TOC Europe 2026, held at the Hamburg Messe from 19 to 21 May, opened on what has become a familiar theme: AI sitting at the top of every terminal operator’s agenda, and the pressure to act being felt from boards, customers and competitors alike. The harder question, posed by speaker after speaker, was what to actually do about it.
Subbu Bhat, interim president and head of engineering at Tideworks Technology, was willing to say what most of the marketing would not. “Using LLMs per se in day-to-day decisions, particularly in an automated fashion — I have not seen it at scale yet,” he told Container Management in Hamburg. “It’s been mostly in experimentation, in assist or recommend mode, potentially. Not completely integrated into operations yet.”
That calibration framed the longer conversation about what AI in container terminals actually requires — and how Tideworks, the technology arm spun out of Carrix some 25 years ago, is repositioning to deliver it.
Five principles for trustworthy adoption
Bhat brought to TOC Europe a developed framework for AI adoption built around five principles. The first is to audit before automating. Periodic time studies — asking staff what they are working on, for which customer, and how long it will take — surface where automation will deliver returns and where, more often, a broken process should be fixed first. No new technology required.
The second is to treat data quality as a prerequisite. “AI does not filter out noise; it amplifies it,” he said. Tideworks’ own market data makes the point in numbers. In a survey the company conducted with Port Technology International, covering 121 terminal professionals across container, bulk and ro-ro operations and annual throughputs from under 100,000 teu to more than five million, 86% of respondents said they used TOS and planning tools, but only 30% leveraged real-time analytics. Fifty-three percent reported internal integration challenges; 46% reported external ones; 58% still relied on manual data practices. Against that, AI interest was running well ahead of the foundations needed to support it — 43% of terminals overall were prioritising AI investment, and among large terminals over a million teu that rose to 64%.
The third principle is the one he returned to most frequently in conversation: sequence adoption as “assist, recommend, automate.” AI starts by aggregating data a planner cannot compile fast enough — past voyages, weather patterns, crane availability, gate activity — and surfacing a recommendation in seconds. The planner decides. As trust accumulates, the model is allowed earlier and more forward-looking guidance. Only then, and only on specific, well-defined tasks, is automation applied. “Making decisions [with AI] is a little far away,” Bhat said.
The fourth principle is to keep humans in the decision loop — framed as accelerator rather than brake on adoption. The fifth is explainability: a black-box recommendation leaves a planner with only the choice to accept or reject, while a transparent system gives them grounds to assess, challenge and act with confidence. Bhat is careful with the noun.
“I like the word confidence better,” he said when asked about operator trust in AI. “There has to be confidence, and that’s when it will adopt at scale.”
In conversation, Bhat condenses the same logic into three operating prerequisites every terminal needs in place before AI can scale: “clean data, trustworthy data; an interconnected system; and the operators with the operational excellence and the data-driven decision mentality. Those three pillars have to be in place before AI can scale,” he said. Most customer conversations, he added, are not about enabling AI at all. “It is more about getting those fundamentals in place.”
From data platform to data layer
The framework points to a quieter but significant repositioning. In 2024, Tideworks moved toward exposing a TOS data platform. By 2025, that strategy had changed.
“We did go down the path of potentially exposing our own data platform in the 2024 timeframe,” Bhat said. “We have since pivoted to a slightly different strategy because what we have found is that customers are not interested in a TOS data platform. What they want is an enterprise data platform for their enterprise. And we don’t have all the data that is relevant for their enterprise.”
The admission is grounded in scope. Tideworks holds inventory, moves, yard layout and container locations — but not labour rosters, equipment maintenance schedules or vacation cover. Larger operators want a platform that spans those domains, and Tideworks cannot be that platform.
The new model recasts Tideworks as a clean data source and an execution endpoint. “We are going to give them the data access to our system, to our entities, via APIs or events. They can ingest it into their data platform; they can make use of that for analytics; if they want to build their own LLMs, whatever they want to do,” Bhat said. Intelligence built outside the TOS can be sent back in. “They can ingest [decisions] back to us via API, which will then influence the execution on the terminal.”
Asked directly whether the shift was a course correction, Bhat agreed. “The company is learning and evolving,” he said, pointing to Tideworks’ origin inside Carrix’s IT group, where bespoke implementations carried a heavy services tail, as the pattern the company has been steadily dismantling. The extensibility layer that now sits between a stable core TOS API and the integration realities of Konecranes, ABB, Siemens and other hardware vendors is the product expression of that shift. “We want to be a software vendor, not in the services business,” Bhat said.
Commercial reality: AI as factor, not line item
For all the conference-floor noise, AI is not a separate line item in Tideworks contracts today, and Bhat is direct about it. “Our algorithms, AI and optimisation capabilities today are typically embedded within the broader platform rather than sold as standalone line items. Customers are ultimately investing in the operational value the software delivers, with AI increasingly shaping discussions around planning, visibility and decision-support capabilities,” he said. “While AI is not typically a separate pricing component today, it is becoming an important factor in expansion and renewal discussions, particularly as customers look to unlock greater efficiency and insight across their operations.”
What that looks like in practice is bounded, concrete deployments rather than headline-grabbing autonomy. The most-cited example is a pilot in Panama in which a machine-learning model estimates container pickup time. The system draws on past vessel calls, gate-appointment patterns and trucker completion rates, refining its probability window as new information arrives. When a gate appointment is made against a container and historical data shows 80% of truckers meet their appointments, the system reflects that in real time. The output is a better-informed planner, not a replaced one.
“The Panama initiative is a pilot focused on improving truck pickup-time predictability using operational data. Our analysis based on historical data shows promising results in ETA accuracy and operational visibility. We are exploring additional AI/ML-assisted use cases with other customers as part of broader efforts to improve operational efficiency. We have also enabled AI-based chatbots within our support portal, enabling customers to find answers more quickly while reducing the volume of inbound support enquiries,” Bhat said.
A second example is automated gate processing that Tideworks has deployed at West Coast terminals. Appointment data is integrated with internal systems so that gate transactions which would once have required a clerk are now completed without manual intervention for the majority of standard cases. Exceptions still go to humans; the automation handles what is routine and predictable. Both deployments sit firmly inside the assist phase of Bhat’s sequence — useful, in production, and a long way from autonomous decision-making.
Where Tideworks meets the market
Tideworks does not chase every deal. Bhat is direct about the customer segment: small and medium terminals, many under a million teu, with a meaningful number below 500,000. The differentiator is not price. “You have to compete on the value you bring to the table,” he said. Customer support — 24/7, relationship-driven — is, in his telling, the consistent reason customers stay.
Where the company does meet larger competitors, the win condition is specific. “We typically encounter Navis and Kaleris in marine terminal modernisation initiatives. Tideworks differentiates through a flexible product suite that can be configured to each terminal’s specific operational needs and processes, paired with a highly engaged, hands-on customer partnership and support model. As a result, operators often choose Tideworks when they are looking for a true technology partnership, rather than a traditional client–vendor relationship,” Bhat said.
The other side of AI
Bhat spoke on the TECH TOC theatre panel “The other side of AI: Navigating opportunities, risks and reality.” On the evidence of the conversation around it, his contribution did little to add to the hype.
“I think this industry has real hunger for more data, for a more data-driven approach to make decisions,” he said. “But we have to get incrementally better. Otherwise you end up doing things that may not be the right things to do.”
For a vendor whose marketing peers spent the same week claiming AI-driven autonomy is around the corner, that line is, in its own quiet way, a competitive position.