The Next Robotic Fulfillment Disruption Will Be AI-led

On its 30th birthday, e-Commerce is growing bipolar:

On the digital front end, it’s highly advanced and constantly evolving. On the other hand - the fulfillment backend - it largely uses the same 10-20 year old technologies.

The warehouse industry is employing around 5 million people in the US and Europe. To put things in perspective, one of the most successful warehouse robotics companies - Autostore - recently celebrated the shipment of robot number 50,000 (deployed in a few hundred sites). 

Adding other leading automation and robotics companies to the count, we can estimate that around 200,000-300,000 robots are in operation in the US and EU. 

That’s a 20 to 1 Person-Robot-Ratio (PRR) without accounting for Amazon. 

That’s a very low figure. It supports the estimate that around 90% of fulfillment centers are still without robotic assistance.

Amazon is the bold exception: employing around 1 million warehouse workers, it runs the world’s largest fleet of over 500,000 warehouse robots. A 2:1 PRR. 

These 3 ground truths are foundational to the conversation: 

  1. E-commerce keeps growing
  2. Labor is becoming more expensive and harder to find
  3. Robotics are becoming more available and at lower prices

Yet the robotic e-commerce revolution hasn’t happened broadly, outside of Amazon.

The Price-Productivity Paradox

As can be seen in the matrix, the supply of robotic systems is clearly divided into two types: the Classic - fixed-infrastructure, expensive, state-of-the-art systems that power highly productive operations, alongside Mobility - low cost, ground roaming systems that support only modest overall productivity gains. Little is happening in the middle, which drives many players into a “wait and see”. 

Option #1 - Too Expensive 

The origins of warehouse automation is in industrial manufacturing. These “classic” warehouse systems were designed to play in an area of the supply chain where you have perfect repetition of similar tasks (e.g. producing or storing the same products again and again), at massive scale. This is what enabled such systems to be state-of-the-art, millimetric accurate, therefore expensive and complicated to run. This group is represented by AS/RS (Automatic Storage and Retrieval Systems) systems called Shuttle or Multi-Shuttle, as well as Cube Storage systems - all rely on robotic units to move within bespoke racks and rails for maximum accuracy and speed. This is essentially the easiest environment for a robot to operate in: doing the same things over and over again and costing as much as it needs to.

These systems are not a natural fit for the type of e-commerce that has large, mixed baskets, and/or same-day deliveries.

Option #2 - Too “Basic”

When Amazon purchased Kiva in 2012 - the upstart robotics company was known for a very different kind of warehouse robot: an Automated Ground Vehicle - AGV - based on the novel idea that the robot can simply move on the floor instead of inside a racking system. 

Today there are over 150 vendors of AGV and AMR robotic systems that all move on the ground, cost much less compared to ASRS and are easier to deploy and operate. So why haven’t they created the long awaited tipping point in warehouse robotics? 

3 reasons come to mind: 

  1. Modest productivity - units which move on the ground (instead of rails) are slow, and consequently so is the productivity that was based on such robots
  2. Low density - the storage units had to be mobile as well, creating both a vertical (height) dimension limit as well as horizontal density limit, resulting in less storage space available 
  3. Limited scope - ground roaming units are designed to focus on moving goods to fulfillment and packing, but the overall end-to-end fulfillment operation is much broader and includes flows which AGVs don’t usually support such as dynamic decanting, consolidation and dispatch

So while the cost of the single mobile robot unit is undeniably low, the true cost of the desired logistical output isn’t as low due to the inefficient labor and waste of valuable storage space. 

What made Amazon’s reliance on AGVs so successful at such a scale is the huge investments they made throughout the past decade in optimizing this type of hardware with software, operations design and AI.

To sum it up, classic automation supports high productivity but is too expensive, while ground robots don’t deliver enough productivity and density gains to make the numbers work, despite being low-cost

Why we founded Finally

The continued rise of e-commerce, alongside the abundance of AGV vendors, and the increasing costs of labor are all part of a perfect storm impacting the fulfillment industry. We saw firsthand how classic AS/RS systems haven’t been adopted by a large proportion of fulfillment centers, at the same time as AGVs fail to empower the kind of productivity gains to make the cost and effort worthwhile. 

The formula was hence right in front of us: build a robust AI and software layer over AGV hardware to step change their productivity and account for the entire scope of operation. We knew that only by unlocking an order of magnitude improvement in the output of AGV robots could we make robotic fulfillment a widespread reality, finally.

Unlocking a foundational fulfillment economy 

Unlocking a foundational fulfillment economy

Our SFC (Store Fulfillment Center) products all champion the same groundbreaking values: 

  1. Lowest overall cost-to-fulfill
  2. Highest density physically possible at a given height
  3. End-to-end product, out of the box 
  4. Robust support and maintenance program

Our vision is that labor should be better utilized in logistics and all around retail, as low cost robots can finally support high performance e-commerce order fulfillment. And if the 2:1 ratio at Amazon is any indication: 2-3 million robots will be deployed in the next 10 years with the power of AI to step-change their outputs. This is the next revolution of robotics.