FourKites unveils Loft AI platform for enterprise workflows

FourKites has launched Loft, a development and orchestration platform designed to generate and run custom business automations across enterprise systems using AI agents.

At the centre is Sophie, which FourKites describes as an AI developer agent that turns operational requirements written in natural language into automated workflows. FourKites teams review the automations before deployment.

FourKites is positioning Loft as a response to slow AI rollouts in large organisations and the effort required to maintain production systems. It said implementations that once took months can now be completed in days.

Agents in production

Agentic AI has become a major theme in enterprise technology, but there are still few examples of large-scale deployments beyond pilots. FourKites said it has spent the past 12 months running "Digital Workers" across Fortune 500 supply chains, focused on automating workflows rather than operating as chatbots.

FourKites cited customer results, saying Coca-Cola reduced inquiry response times from 90 minutes to seconds, and that First Solar now runs half its workflows autonomously.

Loft expands beyond supply chain software, according to FourKites, and is designed to orchestrate actions across enterprise applications, including ERP, IT service management, transport management, warehouse management, and CRM.

Platform approach

FourKites said Loft combines internal enterprise data with external intelligence from its Intelligent Network, which it described as drawing on more than 500,000 trading partners and millions of daily supply chain events.

The platform includes Agent Operating Procedures (AOPs), which FourKites said record why an agent made a decision and who approved it. The goal, it added, is to preserve organisational knowledge that might otherwise be buried in email threads and messaging apps.

FourKites also highlighted the use of external intelligence in decision-making, including supplier performance history across the network and real-time patterns from other customers, alongside precedents from similar situations.

"Most enterprise operations still run on fragmented systems held together by spreadsheets, shared inboxes, and email chains," said Josh Jewett, Operating Partner at NewRoad Capital Partners and former CIO of Dollar Tree and Family Dollar. "In that environment, critical decisions don't live in systems at all - they live in Slack threads and people's heads. When AI is layered on top of that fragmentation, it can observe problems but can't reliably act on them. That's why so many AI initiatives stall. The missing piece isn't just intelligence, it's a platform that turns decision logic into durable, reusable workflows that can actually run the business."

Scaling and maintenance

FourKites framed Loft as an answer to post-deployment challenges that can slow adoption, such as performance monitoring and model drift. It also argued that enterprises often face a choice between building internal teams to maintain systems or relying on vendor engineering capacity.

It also pointed to industry research on the difficulty of scaling AI, citing McKinsey & Company figures that 88% of organisations have deployed AI in some function while 7% have scaled it enterprise-wide. FourKites also cited a Gartner prediction that 40% of agentic AI projects will be abandoned by 2027 due to complexity and unclear returns.

In Loft's development flow, FourKites said Sophie evaluates whether an existing workflow can be configured, whether building blocks can be combined, or whether custom code is needed. Engineers then review the outcome before deployment, it said, and Sophie continues to monitor and improve performance over time.

"We didn't build AI features on top of legacy software. We built an AI-native system from the ground up," said Mathew Elenjickal, founder and CEO of FourKites. "When our AI agents do the work, we record how decisions were made - not just what happened, but the context, the prior cases that informed it, and who approved it. That reasoning doesn't live in your TMS or ERP. It's scattered across Slack threads, email chains, and people's heads. Until now."

Network data

FourKites is emphasising the role of external data in automation decisions, alongside internal systems of record. It said the Intelligent Network holds real-time performance information across 500,000 trading partners in 176 countries and processes about three million daily supply chain events.

"The real value starts with the network," said Charles Brennan, Senior Analyst at Nucleus Research. "FourKites' Intelligent Network provides the external, real-time data foundation that AI and automation depend on. Loft gives enterprises a practical way to operationalize that data by embedding it into governed, repeatable workflows that connect external supply chain conditions directly to internal systems and decisions."

FourKites said Loft also serves as the home for its Digital Workforce, which it described as specialised agents such as Tracy for logistics execution, Sam for supplier collaboration, and Alan for appointment scheduling. FourKites said these are already in use at dozens of Fortune 500 firms.

Finally, FourKites said Loft draws data from more than 200 transport management providers, ERP systems, and CRM platforms, and that custom agent development through Loft is now available.

"We are moving enterprises from dashboards that merely track problems to systems that autonomously solve them," said Elenjickal. "The goal isn't to drop AI into existing silos. It's to capture the reasoning that lets those systems work together, and to preserve it so each decision makes the next one easier."