Legacy TMS Modernization and AI Dispatch Intelligence for a Logistics Company

Client Background

The client was a mid-sized logistics and supply chain company operating across third-party logistics, port drayage, warehouse distribution, and intermodal transportation.

The company relied on multiple core business systems, including TMS, WMS, dispatch tools, warehouse systems, and settlement modules. Some systems had been deployed years earlier and were built on complex legacy architectures with limited data interfaces. Parts of the environment still depended on AS400 / IBM i or highly customized relational databases.

These systems supported daily order intake, dispatching, warehouse operations, transportation execution, and settlement workflows.

However, they did not provide a unified operational data layer.

When port congestion, chassis shortages, warehouse dock delays, or lane disruptions occurred, management struggled to see the full cross-functional operating picture in real time.

To protect client confidentiality, company identifiers, system architecture, transportation data, warehouse data, and capacity network information have been anonymized and sanitized. This case study is based on real enterprise AI delivery experience and presented through a representative logistics and supply chain operations scenario.

The Challenge

The client did not lack operational systems.

The real issue was that critical data was locked inside multiple legacy systems and could not be brought together quickly at the decision layer.

For traditional logistics companies, the digital challenge is often not whether individual systems still work. It is whether those systems can work together when operational pressure is high.

Legacy Systems Ran the Business but Were Hard to Extend

The client’s TMS and WMS were deeply embedded in daily operations.

They handled orders, dispatching, inventory, warehouse activity, and settlement data. Replacing them outright would have been expensive and risky.

A full “rip and replace” project could create downtime, migration risk, and user adoption challenges.

The client wanted to keep its core systems while adding modern data and AI capabilities.

TMS and WMS Data Were Disconnected

The transportation system contained vehicle, driver, load, route, and ETA data.

The warehouse system contained inventory, dock assignments, inbound and outbound activity, storage locations, and order status.

In practice, dispatchers and managers could not easily see how transportation status and warehouse status affected each other in real time.

For example, a truck’s estimated arrival time and warehouse dock availability were often managed separately, making it difficult for transportation and warehouse teams to coordinate ahead of time.

Port Congestion and Turn Time Needed Earlier Warning

In port drayage and cross-border logistics, port turn time, chassis availability, yard congestion, and container release status directly affect cost.

If the company relied only on after-the-fact reports, dispatch teams could not respond before costs escalated.

The client needed earlier visibility into issues such as rising average turn time at a port, increasing chassis split charges, or growing delays on a specific lane.

Management Could Not Ask Complex Operational Questions Easily

Executives and operations leaders frequently needed to answer cross-system questions, such as:

• Which port has the highest turn time this week?
• Which lanes are seeing abnormal detention or demurrage charges?
• Which warehouses are approaching dock capacity limits?
• Which customer orders are most affected by port congestion?
• Should some volume be redirected to another port or warehouse?

Previously, these questions often required IT or data teams to pull data, write SQL, and prepare reports manually.

The response cycle was too slow for fast-moving logistics decisions.

What ZenAI Built

This project was not about replacing the client’s TMS or WMS.

The goal was to build a non-disruptive modern data and AI dispatch intelligence layer on top of existing systems.

ZenAI combined non-intrusive data capture, a modern data lakehouse, an AI-native semantic layer, and a metadata RAG copilot to connect transportation, warehouse, port, and capacity data without disrupting core operations.

The system allowed management to ask operational questions in natural language and receive data-backed dispatch and routing insights.

1. Non-Intrusive Data Capture

ZenAI first assessed the client’s legacy system architecture.

For older systems running on AS400 / IBM i, DB2, or highly customized databases, ZenAI used non-intrusive change data capture, or CDC, to extract operational data.

Instead of putting heavy query load on fragile production systems, the system captured data changes through logs or change records.

The platform synchronized:

• Transportation tasks
• Vehicle and driver status
• Order status
• Warehouse inventory
• Dock assignments
• Inbound and outbound activity
• Port pickup and container status
• Fee and settlement data

This gave the client near-real-time data flow while reducing impact on core operational systems.

2. Modern Data Lakehouse

Data extracted from legacy systems flowed into a modern data lakehouse.

ZenAI designed a data layer based on Snowflake, ClickHouse, or a similar architecture depending on the client’s environment.

The lakehouse unified transportation, warehouse, order, port, and financial data that had previously been fragmented across TMS, WMS, and settlement systems.

This foundation made cross-system analysis, AI querying, and dispatch optimization possible.

3. AI-Native Semantic Layer

Logistics databases are often highly customized and difficult to interpret.

A real database may contain years of historical tables, status fields, internal codes, and business-specific logic.

If a large language model is allowed to guess the database structure directly, it can generate unreliable queries and incorrect conclusions.

ZenAI built an AI-native semantic layer that converted low-level database fields into business-approved metrics and dimensions.

Examples included:

• Average port turn time
• Chassis split rate
• Dock utilization
• Order delay rate
• Customer-level detention cost
• Warehouse inventory turnover
• Lane on-time performance

The semantic layer gave AI a reliable set of metrics to work with instead of forcing it to guess database fields.

4. Metadata RAG Copilot

ZenAI built a metadata RAG copilot on top of the semantic layer.

Executives and operations teams could ask questions in natural language, such as:

“Compare this week’s average turn time at LA / LB and Oakland, and explain what is driving the increase in chassis split charges.”

The system retrieved metric definitions, data lineage, and business rules, then generated a reliable query.

The results were explained in natural language and linked back to the relevant data definitions.

This reduced the need to wait for IT teams to prepare one-off reports.

5. AI Dispatch Intelligence

On top of the data and semantic layers, ZenAI configured AI dispatch intelligence capabilities.

The system could analyze the impact of different operational strategies using port, warehouse, capacity, and order data.

For example, it could help evaluate:

• Whether some volume should be shifted to another port
• Which orders could be rerouted
• Which warehouses could absorb redistributed freight
• Whether chassis capacity was sufficient
• Whether a change would increase inland transportation cost
• Which customer orders should be prioritized

AI did not replace dispatchers. It provided analysis, risk signals, and recommended options.

6. Executive Operations Dashboard

ZenAI built an operations dashboard for management.

The dashboard covered:

• Port turn time
• Chassis utilization
• Transportation task status
• Warehouse inventory
• Dock utilization
• Detention and demurrage
• Customer order risk
• Lane on-time performance
• Exception trend alerts

Management could drill down from high-level metrics into specific orders, routes, or warehouses, improving operational visibility.

How the Platform Worked

The system was designed around cross-system logistics analysis and dispatch decision workflows.

Phase 1: Legacy System Data Access

The system extracted data from TMS, WMS, dispatch tools, and settlement modules through CDC or read-only interfaces.

The synchronization process was designed to minimize impact on frontline operational systems.

Phase 2: Data Lakehouse Integration

Transportation, warehouse, order, port, and cost data flowed into a unified data lakehouse.

The system cleaned, linked, and standardized the data.

Phase 3: Semantic Metric Modeling

ZenAI worked with the client’s operations team to define metrics and business rules.

Low-level database fields were packaged into understandable and reusable semantic metrics.

Phase 4: Natural Language Analysis

Business users could ask operational questions in natural language.

The system used metadata RAG to retrieve relevant metric definitions and data logic, then generated the appropriate query.

Phase 5: Dispatch Impact Analysis

The system combined port, warehouse, capacity, and order data to analyze how current exceptions affected cost and delivery.

AI helped generate possible dispatch or routing adjustment recommendations.

Phase 6: Human Decision and Execution

Final dispatch decisions remained with operations leaders and dispatch teams.

AI provided data support and recommendations. Humans kept final control.

Project Snapshot

Key Changes

• Data synchronization: Critical operating data moved from T+1 reporting delay to minute-level synchronization.
• System modernization: The client kept its existing TMS and WMS while adding a modern data and AI analysis layer.
• Natural language analysis: Management could ask complex operational questions using natural language.
• Dispatch intelligence: The system provided adjustment suggestions based on port, warehouse, and capacity data.
• Risk control: Core business systems were not replaced, and data extraction was designed to be non-intrusive.

Core Technologies Used

ZenAI combined legacy system modernization, data lakehouse architecture, and enterprise RAG.

The project involved:

• AS400 / IBM i legacy system analysis
• DB2 data integration
• Non-intrusive CDC data capture
• Data lakehouse architecture
• TMS / WMS data integration
• AI-native semantic layer
• Metadata RAG
• Natural language BI
• AI dispatch intelligence
• Permission control and audit mechanisms

Business Impact

The project helped the client gain more real-time, unified, and intelligent operational decision-making capability without replacing core systems.

Data Visibility Improved

Before the platform, management relied heavily on T+1 or monthly reports to understand operational performance.

After implementation, key transportation, warehouse, and cost data could be synchronized at the minute level.

This helped the client detect port congestion, fee anomalies, and capacity risks earlier.

Legacy Systems Gained Modern Data Capabilities

The client did not need to replace its existing TMS and WMS.

ZenAI created a non-intrusive data layer that extracted key data from legacy systems and connected it to a modern lakehouse and AI analysis layer.

This reduced replacement risk and protected the client’s existing IT investment.

Management Queries Became Faster

Previously, when executives asked a cross-system operational question, IT or data teams often had to write SQL, run reports, and prepare results manually.

With the new platform, business users could ask questions in natural language.

The system generated queries based on the semantic layer and returned results with business logic explanations.

Dispatch Decisions Became More Data-Driven

The AI dispatch intelligence layer could analyze the impact of different options using port, warehouse, order, and capacity data.

For example, when turn time increased at a port, the system could help evaluate whether some freight should be redirected to another port or warehouse.

This helped dispatch decisions move from experience-driven judgment toward data-supported operations.

Cost Anomalies Became Easier to Detect

The system continuously monitored metrics such as chassis splits, detention, demurrage, lane delays, and warehouse congestion.

When abnormal trends appeared, management could intervene earlier instead of discovering the issue after month-end settlement.

Why This Project Mattered

For logistics and supply chain companies, competitive advantage is not only about trucks, warehouses, or customer relationships.

It is also about data response speed.

When port congestion, capacity volatility, and warehouse pressure happen at the same time, companies need to understand current conditions quickly and adjust operations with confidence.

ZenAI did not replace the client’s core systems.

It built a modern data and AI decision layer around them, allowing the client to keep operational stability while gaining near-real-time visibility, natural language analytics, and AI-assisted dispatch intelligence.

Frequently Asked Questions

Does this system replace the existing TMS or WMS?

No.

The solution usually works as a modern data and AI analysis layer connected to existing TMS, WMS, dispatch, and settlement systems. It does not require replacing the client’s core operating systems.

What is non-intrusive CDC?

CDC stands for change data capture.

Non-intrusive CDC synchronizes data through database logs or change records, reducing the need to place heavy query load on legacy systems. It is often useful for legacy modernization projects.

Why is a semantic layer needed?

Logistics databases are complex and highly customized.

A semantic layer turns low-level fields into business-approved metrics, reducing the risk of incorrect AI-generated queries and inconsistent reporting logic.

Can management really query operations data in natural language?

Yes.

With a semantic layer and metadata RAG, the system can translate natural language questions into reliable data queries and return explainable results.

Can this be deployed privately?

Yes.

For companies handling capacity networks, customer orders, warehouse data, and settlement information, ZenAI can design private or private cloud deployment architectures.

Build a Modern Data and AI Dispatch Layer for Your Logistics Operations

If your team is struggling with legacy TMS / WMS systems, data silos, port congestion, dispatch delays, or cost anomalies, ZenAI can help you build a secure, controllable, production-ready data and AI dispatch intelligence platform without replacing your core systems.

Explore more ZenAI case studies, learn more about ZenAI, or contact us through the ZenAI website to discuss your project.