When AI meets operational chaos, the happy path is off a cliff

Josh Clement-Sutcliffe is AI and Automation Product Manager at Zencargo, a London-based digital freight forwarder 

The standard advice is to pick a simple process, automate it, scale from there. Global freight forwarding doesn’t have a simple process. A single shipment can involve dozens of parties across multiple countries, each with their own regulatory requirements,systems, and standards. Communications still run on email, fax and phone. There’s no shared infrastructure or common data format, and nothing that makes one supplier’s documentation look like another’s. So there’s no playbook to reach for. It’s a business where  a container ship can block the Suez Canal, and every route, timeline and customer commitment has to be renegotiated overnight.

Zencargo’s early automation work targeted the happy path anyway, where everything arrives in the right format, from the right party, through the right channel. But automating that first, then expanding outward didn’t work. “The AI and automation just couldn’t figure out the context and what to do given the amount of variables which had to go into those particular processes.”

Josh describes falling into the sunk cost fallacy on a couple of projects, which is natural enough when there’s no industry precedent. The evidence eventually pointed one way: go back to the data.

Last August, they put a plan in place. First, data foundations: getting inputs clean enough for AI to act on. Second, visibility and trust: letting operators see what the system is doing. Third, what he calls partnership, where AI automates predictions, risks and optimisations at scale. 2025 was the foundational year. 2026 is about operationalising and scaling, with visibility and trust as the focus.

What Zencargo built and how it works

The data foundations stage meant solving a specific problem first: information arriving into Zencargo came as email attachments in dozens of formats, across every language, with no shared standards. Josh felt there was nothing in the market built to handle it. So they built it themselves: a product called Luca Scan, trained and fine-tuned on supply chain operations.

“Document parsing might not sound the most interesting, but when you are really working across so many different document types, across all languages across the globe, across different standards, it becomes very, very challenging.”

Luca Scan takes unstructured data from emails and attachments, makes it structured, and feeds it into Luca, the broader orchestration system he describes as “the brains of what we’re doing with AI within Zencargo.” From there, Luca reads intent and decides what needs to happen next. A booking arrives by email, goods from China to London, and Luca works out what’s required: goes to the booking service, checks whether it can progress or something’s missing, and either continues or sends a request for more information.

The current focus is inbound. Next is outbound: Luca recognising when a specific data point is needed before a container can be loaded and triggering the request itself. Beyond that, the same capability extending across whatever channels clients use, Teams, Slack, others.

Operators seeing it work before they’re asked to trust it

Operators who had previously read supplier emails, interpreted the contents and manually entered data into the platform found that work being done for them. Attachments went to the correct shipment automatically. Luca started making suggestions, corrections, potential optimisations, and operators could see exactly what it had done. The copying and pasting disappeared.

“It definitely becomes easier when it starts, or moves away from being theoretical. Starting to show the impact and making sure that you can see and actually crucially just trust that impact.”

But visibility creates its own problem.

“One of the lines which we’re having to tread lightly on is how do we define sending lots of information and noise to operators and then they’re just going to blank it out because it’s going to be too much – and then making sure we’re actually sending what is useful at the right time.”

Zencargo now has a lot of data. Getting it in front of the right people in a useful form is one of their main challenges for the year. Until then, much of the organisation is still working from what they’ve heard, not from what they can see.

Culture, hiring and the day-to-day reality

Zencargo has executive backing for AI, KPIs tracking adoption across departments, and a leadership team that gets involved when things stall. Josh says this makes his job a lot easier. The expectation from the top is clear: becoming the AI freight forwarder means using these tools across the business, not just in product and tech.

“At a high level, I say that it works well within the organisation. But of course, when it comes to day to day, you do have to bring people along. It’s just another example of a product manager role still being a product manager role, despite the new technology.”

Errors will happen. That’s accepted. Auditability is built in to catch mistakes early, before they compound. The problem predates the technology.

Zencargo has also changed how they hire: candidates are encouraged to use AI in technical tests, and the follow-up question is whether they can explain what’s happening under the hood. Curiosity about these tools is what they’re looking for across the organisation, not just in technical roles.

A case in point: one product manager used AI to teach themselves to code, and is now building an internal training academy on AI in freight forwarding. Find where it’s powerful, and follow it.

The Road to 2030

“When you realise you found something which could be really powerful, you’ll get that light bulb moment. And then if you can prove that value, that makes the transformational shift internally so much easier.”

For Zencargo, that moment came last August. The data foundations, the visibility work, partnership. The whole journey is headed toward an autonomous supply chain by 2030.