases written by AI
AI is used to write 90-95% of test cases and automation scripts, eliminating manual test case writing by developers.
“testing has become like today, almost like 90, 95% of our test case being written is being written through AI, right? There's no manual test cases that are getting written or automation that is getting written, you know, by the developers.”
Akash Gupta · Hear it at 45:16
Engineering & simulationAI coding tools10x faster to 80% completion
AI accelerates the initial development phase of features, reducing the time to reach 80% completion by 10x.
“I think 0 to 80% of any feature is become like 10 X foster, right? I think 80% to 100% still needs a lot of work, but I think 0 to 80% of any feature set is become much faster.”
Akash Gupta · Hear it at 45:16
Engineering & simulationAI coding tools5 algorithms per day
AI allows engineers to rapidly prototype and compare multiple algorithms, enabling them to try five different approaches per day.
“Other thing that has become pretty fast is trying different algorithms, right? Because, you know, earlier, your time to prototype, algorithm or train a model used to be much longer. Today you can literally try 5 of them par Lily and then go from 80 to 100% on the one that is showing good results…”
Akash Gupta · Hear it at 45:16
Engineering & simulationGrey Matter (GreyOrange)
Machine learning algorithms are used to learn inventory behavior, work patterns, and seasonality to make dynamic decisions on slotting, task allocation, and spatial storage without hard-coding rules.
“So it basically, you know, analyzes all of these parameters and makes sure that then the decisions around slotting decisions around task, you know, task allocation, decisions around, you know, just the spatial storage of inventory, you know, all these decisions gets taken through different machine learning, you know, algorithms…”
Akash Gupta · Hear it at 40:31
Engineering & simulationGrey Matter (GreyOrange)
Large language models are used to interpret natural language intent from warehouse managers, automatically adjusting system parameters to prioritize storage efficiency or throughput.
“And then you can say, you know what, I'm ready to compromise with storage efficiency. I need more throughput from the system and system can rejug itself for that. So I think through large language models, the intent can be, you know, conveyed in a lot more natural, you know, way…”
Akash Gupta · Hear it at 40:31
Engineering & simulationGrey Matter (GreyOrange)
Agentic AI capabilities allow the system to dynamically select the best solution method (e.g., solver vs. multivariable equation) for NP-hard problems based on the specific context and time.
“What you can do is you can literally have 4 or 5 different solution or different, you know, kind of ways of solving that problem and put a agent at the top, which can basically decide that at that particular time, which one of these solution is the better solution to apply to that problem…”
Akash Gupta · Hear it at 45:16
Marketing & salesFoundry (GreyOrange)10 scenarios in 2 days vs 2 months
Simulation and digital twin technology is used to run multiple warehouse scenarios rapidly, accelerating the sales process by providing objective boundaries for system performance.
“what you have seen as something like foundry is accelerating our sales process, um, you know, quite a bit because you are able to kind of run 10 different scenarios in 2 days there than 2 months, right?”
Akash Gupta · Hear it at 57:47
Running the companyGrey Matter (GreyOrange)
Predictive analytics and simulation are used to forecast congestion and proactively move inventory to prevent bottlenecks before they occur.
“what we do today is predict that, you know, even that's going to happen one hour later, and literally even right now, start moving some of that inventory in different aisles so that that doesn't happen, right?”
Akash Gupta · Hear it at 55:47
companies scaling up
Other shows on the same ground, closest first.
Joel Thomas and Markus Thiel from Siemens Intralogistics discuss the evolution of warehouse operations from manual handling to digital-first design, driven by e-commerce growth and labor shortages. They explain how Siemens uses its own European factory to test and refine technologies like AMRs, edge computing, and digital twins before offering them to customers. The conversation highlights the industry's shift toward standardization to manage multi-vendor fleets and the emerging role of software-defined automation and IT/OT convergence in solving engineering challenges.
Michael Patrick Perry, Dexterity’s VP of Marketing, discusses how the company has scaled from a small team to nearly 260 employees, moving from bread packing to full warehouse automation. He explains Dexterity’s software-first approach, which allows robots to handle complex, unstructured tasks like truck loading and palletizing by combining vision, force control, and planning. Perry highlights the democratization of robotics, enabling smaller companies to deploy flexible solutions quickly without massive capital expenditure, while also detailing how AI is used to create customized marketing experiences for enterprise clients.
Melonee Wise of Agility Robotics discusses the practical logistics applications of humanoid robots, emphasizing a tiered approach to complexity and the critical importance of new ISO safety standards for cooperative deployment. Glen Guernsey of JR Automation highlights advancements in warehouse automation, specifically random SKU palletization and automated truck loading/unloading, driven by labor shortages. Zoie Rittling of OnLogic and Michael Maxey of ZEDEDA explain the concept of 'right-sizing' AI at the edge to eliminate mundane operational tasks without over-engineering. Finally, Robert Ponsonby of Arduino details how open-source interoperability allows manufacturers to integrate legacy systems with modern cloud infrastructure, accelerating deployment and reducing costs.
Michael Perry, Vice President of Marketing at Dexterity, discusses how AI-enabled robots are transforming warehouse operations by handling variability that traditional automation cannot. He details Dexterity's platform, which integrates vision and touch to allow robots to pick, pack, and palletize mixed-SKU goods in existing 'brownfield' facilities without extensive infrastructure changes. Perry highlights a specific success story involving a bread fulfillment center where one robot managed over 10,000 SKUs, including deformable items like tortillas and loaves of bread, while also explaining the future of collaborative robot teams.
Adi Dalvi, VP of Sales at OSARO, explains how the company achieved calculated growth by prioritizing product robustness over speed, integrating vision software with articulating arm robots to automate warehouse piece-picking. He details OSARO's three-step criteria for selecting early customers—focusing on size, risk profile, and collaboration—and outlines the four distinct founder backgrounds (industry, startup, VC, and academia) that contribute to the company's success. Dalvi also shares strategies for extending runway through investor gate checks and negotiating case studies into initial contracts to build market credibility without heavy early marketing.
The summary, chapters and topics are generated from the episode’s audio by this directory, not written by The Robot Report Podcast. The episode itself belongs to its publisher.