Omni Talk Retail – Technology Spotlight Series
Episode Title: The Now, Next & Future Of AI's Impact On Warehouse Operations With Dematic's John Mabe
Release Date: October 6, 2025
Guests: Chris Walton (Co-host), Anne Mezzenga (Co-host), John Mabe (Product Manager, Dematic)

Episode Overview

This episode explores the evolving role of AI in warehouse operations. Chris Walton and Anne Mezzenga, the hosts, welcome John Mabe from Dematic to clarify how different types of AI are currently used, what practical impacts they have, and which innovations are on the near and distant horizon. The discussion is structured around three main AI categories: optimization AI, vision/perception AI, and large language models (LLMs). John provides insight into real-world adoption, limitations, and forecasts for the future—including the realistic timeline for humanoid robots.

Key Discussion Points & Insights

1. Defining AI in Warehousing

(02:05 – 02:56)
John Mabe: Emphasizes that AI is not a single technology but a toolkit with three main categories:

• Optimization AI ("the brains"): Decision-making for tasks like inventory management and labor forecasting.
• Vision and Perception AI ("the eyes"): Cameras and sensors interpreting real-time warehouse conditions.
• Generative AI / LLMs ("the interface"): Natural language interfaces (like ChatGPT) for querying warehouse systems and data.

"It's not just a single technology, it's more of a toolkit." – John Mabe (02:09)

2. Optimization AI: History, Use Cases, and Adoption

(02:56 – 06:55)

• Not New, But Evolving: Classic optimization has long been used for inventory slotting and labor planning (ABC classifications, rules-based systems).
• AI Adds Adaptivity: Machine learning enables systems to learn non-obvious patterns, predict velocity across SKUs (including slow or medium movers, seasonality), and respond to events like promotions or unexpected demand shifts.
• Limited Current Adoption:
  • Widespread use of mathematical optimization, but only larger players are fully leveraging true AI due to the need for structured, high-quality data.
  • Smaller companies are in the “crawl” phase, using AI for recommendations rather than full automation.

"Where I started to come in is adding better inputs and adaptivity... really truly predict the velocity of SKUs." – John Mabe (03:41)

"It's mostly the bigger players that are using it that have access to really good data and have their data structured in a way the AI can leverage it." – John Mabe (05:19)

3. Vision and Perception AI: Eyes of the Warehouse

(06:56 – 11:05)

• Current Deployment:
  • Used in process integrity (e.g., ensuring totes are positioned correctly), detecting jams, and monitoring for safety/ergonomics.
  • Mostly deployed as fixed-position cameras; also increasingly integrated into robots and AMRs.
• Pathway to Full Autonomy:
  • Vision AI is foundational for “lights out” (fully autonomous) warehouses.
  • Major challenge: Robots require split-second, complex understanding for tasks like picking—knowing how to handle different products safely and efficiently.

"It's really about how can we minimize errors that take a lot of time to fix." – John Mabe (07:16)

"It has to have a deep understanding of how to interact with that product." – John Mabe (10:10)

4. Robotics and the Humanoid Hype

(11:05 – 13:47)

• Humanoid Robots:
  • Exciting media coverage; potential because they can operate in human environments without major warehouse changes.
  • Technology is past the “sci-fi demo” stage—real pilots are running, mainly simple tasks for now.
• Timeline:
  • Increasing complexity of tasks expected over the next five years.
  • Evolution likened to the smartphone: general-purpose robots will slowly centralize many roles.

"When I was 10, like I watched the Jetsons all the time. I wanted Rosie to clean my room and I wanted a fly car... we're not there yet, but we are past that sci-fi demo stage." – John Mabe (12:10)

"General purpose usually wins in technology. So I feel like we'll get there. It will take time." – John Mabe (13:29)

5. Large Language Models (LLMs): The Interface Revolution

(14:20 – 15:55)

• New Frontier:
  • LLMs will dramatically reduce the time needed to gain insights: instead of manual analysis, users can pose questions like “Where’s the bottleneck today?” and receive synthesized, actionable answers, including visualizations.
  • LLMs can draw from disparate data sources, unlike traditional static dashboards.
• Faster Adoption:
  • Generative AI is software-based and easier to deploy quickly compared to physical automation.

"It takes a long time to get to insights and our feelings. This kind of LLM models, you'll be able to ask questions... and display that back to the user within seconds." – John Mabe (14:37)

6. The Road to Full AI-Driven Warehousing

(15:55 – 19:25)

• The Human-in-the-Loop Model:
  • Starting with decision support: AI recommends, humans approve.
  • Progresses to decision intelligence: AI acts with rules and confidence thresholds (e.g., "If confidence is above 90%, execute automatically").
  • The ultimate goal is a multi-agent, largely autonomous system, but always with human oversight for strategic direction and safety.
• Today’s Reality:
  • Pilots are underway—Dematic has real customers currently piloting AI forecasting models.
  • Decision support is already in play; transition to more autonomous decision intelligence is coming soon.

"The crawl side is more around decision support where a human is in the loop." – John Mabe (16:05)

"That's happening today..." – John Mabe (18:41)

7. The Adoption Timeline & Takeaways

(20:24 – 22:42)

• Logical AI Adoption Pathway:
  1. Optimization AI: Already established; start here for measurable ROI.
  2. Vision & Perception AI: Scaling in automated, less so in manual operations.
  3. LLMs/Generative AI: Rapid deployment anticipated—you can unlock value with software upgrades, not new equipment.
  4. Integration/“AI Brain”: All these elements will eventually converge into a single intelligent orchestration system.
• Strategic Reminder:
  • AI is a toolkit, not a monolith.
  • Rollout progresses stepwise, following ROI and data maturity.

"AI is not just a single technology. It's a toolkit... brains, eyes, and the way you can interact with your software." – John Mabe (20:24)

"Eventually all these pieces will come... and they'll all converge into a more single intelligent system." – John Mabe (21:27)

Notable Quotes & Memorable Moments

• On Defining AI Categories:
  "Optimization is the brains of your operation. Vision and perception are the eyes, and LLMs are the way you can interact with your software." (20:24)
• On Humanoid Robots:
  "We're past the sci-fi demo stage... real warehouse and manufacturing pilots happening today." (12:10)
• Human-in-the-Loop:
  "You need to build trust and validate the technology... guardrails need to be in place for safety." (17:54)
• On Pace of LLM Adoption:
  "...the last part of this, the LLMs, might actually take off faster than everything else, which has been in place for a while, which also makes sense because it's... software." (19:44, Chris Walton)
• Overarching Vision:
  "All these elements will converge into a more single intelligent system… orchestrating all the processes in the warehouse." (21:27)

Timestamps for Key Segments

• 02:05 – John defines AI categories for warehouse operations.
• 03:24 – Discussion of traditional vs AI-driven optimization and adoption by large vs small players.
• 06:56 – Introduction of vision/perception AI: current uses and future possibilities.
• 09:03 – Fixed vs robot-mounted vision systems, robotics landscape.
• 11:45 – Humanoid robots: realistic expectations and timeframe.
• 14:20 – The role and future of LLMs in warehouses.
• 15:55 – Decision support, human-in-the-loop, and roadmap to autonomy.
• 18:41 – Current pilot projects in forecasting and decision support.
• 20:24 – Summary and logical order of AI technology adoption.

Conclusion

John Mabe breaks down complex AI terminology and demystifies its real-world application and trajectory in warehousing. The conversation offers a balanced outlook: optimization and computer vision are steadily advancing (with practical, quantifiable value—especially for larger, data-mature operators), while generative AI (LLMs) is poised for rapid, near-term impact due to its flexibility and ease of integration.

Final Thoughts:
AI in warehousing is a journey, not a simple switch. Each technology—optimization, vision, and LLMs—will continue to become more robust and connected, eventually forming a cohesive, largely autonomous system but always with human oversight and strategic control.

Contact:
John Mabe welcomes follow-up via email (johnmabe@dematic.com) or LinkedIn.