FourKites is a company that is evolving rapidly to meet its customers’ needs. Founded in 2014, it pioneered supply chain visibility solutions. Over the last ten years, it has accumulated a huge network of customers, carriers, and 3PLs to offer visibility into supply chain data worldwide.
With the supply chain disruption of the last few years and the advent of usable AI tools, supply chains are evolving from visibility to orchestration. Enterprise Times spoke to Himanshu Mehrotra, SVP Product Management at FourKites, about the company and its ambitions.
Over the past ten years, FourKites has moved from visibility of transportation to execution and yard management. It has developed digital twins which enable customers to operationalise external data visibility of external risks and see how they impact their supply chain.
The next step is leveraging AI. Mehrotra explained, “With the advent of AI in the last couple of years, we have been able to take it to the next level by not just giving them visibility and impact analysis, but the AI agents help them to execute and mitigate against those disruptions.”
External data is a huge challenge still.
Customer challenges are evolving. However, data was and still is the biggest challenge, or the lack of it. Not the internal data but access to external data. Organisations have data from planning, transportation and warehousing. But there are still gaps between these often held in data silos. Mehrotra believes that, like real gaps, things fall between them and are lost. Simple things are not getting done.
The challenge is that most organisations’ supply chains are made up of suppliers, customers, carriers, and 3PLs. Despite needing to collaborate, most organisations do not have access to data held within them. If an organisation orders 4,000 items, it wants to know how many and when they will appear in the warehouse. When dealing with 3PLs that run the warehouse, firms want to know what is happening in the warehouse.
Mehrotra explained. “I want you to tell me beforehand what’s happening within my warehouse. What are the constraints? Am I running everything on time? If I’m not running on time? Where exactly are those hurdles so that I can make better decisions within my own supply chain to do that?”
I asked Mehrotra whether this included access to weather data and other data outside their supply chain. He replied, “Many of our existing customers are more concerned about things like Red Sea avoidance or the strikes at ports. Those are more impactful than just weather data for them.” He explained that they can build in resilience for poor weather but unexpected events highlight the lack of resilience in supply chains.
Languages are a challenge in Europe.
In Europe, however, collaboration between different parties has additional challenges. One is the language barrier. European transportation works on brokerage, which means that if something is contracted, it might be a third party that does the delivery. The driver could be Romanian and doesn’t speak any other language.
The other factor is that 3PLs make decisions by themselves. Across Europe, they may use a ferry, train, or road, a decision they often make without referring to the end customer. The problem is that the end customer has no visibility of where their goods are or what delays might be incurred.
Can AI help solve some of these challenges? Mehrotra certainly believes so. He said, “All of that communication can happen through an AI agent in their natural language.” It means that while an organisation might have a customer asking in German, the AI can deal with the query and liaise with the Romanian driver in their language.
It is not just about language. An AI can interact with people in the channel of their choice, be that email, SMS, WhatsApp or even voice.
The Emergence of Digital Workers
In FourKites’ terms, a digital worker is a vertically trained agent. They conduct specific tasks for which they have the industry knowledge and data to perform tasks such as track and trace. FourKites, through the extensive data it holds within its network, has access to far more data than any human can remember. This is where the advantage of digital workers exists. However, they are limited by their process knowledge.
Once the agent is trained on the data and has the knowledge, it can converse intelligently. Mehrotra explains that there are different levels of an agent: “The AI agent can be a very simple AI agent from your customer services agent answering your questions, or it could be executing workflows.”
He gave an example of why Digital agents improve efficiency. If a customer wanted to delay delivery by three days, it would take between five to six hours to complete the task because of the volume of enquiries and the need to read through them all. With an AI agent, the process can be automated and take minutes rather than hours. This simple automation of repetitive tasks is straightforward.
However, Agents are getting more intelligent. They can now provide recommendations to the customer, on which they must make the final decision. The human-in-the-loop is important. People should be wary of giving complete autonomy to AI.
How to Deploy Digital Agents
I asked Mehrotra if there were some simple steps that a customer must undertake before deploying digital agents.
- The customer must understand the problem they are aiming to solve.
Why? Digital agents have not reached the level of general intelligence. FourKites has already launched three digital agents. The first is Tracy, the track and trace digital agent; Sam, a supplier collaboration agent; and Alan. The third is an appointment manager agent who books appointments on your behalf for your customers at your customer sites.
- Once the problem is known, the workflow must be understood for that use case. Mehrotra noted, “There are lots of exception workflows that people think happen, and everyone thinks we are very unique. The reality is they are not.”
- There must be data to support the use case; without sufficient data, AI will give poor results.
- The agent should then be configured for that use case.
- The last step is to ensure that change management is in place to align everyone with the new processes so that the organisation can execute them.
Intelligent digital agents will make a huge difference to organisations. Mehrotra gave an example of a food industry organisation that requires suppliers to load trailers and organises carriers to pick them up.
Since each load has a short shelf life, it can be difficult to manage. An AI agent can contact each point in the supply chain to determine if the goods are loaded on the trailer and when they need to leave. It can identify when the trailer is shipped and, if it has not been picked up, proactively contact the carrier to ensure that it is delivered in time to the destination.
The digital agent takes over the communication, Mehrotra explained, “It’s very minimal impact to anyone. But there is an element of change for the carrier to receive this information directly, rather than a phone call, receiving it either via email or directly into their systems (from the Digital Agent).”
Building trust with AI agents
FourKites takes a sensible approach to building trust. Mehrotra explained that in the first instance, AI helped people to automate processes that people spend hours and hours doing, such as extracting information from documents. This is the first step its agent can take, which starts to build trust in the AI Agent.
The second step is to have the agent perform the simple tasks that they would normally do with this information. Often, these tasks are mundane, manual, and time-consuming. The third level is the insights layer. Mehrotra explained that this is where the agent will alert the user to where some goods, if not shipped, may spoil. It prompts the user to take action, such as arranging a carrier.
The agent does not directly take control, but it informs the person who makes the decision and takes action. The final stage is where trust has been built enough for the agent to be fully autonomous and allowed to do certain tasks, but still have human oversight, especially where exceptions occur.
A case study
FourKites’ digital workers are already being used. Mehrotra shared an example of how they are making a difference. A global electronics component manufacturer and distributor had a challenge where they had only purchase order visibility through integrated systems of 35 out of 1,300 suppliers.
In addition, suppliers were not following the corrected processes for using specific 3PLs or forwarders when shipping goods. It means that the customer had over US$8 – 9 billion in inventory sitting in their warehouses because suppliers were shipping goods with no visibility. Also, transportation costs were high because the supplier often engaged in express deliveries, as they felt it was the right approach.
The challenge was that it was impossible to integrate with most supplier systems. With digital workers in place, they are now getting visibility of what suppliers are shipping and when they are shipping. Digital workers are communicating directly with each supplier, something the organisation did not have the traditional workforce to cover.
They can now ensure that the supplier books the shipment with the right supplier. As they now know when goods may arrive, they can also forecast the tariff cost impact they might see.
The Digital Agents can communicate in multiple languages across multiple channels. They can request the commercial invoice, packing list and other details that the digital agent consumes and then propagates to their internal systems. They now have visibility of over 800 suppliers.
Mehrotra concluded, “All of that happened within a period of four to five months. They now get full visibility of that, including freight cost and duties paid and the quantity on time and in full.”
“Already, they have reduced freight costs by 12% and expect to see a further 8% reduction in the coming months. They will also reduce the overstocking issue within the warehouses.”
What is next?
Mehrotra was keen to explain that FourKites are working on several initiatives, some of which are not yet public. They are looking at detention and demurrage, natural language analytics and tariff features. In addition, they will add more agents. Tariffs, with the recent trade war, are a key concern for customers who are looking for ways to reduce their impact.
I asked Mehrotra what steps supply chain leaders should take to improve the effectiveness of their supply chain.
He answered, “The first thing is to bring visibility. If you can’t see what’s happening, you cannot solve it. The second one is to find out the gaps. What are the reasons your KPIs are not being met? Then, fill those gaps, either through processes, solutions or agents. I’m not saying every solution is an AI agent or every solution is a software system. It could be process, it could be anything, but to me, that’s the way to do it.”
FourKites has always been known for visibility. By solving integration challenges, it helps to break down the information silos that exist within and outside organisations.
They are now leading the way with the Intelligent Control Tower. Mehrotra added, “Most control towers were just giving predictive visibility. While they may have made recommendations, none of the control towers were able to execute and orchestrate. That was the biggest missing link. FourKites is now able to operationalise it.“
The Book Question
What was the latest book you read?
“I reread Zone to Win (by Geoffrey A Moore, Amazon Aus, UK, US). It’s a very classical book that my previous CEO at my previous company had us read. I loved it so much. We are in that particular kind of evolution in FourKites. We need to use a very similar concept in the way we need to build our business. I’m making sure that my team and I follow those principles.
“Zone to Win is all about companies like ourselves, who have got a very mature solution that also needs to be developed and launch new products. How do we ensure that we zone those in various ways and zone capacities to make sure that they don’t steal from each other? Make sure that a particular product is able to move from an incubation zone to a growth zone.
“So essentially, these are the two extremes of those four quadrants: incubation and the growth zone, or a revenue zone.“
