If you’re like many logistics executives, you’re likely very curious and interested in how AI can improve your logistics operations and deliver business value, but you’re also probably a bit skeptical or confused by all the noise and buzz in the market. Therefore, it’s not surprising that many companies are still in the “early awareness” stage — that is, they’re trying to learn as much as possible to figure out where and how to get started with AI in logistics.
Pando.ai and JBF Consulting just published a research report titled, “The State of AI in Logistics 2025” with data and insights they gathered from in-depth interviews with senior supply chain and logistics executives across several industries.
Abhi Manohar, CTO and Co-Founder of Pando.ai, and Mike Mulqueen, Executive Principal for Strategy & Innovation at JBF Consulting joined me on a Talking Logistics episode to explore some of the research results and key takeaways.
The Evolution of AI
One of the misconceptions in the market is that AI is something completely new, but in reality, aspects of AI have been used in supply chain and logistics for decades. The report discusses this “evolution of AI” in terms of five evolutionary stages. I asked Mike to walk us through those stages and where we are today.
Mike began by clarifying that “AI is a lot of different things” and that even early heuristics fall under the AI umbrella. The first stage, Automation, involves basic rule-based automation. “We would get a bunch of orders in and build them into routes on old desktops running DOS,” Mike said, recalling his days working at a route optimization company. He traced these early techniques back to the 1950s.
The second stage is Predictive, which uses historical data to forecast demand. “Fantastic demand planning applications [like Mangustics back when I worked there] would be able to predict demand based on patterns,” he noted.
The third generation — Prescriptive — builds on the previous stages of automation and predictive. “It’s not just about knowing what will happen, but deciding what to do about it,” says Mike. “For example, if five trucks are predicted to arrive at a facility at the same time, a prescriptive system might recommend contacting trucks A and B and asking them to delay their arrival by two hours based on their contents. That’s a simple logistics example of prescriptive AI in action. It’s where AI starts making actionable decisions, not just forecasts.”
The fourth stage is Generative AI, where large language models like ChatGPT and Gemini come into play. Mike sees this as the first truly disruptive leap: “Up until that point, they were evolutionary. I think the Gen AI piece has turned everything revolutionary.”
Finally, Mike described Agentic AI as the emerging fifth phase. Unlike Generative AI, which waits for user prompts, Agentic AI is proactive. “It doesn’t necessarily require any input from me,” explained Mike. “But we’re still in the early stages of Agentic AI, and I think we’ll see a lot of guardrails and manual oversight at first to ensure it’s doing what we want it to do. But this is the next phase — and it’s a truly exciting one.”
What Factors Are Driving Increase in AI Investments?
The research revealed that 75% of companies plan to increase their investments in AI significantly over the next two years. I asked Abhi what factors are driving this acceleration.
Abhi identified talent burnout and knowledge bottlenecks in logistics organizations as key drivers. “The biggest constraint is that the knowledge is proprietary and it’s held by very, very few people in the organization,” he said. When those individuals leave or change roles, it becomes difficult to transfer their expertise.
AI, he explained, offers a way to capture and scale that knowledge. “Imagine a logistics user asks an agent for something, and if the response doesn’t match what they expect, they nudge it — ask more questions, give feedback. Over time, the agent starts to understand the mindset of the user.” In this way, AI can help alleviate or remove the dependency on individual knowledge holders.
Abhi also mentioned that growing trust in AI — especially as models become more explainable — will further drive adoption: “Being able to articulate the exact reason for why the agent made a decision will help with adoption.”
Mike, Abhi, and I covered a lot more ground in our conversation, including:
- What are some real-world use cases where AI agents are gaining traction?
- If I’m a logistics executive and I need to build the business case for AI in logistics, what are the key metrics I should focus on? Is it just about cost reduction?
- Looking ahead to 2026 and beyond, what excites you the most about the future of AI in logistics? What does the next stage of evolution look like?
Therefore, I encourage you to watch the full episode — and download the report — for all their insights and advice on this topic. Then post a comment and share your perspective about the role of AI in logistics!
