DEEPER DIVES: The Truth About AI In Logistics & Supply Chain

Good Morning,

First, as always, thank you for joining.

Last week was warehouse week. I connected with a few new customers and attended a cool virtual networking session co-hosted by Kevin Lawton and The New Warehouse. If you’ve never used the Remo platform, you should check it out if you are looking to host your own event.

Here’s what this issue brings:

• AI is everywhere. Pretty much every service provider that you’ll talk to will tell you that their product is “AI Driven”.
  
  I think I break down what that (likely) means for the majority of providers out there. We’ll also jump into the different types of “AI” out there and where they are being applied
  

• Loop Returns announced Offset, a program designed to “help” merchants offer free returns… by passing the costs onto the customers
  

• Google’s Looker Studio is one of the best data visualization tools you can use and it’s free

OpenAI And Anthropic Are Losing Billions On AI. Do You Really Think That Start-Up SaaS Uses Their Own Model?

I’ve been doing a ton of data work lately.

Data analysis, dashboards and figuring things out.

One of the tools I use daily is KNIME. It’s an analytics platform for processing large amounts of data and creating automated workflows.

While extremely powerful as an ETL (extract, load, transform) tool, sometimes some “simple” things you can do in a spreadsheet are overly complicated in KNIME.

Imagine my surprised when I saw a post the other day showing some features of K-AI, KNIME’s experimental AI chatbot.

Thinking this was super cool, I explored more about it.

While integrated directly into the platform, it seems that it leverages OpenAI’s model. Whether it has additional layers of training or fine tuning is not clear.

But this got me thinking.

How many other “AI solutions” being sold are simply integration layers that are all leveraging the same back-end models?

When most people think about AI, they envision robots that work independently, systems that "think" like humans, or technology that invents new solutions in real time.

The truth is far less glamorous.

From the research, it seems like the AI that powers the logistics industry today doesn’t fit those expectations.

Misconception: AI in Supply Chain Isn’t What You Think

Most AI solutions in Supply Chain and Logistics rely on data processing and predictive models.

These tools use huge amounts of information to find patterns and make forecasts.

But they don’t create new ways of doing business.

Instead, they improve efficiency by automating repetitive tasks and providing data-driven insights.

While this can lead to significant gains in operations, it isn’t the human-like intelligence that many expect when they hear the term "AI."

And in the majority of cases, the quality of the answer is limited by the statistical data fed into it.

OpenAI & Anthropic’s Financial Reality. AI Is Mad Expensive

The two biggest names in AI right now are OpenAI and Anthropic.

OpenAI has partnered with Microsoft and Anthropic with Amazon.

OpenAI's GPT and Anthropic’s Claude models have received widespread attention, but operating these models is expensive.

It was reported this week that Open AI is tracking towards $3.7 Billion in revenue this year.

But they are also anticipating a $5 Billion dollar operating cost loss.

Anthropic’s story isn’t much different, just a smaller scale.

They are projecting revenue of $800 Million for this year. And they are only losing half as much money … about $2.7 Billion.

Because of this, both of these companies are looking for more money and are looking to close new funding rounds that would increase the valuation of both companies.

OpenAI is also exploring restructuring the company. They are considering transitioning to a for-profit model (first) while maintaining a separate non-profit division.

My point is this, their day-to-day costs reach hundreds of thousands or millions of dollars.

Despite their innovative technology and market dominating positions, they are facing huge financial challenges.

They burn through significant cash just to keep the lights on.

Maintaining AI at this level demands huge investments in infrastructure, computing power, and specialized talent.

This should shed light on why true AI development is rarer than what everyone wants you to believe.

In the context of supply chain technology, many smaller companies simply can’t compete when it comes to building and maintaining advanced AI systems in-house.

Providers Are Relying On External AI Services

Smaller providers are not developing original AI in most cases.

Instead, they lean heavily on third-party services like Google Cloud, Amazon, Microsoft Azure, Anthropic and OpenAI.

These platforms allow companies to integrate AI capabilities without needing to develop them from scratch. It’s a practical solution for businesses with limited resources.

Developing proprietary AI requires extensive expertise and resources (which are often beyond the reach of mid-sized logistics companies).

So, what they do is use existing AI models, applying them to their own datasets to automate processes like inventory tracking or route optimization.

The Different Types of AI Technologies

With that base understanding, let’s get some context to what AI in Supply Chain and Logistics really means.

Especially since there isn’t a day that goes by where you aren’t reading about it.

Understanding the types of AI used in logistics helps clarify their actual role.

• Machine Learning (ML): Common in predictive models that learn from data.

• Deep Learning: Used for more complex tasks like image or voice recognition.

• Natural Language Processing (NLP): Drives customer service bots that respond to text or speech.

• Computer Vision: Helps machines see and sort packages in automated warehouses.

• Reinforcement Learning: Applied in tasks like route optimization.

• Predictive Analytics: Enables businesses to forecast demand and prepare accordingly.

While each of these technologies plays a vital role in logistics, they are often marketed under the broad term "AI," even though their scope and capabilities vary significantly.

So How AI is Used in Supply Chain and Logistics

You’ve heard the term tossed around, but have you considered how AI is already influencing your daily operations or how it could change the way you work?

AI is becoming an integral part of supply chain and logistics, often in subtle ways that may not be immediately apparent (and not in ways some people want you to think).

From demand forecasting to automating warehouse tasks, AI technologies are quietly revolutionizing processes across the industry.

Demand Forecasting

• AI Type: Machine Learning

• Use: Analyzes past sales trends to predict future demand

• Challenges: Requires high-quality data, which can be a limiting factor

• Benefits: Reduces overstocking and stock-outs, improving cash flow and customer satisfaction

Route Optimization

• AI Type: Reinforcement Learning (make decisions by taking actions in an environment and receiving feedback in the form of rewards or punishments)

• Use: Identifies the most efficient delivery routes in real time

• Challenges: Must integrate with external data sources, such as GPS and traffic data, to be effective

• Benefits: Reduces delivery times and fuel costs, driving operational efficiency

Warehouse Automation

• AI Type: Computer Vision, Robotics

• Use: Automates tasks like sorting and packing

• Challenges: High initial costs and complex integration with existing systems

• Benefits: Increases accuracy and reduces labor costs

Predictive Maintenance

• AI Type: Predictive Analytics

• Use: Anticipates equipment failure based on usage patterns

• Challenges: Requires continuous data collection from IoT sensors

• Benefits: Reduces downtime and maintenance costs

Supplier Risk Management

• AI Type: Machine Learning

• Use: Analyzes data on supplier performance and market risks

• Challenges: Limited by the transparency of the data supplied by third parties

• Benefits: Minimizes supply chain disruptions by proactively identifying risks

Customer Service Automation

• AI Type: Natural Language Processing

• Use: Automates customer queries, particularly about order status

• Challenges: NLP systems still struggle with complex queries

• Benefits: Reduces the workload on customer service teams, cutting costs

Business Risks of Relying on Small AI Providers

One of the biggest risks businesses face when investing in AI-based logistics solutions is the instability of smaller providers.

Many of these companies rely on external AI services, and if the cost of using third-party platforms rises or the service changes what they provide access to, the providers might find it difficult to adapt.

Worse, some small software providers may not have the customer base needed to stay viable.

This creates a risk for companies who invest in a solution that could disappear if the provider can no longer support their operations.

There’s also the problem of vendor lock-in.

Companies that build their processes around one AI provider might find it difficult to switch if the provider changes terms or shuts down. This dependency on a single AI platform can be costly and risky in the long run.

What It Takes to Leverage AI Effectively

For businesses to benefit from AI in their operations, they need to address several key areas.

1. Data Quality
   
   AI systems rely heavily on clean, structured data. Without it, AI outputs are unreliable.
   

2. Training 
   
   Employees need to understand how to manage and interpret AI-driven insights.
   
   It’s not just about using the technology, it’s about applying it effectively.
   

3. Strategic Vendor Selection
   
   Choosing the right AI vendor is crucial. Look for vendors that are transparent about how their AI systems work, what kind of data they require, and the results they can realistically deliver.
   

4. Change Management
   
   Integrating AI into operations requires a shift in how teams work, and this can be met with push back.
   
   You need to invest in training and change management to help teams adjust.

AI has enormous potential to improve your supply chain and logistics operations.

But it’s important to approach this technology with realistic expectations.

Most of the AI solutions available today are tools for optimizing existing processes, and not reinventing them.

For businesses to make the most of AI, they need to focus on data quality, vendor transparency, and ensuring they have the skills in place to interpret and act on AI insights.

By approaching AI with caution and doing your due diligence, you can leverage this technology to drive meaningful improvements without falling for over-hyped promises made to close sales.

Retailer’s Are Looking To Solve Returns With An Insurance

Loop's Offset is a new pre-purchase returns solution designed to help merchants cover the costs of returns while offering customers a premium returns experience.

Customers are given the option to pay a small fee upfront during checkout in exchange for “free” returns later.

In one example shown on Loop's website, the Offset fee is shown as $1.98 for a product priced at $109. This suggests that the fee might be around 1.8% of the purchase price.

There are two options for retailers looking to use the program.

Offset Free

Loop covers all costs related to returns, including software and shipping.

Loop collects the Offset fees directly from customers who choose to opt-in.

Merchants don't pay for return costs but also don't receive the Offset revenue.

Offset Flex

Merchants collect the Offset fees from customers who opt-in.

Merchants use this revenue to cover their return software and shipping costs.

Merchants have more control over the experience, allowing them to customize it for different customer segments or regions.

When I was younger, I used to work for Future Shop / Best Buy.

For those of you that know the business of those stores, you know just how much money was being made on extended warranties.

Like every type of insurance, it’s profitable for one reason.

A certain percentage of customers that buy it never use it.

Selling these services makes sense because you end up collecting more revenue than you ever pay back out.

This is EXACTLY what Offset is looking to do with retail returns. Even though this isn’t said so clearly, this is precisely what this means “Merchants don't pay for return costs but also don't receive the Offset revenue.”.

To generate revenue from Offset, it means that there will be more money collected than will be paid back out.

What’s worse, is that retailers are positioned to make this a default choice that customers feel they almost have to take.

Many retailers have started charging for returns. Overall I support this move to limit the abuse that is happening in the system.

The problem that I have now with a program like Offset, is that retailers can create a big variance between the Offset fee and the cost of a return - essentially “forcing” customers to opt-in on every purchase.

Retailers are further incentivized to do this by keeping the revenue from customers from this program because they know that 60% of the customers aren’t returning items in any business negative level.

I also see these type of program as a step backwards when it comes to sustainability and environmental impact.

By removing the pain of the returns, and making the activity a potential profit center, you only further encourage retailers to keep promoting rapid purchases and fast fashion.

This is disappointing news.

Revisiting Google’s Looker Studio

I got into pretty heavy dashboard use about 5 years ago.

It completely changed the way I thought about sharing data with different teams.

Numbers are great.

But pictures to say a 1,000 words.

The biggest challenge most people have when it comes to data visualization tools is finding something that makes sense when you are getting started.

There are some great enterprise level options, but most small businesses are going from spreadsheets and PDF reports to hiring data scientists.

Power BI is one of the most well rounded options for people to get started with. But it has one major drawback that has ALWAYS frustrated me.

It’s way harder than it should be to share information (at least, in a straightforward and cost effective way).

And the whole point of visual dashboards is for them to be shared and communicate the same information to groups of people.

I’m working with 4 different clients right now on different data analytics projects to improve their business profitability and operational execution.

With two I’m working in Power BI.

But for the other two, I revisited Google’s Looker Studio.

Why?

Because Looker Studio immediately publishes to the web. It’s extremely easy to share.

And, for organizations just dipping their toes into visualizing data, the actual dashboard software is free (Google hopes that you will see the benefit of the service and start using their Cloud services to transact and store your data - this is certainly NOT free).

The drawback for Looker versus other options on the market however is on the ETL (extract, load, transform) side of data management.

There’s basically nothing in Looker.

But if you already have your data in the table formats you want and have the dimensions you need to segment your data, it’s been a dream.

I was able to do an entire analysis for one 3PL customer on a recent shipping and rate issue. In a day we were able to go from raw data to a full dashboard that allowed the problem to be discovered and understood.

If you are looking to get into a new level of understanding your business data, check out Looker Studio today.

That’s it for this week. Thanks for being here.