Tamim Fannoush is the founder of Deep Current AS.

On 24 September, I spoke on the main stage at the 1st ALP Virtual Camp. My session, “AI in Logistics: From Documentation to Operations,” was built around one question:

Where do freight forwarding problems actually begin?

My thanks go to Aydin Caliskan, Founder of All Logistics People (ALP), for inviting me to speak. It was an engaging hour, and I wanted to share not only what I presented, but also some of the thinking the discussion reinforced for me.

The more I spoke about what we are building at Deep Current, the more I found myself coming back to a simpler question:

What actually needs to change before AI can become part of everyday logistics operations?

That question has been at the centre of our thinking for some time. It is also the reason we published our Levers of Digital Sophistication report earlier this year.

The report made a distinction that I still believe is important: the gap in logistics is no longer really about whether companies are interested in AI. It is about whether they can operationalise it.

A forwarder’s day starts in the inbox

Ask any forwarding team how their morning begins and you will hear a familiar story. Bills of lading, invoices and customs documents are checked and re-keyed by hand. Customers chase by email and phone for updates the team has already sent. Skilled people who should be planning and selling spend their days firefighting.

My starting point for the session was this: Delays, rework and compliance risk often begin upstream, in the emails and documents that feed every shipment.

If the data entering an operation is broken, adding intelligence further downstream doesn’t solve the root problem. This is the same operational reality that sits behind our first lever of digital sophistication: integrated digital foundations. (https://deepcurrent.no/five-levers-of-digital-sophistication/)

Where the data breaks

Every manual touch between the inbox and the system is a chance for a mistake to slip through, and errors get more expensive the further they travel. The problem is not necessarily that companies lack systems. It is that information still has to be manually moved reliably between them.

When people become the bridge between emails, documents, spreadsheets and operational systems, they effectively become the integration layer.

That was one of the strongest ideas in our Levers of Digital Sophistication research, and the ALP conversation reinforced it for me: before we ask AI to make better decisions, we need to give it access to the accurate data sets and vetted information.

Hype versus reality

I wanted the ALP community to leave with a realistic view of what AI does today, so I split it into two lists.

AI handles these tasks well:

• Reading and extracting data from documents
• Checking documents against rules and requirements
• Answering routine customer questions, 24/7
• Asking for missing shipment details

People stay in charge of:

• Exceptions and edge cases
• Commercial judgement and pricing
• Customer relationships
• Final sign-off where it matters

I don’t see operational AI as a story about removing people from logistics. I see it as a way of removing unnecessary work from the people who understand logistics.

The more repetitive checking, searching and data handling we can take away, the more time experienced forwarders have for exceptions, judgement and customer relationships.

That thinking is also central to the fifth lever in our research: human-AI partnership and governance. As AI moves closer to real operational decisions, organisations need clarity about where AI can act, where humans intervene, and who owns the decision.

Two tools, one principle

At Deep Current, the principle is to structure, validate and complete messy inputs automatically, so the data is system-ready before anyone has to work with it. Our current tools apply that principle to different points in the workflow.

Documus Prime checks that each shipment’s paperwork is complete, correct and compliant before it moves. It identifies the documents a shipment requires, checks details across documents, flags discrepancies and helps teams correct them.

Ada handles customer communication. It replies to client emails around the clock, collects shipment size, destination and timeline, and handles FAQs, tracking and status updates. It takes the first interaction so sales teams can focus on closing.

But the products are less important than the principle behind them. AI should not become another place where an operator has to work.

If someone has to leave their TMS, open an AI tool, copy information across, interpret an answer and then manually enter the result back into the system, we have not really removed the problem.

We have added another step.

That is why one of the biggest ideas in our digital sophistication framework is workflow embedding. AI becomes operational when it sits inside the workflow rather than alongside it.

What customers see

I shared the results our customers report:

• 70% less time spent on manual documentation
• 60% fewer customer query emails for ops teams
• 99% error-free workflows with automated validation
• 30% more shipments handled with the same headcount

Source: Deep Current customer results.

One customer summed up the change in a sentence I like to repeat: “We used to start our days putting out fires. Now, we actually plan ahead.”

For me, that is a more useful measure of AI adoption than how sophisticated the technology sounds. The question is not simply whether AI has been introduced. Has the way the team works actually changed?

Why forwarders trust the tools

Three things matter here.

The tools are built by people with decades of hands-on forwarding experience and tested by logistics professionals before release. They integrate with the systems a forwarder already runs, so nobody has to redesign their operation around us. And Deep Current is ISO/IEC 27001:2022 certified for the design, development and operation of its SaaS.

But there is another part of trust that matters just as much:

Knowing when not to automate.

As AI becomes part of operational workflows, trust cannot come from simply telling people that the system is intelligent. It comes from clear boundaries, validation, escalation paths and human oversight.

That is why governance is not something to think about after implementation. It needs to be part of the operating model from the beginning.

How to start

I am wary of grand transformation programmes, so my advice to the ALP members was simple:

Start with one workflow.

1. Pick one pain point. Choose a high-volume, repetitive task, such as document checks or inbound emails.
2. Measure the baseline. Record time spent, error rate and rework today, so results are visible.
3. Pilot on live shipments. Run it with your team in the loop and tune it to your reality.
4. Scale what works. Roll it out wider, then move on to the next workflow.

This is also consistent with how we approach implementation at Deep Current. The objective isn’t to automate everything at once. It is to identify a real operational friction point, prove the value and then expand.

What the conversation reinforced for me

I came away from the session with an even stronger conviction that the next stage of AI in logistics is not about adding more AI tools. It is about making AI operational.

That requires several things to happen together. Data has to become structured and connected, AI has to move from dashboards into workflows and decision support has to sit closer to the moment of action.

People need to know what the system can do, what it cannot do and when they need to intervene.

And organisations need to measure operational outcomes rather than simply counting pilots. These are the ideas behind the five levers of digital sophistication we identified earlier this year: integrated digital foundations, decision intelligence, workflow embedding, predictive resilience, and human-AI partnership and governance.

The ALP discussion reinforced for me that these are not abstract technology concepts. They are practical questions that forwarders are already dealing with every day.

The bigger picture

I closed the session with the broader case, because smoother logistics moves more than freight. For businesses, it can mean more predictable cargo flows, lower costs and healthier margins. For people, it can mean less firefighting and more meaningful work. For customers, it can mean consistency instead of unpleasant surprises.

And at a broader level, smoother execution can reduce wasted activity across logistics operations.

That is ultimately why we are building Deep Current.

Logistics doesn’t have a data shortage problem. It has a data-flow problem.

Every shipment begins with information. If that information is incomplete, inconsistent or disconnected, everything that follows becomes harder. Our focus is therefore on the layer before the workflow: structuring information, validating it, connecting it and making it usable. When data flows correctly, workflows become smoother.

And when workflows become smoother, people can spend more time doing the work that actually requires them.

A final thought

Events like the ALP Virtual Camp are valuable because they bring independent forwarders together to learn from each other.

Aydin and the ALP team built a programme around that idea, and I am grateful they gave Deep Current a place in it.

The questions and discussion also reinforced something I believe strongly: the people closest to the operation should have a central role in deciding where AI belongs.

Technology should adapt to the reality of freight forwarding and not the other way around.

If you attended the session and would like to continue the discussion, or if you would like to explore a first pilot workflow, please get in touch at tamim.f@deepcurrent.no or visit deepcurrent.no.