AI and cybersecurity lessons for warehouse automation

New capabilities are spreading across the warehouse and logistics sector, raising the competitive standard. Traditional automation technologies offered speed, Hulbert said. “Today, software is starting to make the decisions. There's a lot of talk about the risks of that, and those risks are very different. What we said about automation before is still true, but the stakes have gone up.”

Many warehouses are already adopting new capabilities, bringing risks that leaders must learn to identify and manage.

Automated decisions

AI-driven decision-making might sound like a distant goal for some warehouse operators, but it is increasingly being integrated and released through the WMS platforms that facilities already own.

When warehouses implement AI decision-making through a WMS or elsewhere, they need to ensure their processes maintain accountability. “You need to build the solution with guardrails and explainability, because there is a financial impact,” said Grant Thornton Operations Manager Colin Smith. “If we’re losing inventory, or not hitting certain metrics, there needs to be a human in the loop to track the explainability for that. You need decision logs and tracking that someone can reference for any AI solution, especially for auditability at public companies.”

“When you have AI making decisions, your board and executive team really need to be asking who's approving those decisions, how you’re auditing them and who’s accountable if something goes wrong,” Hulbert said. “Companies aren't asking whether to automate anymore, but they're asking who is accountable when systems underperform.”

Automated decisions can be even more powerful, and risky, when they span system integrations.

Automated decisions

AI-driven decision-making might sound like a distant goal for some warehouse operators, but it is increasingly being integrated and released through the WMS platforms that facilities already own.

When warehouses implement AI decision-making through a WMS or elsewhere, they need to ensure their processes maintain accountability. “You need to build the solution with guardrails and explainability, because there is a financial impact,” said Grant Thornton Operations Manager Colin Smith. “If we’re losing inventory, or not hitting certain metrics, there needs to be a human in the loop to track the explainability for that. You need decision logs and tracking that someone can reference for any AI solution, especially for auditability at public companies.”

“When you have AI making decisions, your board and executive team really need to be asking who's approving those decisions, how you’re auditing them and who’s accountable if something goes wrong,” Hulbert said. “Companies aren't asking whether to automate anymore, but they're asking who is accountable when systems underperform.”

Automated decisions can be even more powerful, and risky, when they span system integrations.

System integration

“We’re seeing a lot of connected solutions,” Smith said. “Clients are building upon the WMS with different robotic solutions and reporting tools, daisy-chaining things together and letting upstream planning and outbound demand drive some decisions in the warehouse with real-time control.”

Beyond outbound picking, warehouses use their integrated data to help reprioritize and redistribute their inventory. This integrated decision-making automation gives them near-term agility that goes beyond accelerating work.

“Supply chain platforms are building in AI agents that can flag at-risk orders, suggest fixes for shortages and analyze cycle counts,” Smith said. “That’s pushing capabilities into the hands of the managers on the floor, without the need for long development cycles from IT.”

Even as warehouses develop processes that apply the new AI capabilities in their WMS and enterprise resource planning (ERP) platforms, they also need to ensure those AI capabilities have access to the right data.

Automated decision-making across systems can create solid-looking recommendations that conceal missed considerations or wrong assumptions caused by data disconnects.

“Some organizations still don't have their data consolidated,” Smith said. “They need to go through a project to build out the middleware and aggregate it into solutions.” Hulbert agreed, “You need to get the basics right before adding AI. Without a strong operational foundation, advanced tech just becomes an ambitious experiment.”

Most warehouse operators cannot afford ambitious experiments, especially if they put security at risk.

Cybersecurity

How we can help you

SERVICES

Artificial Intelligence -->

Ready to talk? We’re ready to listen.

Request a meeting -->

To mitigate cybersecurity risks, warehouse leaders need to have an up-to-date plan for risk management across the enterprise.

Reevaluate risks

To make sure your plan is up to date, reevaluate the risks you have today. “Identify where the risks are, and how to protect against them. Then, have a response plan for when an attack occurs,” Norris said. The response plan can keep a small breach from growing out of control. “A lot of times, nobody knows what to do when an attack occurs. How do you shut it down? What's the right response? How do you lock systems? How do you keep access to the things you need to continue?” It’s important to answer these questions in advance, and to ensure those answers are understood. 

Ensure understanding

Clear communication is critical for cybersecurity plans, and for automation technology overall. Successful technology implementations must involve the users, giving them training and gathering their feedback along the way.

“You have to provide the right training on the front end, so that you can mitigate risks and actually gain the efficiencies you're pursuing,” Norris said.

“Even in some hyper-automated warehouses and distribution centers, there are still gaps in training and understanding,” Smith noted. “That can especially happen when there are strained labor markets and changes in expertise.” However, warehouses cannot always overcome these gaps with training alone. They also need to consider how technology is designed.

Design for success

Start smart

“Buy what works today,” Hulbert said. “Humanoid robots get all of the headlines, but purpose-built automation is what performs reliably right now.”

That pragmatic approach is the best starting point for warehouse operations that want to ensure a technology initiative achieves its planned ROI. “Some companies see a big potential ROI number and jump into heavy automation,” Hulbert said. “Then, they find out that integration is hard, systems are messy and data is complicated. Process design and training are often the hardest parts of implementing technology across your organization.”

“Understand the flow of data and the actual processes within facilities, then train the organization.” Hulbert said that these are still some of the top lessons about how to achieve successful automation. Another lesson is to define success up front.

“Have clear metrics,” Hulbert advised. “You can start small, with one area, and really understand how you're going to measure success, validate it and report on it throughout the life cycle.” Whether metrics are units per hour, cost per line, service-level agreement (SLA) adherence or other measures, leaders need to understand their goals for scaling AI or other automations.

Once companies align their systems and prove success with automation, they can look at how to scale successful automations on the floor and in the back office. 

Invest with precision

As companies scale automation, they might not always expand more of the same solutions. A successful solution on the floor might have diminishing returns in other locations, but it might yield data that empowers new capabilities in the back office.

“At a macro level, how are you looking beyond just the warehousing component?” Smith asked. “How are you making sure that the business is serving the warehouse and vice versa? Look at executive-level reporting beyond historical metrics. New data can inform your planning, forecasting and tracking of real-time results.”

Back-office solutions should still be designed with metrics that clearly measure value. “Make sure that any investment dollar ties back to show that it's validated, because some companies have just thrown money at these solutions,” Smith said. “You need to think in terms of investment, making sure that you see the returns come through in executive-level reporting that's built on top of the data your teams are consolidating now.”

Now is the time to plan precise investments and metrics for success, as warehouse investments start to recover from an industry downturn.

“Companies have been underinvesting the last couple of years, but a lot of them knew they needed to improve their automation,” Norris said. “Now, we’re seeing capital investment come in, companies are playing catch-up and things will get even more competitive.”