Why Transportation Needs Both AI and Traditional Algorithms
Half the “AI” solutions I see in transportation are just rebranded rule engines. The other half are using machine learning to solve problems that algebra could handle better.
Last week, a client forwarded me this TruckingDive article about how AI is revolutionizing freight brokers, with a simple question that cuts to the heart of the issue: “Is what we’re using automation or AI?”
It’s a question I hear constantly from transportation executives. And honestly? It’s the right question to ask—because the answer determines whether you’re investing wisely or throwing money at marketing buzzwords.
The TruckingDive article highlighted five ways brokers are using AI: freight matching, automated RFP quoting, carrier recommendations, customer service chatbots, and trailer pool management. All real applications, all happening today. But here’s what four years of building transportation software has taught me: not all “AI” is created equal, and sometimes the old tools are still the best tools.
Where AI actually shines
Let’s start with what the current generation of AI (Large Language Models like Claude and GPT-4) does brilliantly:
Parsing messy data (like C.H. Robinson’s email parsing): Every transportation company knows this pain. You get bid sheets in Excel, PDF, handwritten notes, emails with tables that don’t quite line up. LLMs can read all of this and extract the important information. No more manually copying data from cell to cell.
Natural language interfaces (the chatbot example): Asking “What were my top 10 lanes last quarter?” in plain English beats clicking through five different reports. This is where chatbots and conversational interfaces genuinely improve workflows.
Carrier recommendations: AI can translate historical data into insights—carriers that perform well on particular lanes, carriers that are reliable during peak season, carriers that can best handle specific equipment types. This pattern matching is right in AI’s wheelhouse.
Summarizing and explaining: After your optimization engine spits out a complex solution, AI can explain in plain English why it recommended specific lane combinations. This bridges the gap between data scientists and business users.
Where traditional algorithms still win
Here’s where I might disagree with some of the article’s implications: for the heavy mathematical lifting in transportation (like freight matching optimization and trailer positioning), traditional algorithms still crush AI.
I’ve tested every new AI model that comes out—GPT-4o, Claude Opus 4, the new “reasoning” models—giving them real freight matching problems. The results? They can suggest approaches, but they consistently fail at finding optimal solutions for multi-constraint routing problems.
My favorite example: An AI model once recommended routing a truck from Michigan to New York by going through Canada—technically shorter distance, but it forgot about this little thing called customs. While creative, adding 6 hours of border crossings to save 50 miles isn’t exactly optimal.
For tasks like network optimization, demand forecasting, route planning with multiple constraints, load consolidation, and trailer pool positioning, purpose-built algorithms (Mixed Integer Linear Programming, metaheuristics like ant colony optimization, specialized time series models) still outperform LLMs by a wide margin.
The future: AI as the conductor
The real breakthrough isn’t making AI do everything—it’s using AI as an intelligent coordinator. Researchers call this “agentic AI,” and it’s where we’re placing our bets.
Think of AI as a Swiss Army knife and traditional algorithms as specialized power tools. The Swiss Army knife is incredibly versatile and perfect for many tasks, but when you need to drive screws all day, you want a dedicated power drill. The magic happens when you have both in your toolkit and know when to use each.
Imagine AI as a conductor of an orchestra:
- It understands your request in plain English
- It knows which specialized tool to call (optimization engine for routing, time series model for forecasting)
- It orchestrates these tools to solve your problem
- It explains the results in terms you care about
This isn’t just theory. We’re building this today at EnrouteAI. When you upload a messy bid sheet, AI reads and structures it. When you need optimal lane selection, traditional algorithms do the math. When you need to understand why certain lanes were chosen, AI translates the results into business insights.
So, is it automation or AI?
Back to my client’s question. The answer is: it’s both, and that’s exactly how it should be.
The TruckingDive article is right about where AI is being deployed today. But the nuance is in understanding which tool to use when. The companies that will win are those that understand this distinction. They’ll use AI where it excels (language, pattern recognition, interface) and traditional algorithms where they excel (optimization, forecasting, mathematical precision).
What this means for carriers and shippers
If you’re evaluating AI tools for your transportation operations, ask:
- What specific problem does this solve?
- Is this problem language-based or math-based?
- Can I explain the AI’s decision to my team and customers?
- Am I paying for genuine intelligence or just marketing buzz?
The hype around AI is real, but so are the benefits—when applied correctly. Don’t let vendors convince you that AI is magic. But also don’t dismiss it because it can’t solve every problem.
The future of transportation tech isn’t AI or traditional algorithms. It’s AI and traditional algorithms, working together.
Technical note: For those interested in the details: The current generation of AI (Large Language Models) are essentially very sophisticated pattern matchers trained on text. They excel at language tasks but struggle with precise mathematical reasoning. That’s why for forecasting, traditional time series models (ARIMA, Prophet, etc.) still outperform LLMs—they’re built specifically for that mathematical task. Similarly, for network optimization, algorithms like Mixed Integer Linear Programming will find mathematically optimal solutions that LLMs can only approximate. The magic happens when we use LLMs to make these powerful but complex tools accessible to everyone through natural language interfaces.
Neil Fernandes is the technical founder of EnrouteAI. He spends way too much time testing AI models on freight optimization problems.