AI-driven warehouse analytics: Transforming inventory data into actionable insights

Introduction

Since the pandemic, e-commerce sales have continued to surge, with the latest reports forecasting sales of $8.09 trillion by 2028, compared to $5.13 trillion in 2022 [1]. Alongside this, demand profiles are changing as consumers want their products faster and labour shortages could lead to 1.9 million unfilled manufacturing jobs over the next 10 years [2]. There is no question that the logistics landscape is transforming dramatically, and to survive, warehouses must respond with new strategies, technology and thinking.

This is why companies such as Dexory are fuelling the transition from conventional, ‘blind’ warehousing to automated, self-learning, ‘adaptive’ warehouses.

‍

The evolution of the warehouse

‘Blind’ warehouses have no real time visibility of stock and instead, rely on manual processes to track inventory, identify misplaced items and conduct stock takes. This is not only time consuming, but can also lead to errors, inefficiencies and an overall reactive approach to problem solving.

However, new scanning technologies such as barcodes, RFID tags, WMS (Warehouse Management Systems) and autonomous robots can convert labour-intensive data gathering into an automated process, improving inventory accuracy, operational efficiency and the number of errors. This ‘observable’ warehouse automates data collection but still relies on manual processing to generate insights.

Whereas the ‘intelligent’ warehouse utilises AI and machine learning techniques to extract valuable insights from collected data, which humans then use to make decisions. This results in a more proactive and agile workflow, accelerating decision-making, with AI’s ability to forecast demand and stock levels unlocking new efficiencies.

The final stage in this warehouse evolution is the ‘adaptive’ warehouse. This employs AI agents to make decisions without human intervention, resulting in an autonomously self-learning, self-optimising and self-executing warehouse capable of making more informed decisions faster, maximising efficiency and resilience.

‍

‍

The tactics to facilitating this warehouse journey

So, what are the key drivers that will enable this transition from blind to adaptive warehousing? Firstly, the observable warehouse needs to achieve extremely accurate data capture. 

‘You have to build trust in the data gathering process,’ says Chris Coote, Director of Product at Dexory. ‘There are three dimensions to this; the accuracy of data, the timeliness of data and the completeness of data. These are the fundamental principles behind every development of our autonomous inventory scanning robots and how we can operate with 99.9% accuracy.’

Dexory’s next-gen sensors for ultimate accuracy [Insert link to ‘New sensor capabilities’ article]

Once this real time data has been collected, the focus for intelligent warehouses shifts to not only extracting reliable insights but turning these into useful actions. This requires developing intuitive, representative and validated AI models, which customers can trust.

‘Context and trust are two essential pillars for the transition to the intelligent warehouse,’ highlights Coote. ‘Instead of simple recommendations such as move product from position A to position B, the AI is surfacing more strategic actions which would previously be human-led. These actions are higher risk, and therefore costly, because they require more work which could have operational knock-on effects. So we have to build trust around how our AI models interpret data and extract insights.’

‘There are two ways of doing this; providing the workings and context behind AI’s decision making, and quantifying the impact,’ adds Coote. ‘We need to do both. If our algorithms suggest an action that will gain 5% efficiency for example, we need to show customers how the AI calculated this and then prove this 5% efficiency gain. If we build trust in the process that got to the decision as well as the impact of that decision, it will be easier for customers to move towards adaptive warehouses.’

‍

Surfacing actionable insights of warehouses

An example of just how valuable AI can be at extracting intelligent insights is Dexory’s ability to identify pallets of different colours. One client, who rents pallets that are a different colour to their own, requested DexoryView to identify and quantify the number of pallets of each colour. This allowed them to track the number of rented pallets and therefore the charges they needed to pay their supplier. 

‘The advantage we have is our robots are continuously scanning racks every day, collecting 36TB of warehouse data,’ explains Richard Williams, Senior Robotics Engineer at Dexory. ‘So, we can easily train and validate AI models. In fact, whilst developing this colour pallet capability, we also created algorithms that analysed damage on pallets, misalignment of stock and even individual items. We might not have a use case for these capabilities yet, but we are continuously pushing to extract the most warehouse management analytics from the data we capture.’

‘Often the challenge is not detecting something, it’s how we turn that into a meaningful insight for our customers,’ adds Williams. ‘Sometimes it can be easy, such as counting the number of pallets, and other times it is more complex and requires additional analysis to surface insights to help customers improve their inventory management.'

‍

‍

Hidden warehouse analytics

Leveraging AI to extract more intuitive observations faster is not the only advantage of intelligent warehouses. AI also has the power to reveal insights hidden within the data that humans may not even know were there. For example, the information collected by Dexory’s robots can be used as metadata to improve overall warehouse management such as sustainability.

‘There’s a growing trend in logistics to reduce the carbon footprint of warehouse operations and this will only continue as legislation comes into force in the future,’ explains Coote. ‘Improving overall operational efficiency already boosts sustainability because it reduces waste, whether that be people’s time, the number of trucks in the loading bay or available space for stock, ultimately minimising overheads.’

‘AI could also assess how HVAC, lighting and power are used in the warehouse to optimise energy usage and schedule robots to operate in the dark,’ continues Coote. ‘These kinds of intelligent efficiencies are just one example of how AI can turn data into action. Beyond energy management, predictive analytics and realistic forecasts derived from inventory data can help customers shift from being reactive to proactive, ultimately optimising their entire warehouse operation’. ‘Exploiting AI to turn data into action is the key to unlocking the intelligent warehouse. By uncovering predictive analytics and realistic forecasts from inventory data, customers can switch from being reactive to proactive, helping them to optimise their warehouse operations.’

‍

References

[1] Y.L., 2024. Global Ecommerce Sales Growth Report [Online]. Shopify

[2] M.S., L.B., J.M., D.S., 2025. 2025 Deloitte Manufacturing Trends: From a Workday Lens [Online]. Deloitte