The Nine Layers of AI: Why Shipping Companies Must Fix the Foundation Now to Stay Competitive Later

By: Moshe Mazuz, Chief Data Scientist

Artificial Intelligence is revolutionizing industries across the board—and shipping stands to be one of the biggest beneficiaries both in terms of speed and the magnitude of change its undergoing. An expected reduction of 80-90% of administrative workload onboard and on-shore, predictive capabilities, AI offers transformative potential. But shipping companies often operate in highly fragmented IT environments, with special versions, and with dozens of special-purpose systems that don’t really talk to each other, each in its own version and format.

In this setting, adopting AI without addressing its foundational requirements is like building a skyscraper on sand. Let’s break down the Nine Layers of AI and explore why ignoring the bottom half of this pyramid—especially in the shipping industry—is a strategic risk.

The Nine Layers of AI

Think of AI implementation as a pyramid of nine interdependent layers. Here’s a quick overview, from the ground up:

1. Data Capture

2. Data Quality & Standardization

3. Data Integration

4. Data Infrastructure

5. Analytics & Visualization

6. Machine Learning Models

7. AI Applications

8. Human-AI Interaction

9. Governance, Ethics & Policy

   
   

Each layer depends on the stability and maturity of the one below it. Let’s explore why ignoring the lower layers now can set the industry up for failure or irrelevance in the coming years.

Layer 1–4: The Underinvested Backbone

1. Data Capture

Most shipping companies still rely on manual entry or siloed digital inputs into a special version ERP. Missing or inconsistent data from vessels, terminals, and offline systems makes it nearly impossible to build reliable AI models later.

2. Data Quality & Standardization

Almost all ERPs today are built in multi versions, and multiple modules so logs don’t speak the same “data language.” Even timestamps can be inconsistent. AI needs clean, harmonized inputs. Junk in—junk out. 

3. Data Integration

If you are using different systems, a limited APIs with a lot of instability is just not going to cut it when it comes to AI.

4. Data Infrastructure

Many shipping firms still lack a centralized data lake or scalable architecture to store and process large volumes of streaming data from sensors and reports. AI doesn’t work without compute and storage muscle. And its needs the full data of thousands of ships to be able to properly train models to give meaningful results.

Neglecting these foundational layers leads to AI initiatives that are brittle, narrowly scoped, or entirely non-functional. Worse, it leads to wasted investment and disillusionment. You can easily scar your team’s willingness to change and trust AI and the damages will take a while to correct.

The Illusion of Progress: Layers 5–9 Without a Base

5. Analytics & Visualization

Dashboards are useful, but they don’t equal intelligence. Pretty charts won’t compensate for unreliable inputs.

6. Machine Learning Models

A predictive maintenance or predictive procurement model trained on noisy, sparse, or unverified data will do more harm than good—prompting unnecessary interventions and creating a contrarian vibe to AI as a whole.

7. AI Applications

Whether it's a Co-Pilot that can let you talk to your data and plan your activities, or predictive procurement, apps built on weak foundations crumble under operational stress.

8. Human-AI Interaction

Your crew and ops team can’t trust a system that gives conflicting advice or pulls data from unknown sources. Trust and usability depend on solid groundwork.

9. Governance, Ethics & Policy

From IMO regulations to the new EU AI compliance requirements, AI must comply with evolving norms—but that’s impossible if you can’t trace your data lineage or verify predictions. We are anticipating massive issues with AI that will be deployed on legacy environments, and the risks are significant given the sensitivity of data in shipping.

The Shipping Challenge: Fragmentation and Legacy Systems

The industry’s reliance on bespoke systems—some decades old—means data is scattered, duplicated, or inaccessible. APIs may be absent. Documentation is patchy. Each integration is a project of its own.

This technical debt limits your AI future before it even begins.

Why Act Now: The Cost of Waiting

Companies that wait to clean up the lower layers will hit a wall as the AI race accelerates:

• Competitors will leapfrog with scalable, integrated AI platforms.

• Customers and regulators will demand transparency you can’t provide.

• Your talent pool will shrink, as employees avoid legacy-heavy environments.

Investing in data architecture, integration layers, and standardization today sets you up to adopt future AI solutions quickly—and responsibly.

Next Steps for the Shipping Industry

1. Audit Your Provider(s)

   
   

   How many data schemes and versions exist? What data do they generate?

   
   

   Where are the gaps?

   
   

2. Prioritize Data Infrastructure Projects

   
   

   Start moving all the data of your systems into a single data layer.

   
   

3. Invest in Data Infrastructure

   
   

   Whether cloud, hybrid, or edge-computing—get your foundation ready.

   
   

4. Adopt Open Standards

   
   

   Push vendors and internal systems toward standards! Single version, single schema, fully integrated systems.

   
   

5. Create an AI Roadmap

   
   

   Align your technology stack to your business goals—think 5 years ahead, not 5 months.

   
   

Conclusion: Don’t Skip the Groundwork

The shipping industry has everything to gain from AI—but only if we’re willing to address the invisible, less glamorous parts of the stack. Ignoring the lower layers today means you’ll be locked out of innovation tomorrow.

Build your AI future from the bottom up. That’s how ships—and smart systems—stay afloat.