News / Logistics tech debate shifts from ‘build vs buy’ to ‘buy vs AI’, says WiseTech 

From left: Sushanth Raman; Ian Arroyo; Zubin Appoo; Eric Johnson

Artificial intelligence may soon write much of the code behind logistics software, but that does not mean freight companies will be able to rebuild complex platforms such as CargoWise overnight, industry leaders told delegates at TPM in Long Beach this week. 

WiseTech Global CEO Zubin Appoo said advances in AI coding tools were already transforming how logistics software is developed. 

“Writing code is no longer something humans do,” he said, arguing that the role of software engineers was shifting toward system design, domain expertise, and product thinking, rather than manual coding. 

The comments came shortly after WiseTech indicated it expected to reduce its product development and customer service teams by as much as half, over time – around 2,000 roles – as AI-driven productivity reshapes how the company builds and maintains software. 

According to Mr Appoo, the turning point came late last year when new AI models, designed specifically for software development, dramatically improved engineers’ ability to fix bugs, review code, and resolve system errors. 

But while AI may accelerate software creation, a discussion panel at the event suggested that replicating the complex digital infrastructure underpinning global freight operations was far more difficult than simply generating code. 

“The code is one-tenth of it,” Mr Appoo said. “Nine-tenths is dealing with the company, building the commercials, and building that relationship.” 

The distinction reflects a broader shift in how logistics companies are approaching technology. 

For decades, freight forwarders faced a straightforward decision when investing in systems: build their own software or buy a platform from a vendor. 

But with AI tools now able to generate working applications rapidly, the industry’s technology debate is evolving – “Maybe the title of this session should be ‘buy versus AI’,” Mr Appoo told the ‘Buy versus Build’ panel. 

In other words, logistics companies may increasingly combine established platforms with AI-built tools and automations, rather than replacing those platforms entirely. 

One reason lies in the scale of the ecosystems behind modern logistics software. 

Those integrations with carriers and forwarders, built over decades, underpin the digital infrastructure through which much of global trade flows. 

“Eighty percent of manufactured trade flows through CargoWise,” Mr Appoo said. 

Panelists noted that while AI may make it easier for companies to build niche applications or automate specific workflows, recreating such a network would require far more than software engineering. 

Connections with carriers, ports, and customs authorities often involve lengthy commercial negotiations, regulatory approvals and operational coordination across multiple organisations – the hardest challenge for potential rivals to WiseTech’s CargoWise. 

When building new integrations with shipping lines or government systems, Mr Appoo said, the technical coding involved represented only a small portion of the effort required. 

That complexity is one reason many logistics companies continue to rely on established platforms rather than building their own systems. 

Ian Arroyo, chief strategy officer of Freightos, said companies should focus their internal development efforts on areas where they create competitive advantage, while relying on shared platforms for industry-wide connectivity. 

“A carrier or forwarder should absolutely be building in the areas that create a unique value proposition,” he said. “But when it comes to the network… that’s where platforms provide value.” 

Another challenge for AI in logistics lies in the nature of the industry’s data. Unlike software development, where AI models can train on vast amounts of publicly available code, much of the knowledge required to run freight operations is not easily accessible. 

Instead it exists as undocumented processes, or “tribal knowledge” held by experienced staff. 

Sushanth Raman, founder of logistics AI company Pallet, said this lack of structured data could limit the effectiveness of AI systems in real-world freight operations. 

In one test discussed during the panel, an AI model processing a bill of lading achieved only around 69% accuracy – far below the level required for automated logistics workflows. 

“Getting something to 80% is easy,” he said. “Getting it to 99% is really, really hard.” 

That gap is critical in an industry where errors in customs filings, documentation, or regulatory declarations can result in costly delays or penalties. 

Panelists also warned that many logistics companies may not yet be ready to fully exploit AI tools. 

Legacy systems, fragmented data, and undocumented workflows remain common across the sector, making it difficult for AI to access the information needed to automate processes effectively. 

At the same time, global supply chains are becoming increasingly volatile, increasing the demand for real-time data and faster decision-making. 

One speaker described how a chief supply chain officer at a Fortune 100 company recently needed to quickly analyse sourcing options following a sudden market disruption, highlighting the growing importance of integrated data ecosystems. 

As a result, logistics software providers are increasingly positioning themselves not only as vendors but also as partners helping companies navigate AI-driven transformation. 

The panel suggested the future of logistics technology would likely combine established platforms that manage the industry’s complex web of integrations with AI tools that allow companies to automate tasks and customise workflows. 

For forwarders considering whether AI might allow them to replace existing systems entirely, the message from the panel was clear: AI may be able to write the code, but building the infrastructure of global logistics is another matter altogether.