Using Technology to Drive Inventory, Productivity, and Service Gains

Ten ways artificial intelligence can improve inventory, labor planning, and overall warehouse performance.

By Ethan Gibble
Contributing Writer

A 2024 McKinsey & Company report found that integrating AI into distribution can yield reductions of 20% to 30% in inventory, 5% to 20% in logistics costs, and 5% to 15% in procurement spend. In looking at how warehousers are successfully using AI, it’s clear those results are not driven by the AI solutions alone. Companies see the best results when they create a strategy, cultivate employee buy-in, and target real operational improvements. This article explores 10 strategies distributors can follow to leverage AI solutions across their warehouse operations. 

1. Starting with Clean Data
Companies can invest heavily in advanced algorithms, but if the underlying data is inconsistent, duplicated, or incomplete, the output is going to be unreliable — and costly.

“There’s so much data a distributor can use, but it’s got to be cleaned up,” said Will Quinn, a warehousing and supply chain strategist who has managed distribution networks for global companies such as Grainger and Coca-Cola. Using consistent formatting for data attributes such as units of measurement is critical to ensuring that an AI solution will be able to properly use that information. If an underbar system uses half-inch pipe fittings, is that spelled out as “one half inch” or 1/2” in the system? Depending on the measurement, Quinn said the same item could be entered in three or four different ways, each of which could be interpreted as a distinct item that needs its own spot in the warehouse.

FEDA and its members are ahead of this issue. One of the key objectives of the FEDA Data Portal has been to provide uniform and standardized product information that distributors and manufacturers can rely on to support their sales and supply chain efforts. While those standards have been in place from the start, the FEDA Future of Distribution Council Product Data Standards & Integration Subcommittee is working through refining the portal’s data fields to ensure they continue to provide complete, uniform data.

Alberto Oca, a partner with McKinsey & Company who specializes in warehousing, logistics, and distribution, noted that although AI is becoming more adept at working with structured and unstructured data, companies must still pay attention to how the information it feeds off is managed post-integration. This helps business leaders proactively catch AI errors or anomalies. “Areas to focus on from a data perspective are ensuring inputs like lead-time variability, service-level targets, and demand segmentation are regularly maintained,” Oca said. 

2. Establishing Operational Alignment
Having quality data is a great start, but organizations also need a plan for what to do with all that information. Oca stressed the importance of setting a value-driven goal up front then exploring how AI can help make it happen. “Define the north star first and the impact you want to achieve instead of focusing on a technology roadmap first,” he said. “It’s important to set the ambition, and then leverage use cases to bring rapid impact to keep momentum and help with self-funding the AI journey.”

With the goal set, businesses should prioritize organization-wide buy-in. Quinn acknowledged that many in the workforce are on edge about AI. Clear, upfront communication can help ease those concerns. “They think companies want to implement it to get rid of people, but it’s a tool,” he said. Just as a forklift moves more freight than a hand truck, AI lets employees crunch a lot more data than if they relied solely on Excel. Helping employees understand that benefit can bring them on board. “Transparency and honesty are the best policy,” Quinn said. “In the absence of information, humans fill in worst-case scenarios. So be transparent — here’s what we’re doing, why, what we’re hoping to accomplish, and how it’s going to help you.”

Messaging starts at the top, Oca added, and these initiatives work best when they are backed by the entire C-suite. Sponsorship from the CFO and COO along with support from the chief information officer or chief development officer illustrates that any automation investment isn’t just a tech roadmap. “If positioned as an IT initiative, it will fail,” he warned. “It has to be coded as the new way of working. Enterprises that run successful transformations put the employee and its value proposition at the forefront, not AI, digital, or tech.”

3. Reskill Employees
Putting employees at the center of AI transformation often requires investment in reskilling. AI may be able to take over much of the grunt work, but supervisors, operators, and planners should be trained to interpret the output — and rapidly determine whether the machine has generated a nonsensical or inaccurate response. This kind of mistake, known as “AI hallucination,” occurs when systems fill in gaps with nonexistent or misinterpreted data. The result might look plausible, but closer inspection can reveal underlying flaws. When it comes to acting on an AI’s analysis or recommendation, humans are needed to validate the information and make the call. “They still have power and need to make decisions,” Oca said. “What’s changing is on what and how.”

Some companies may choose to rely heavily on third-party services to implement AI solutions, but those looking to invest internally may need to adjust their talent strategy. “Bring in data scientists, data engineers, and machine learning (ML) engineers,” Oca said. The supply chain practitioners already in the organization may also need training to better collaborate with these new specialists. “You will need supply chain people who can translate the needs for the scientists and engineers so that they can code the agents, algorithms, and workflows,” he added.

Once the organizational groundwork is in place, distributors can begin applying AI to specific warehouse and inventory processes.

4. Improve Forecast Accuracy
Inventory management is one of the most impactful components of distribution performance — too few items in stock can create service failures, and too many can constrain capital and clutter operations. Manually reviewing and adjusting SKUs may work well in some situations, but as assortments grow, that approach can quickly become inefficient. Here, too, AI offers a solution. Masked language models (MLMs) help strike a more disciplined balance by analyzing historical demand, seasonality, lead-time variability, and demand segmentation with greater precision than traditional planning spreadsheets.

“Planners now need to shift to trust the outputs and make minimal overrides,” Oca said. “Meaning AI does the work and they focus on the exemptions. In situations where they manage hundreds of SKUs, they have no choice but to trust some of these outputs for low-priority SKUs.”

Quinn recalled a distributor that launched a new branch and decided to rely almost entirely on an AI system to drive demand planning for the location. The warehouse racks were often empty enough that workers could cleanly throw footballs through aisles of shelving, but the company never missed any orders. “The potential with demand planning is huge,” Quinn said. “Too much inventory makes the warehouse less efficient, drains capital, and decreases turns. When you have the right amount of product — not so many ‘just in case’ items — cash flow is healthier, turns are higher, and profit margins improve.”

5. Reduce Safety Stock
Planning traditionally operates on fixed weekly or monthly cycles with safety buffers absorbing variability. AI-enabled real-time analytics are reducing that dependency. Oca described the shift as a “tremendous change in terms of how inventory, pricing, and service will now be managed.” Instead of waiting for the next planning run, real-time inferencing reacts to continuously updated demand signals with dynamic readjustment of reorder points and lead times.

What that translates into is the ability to lower safety stock without eroding service. Consider a chain restaurant procuring supplies ahead of a limited-time menu rollout, a situation where demand often ramps up faster than projected. With real-time analytics, inventory rebalancing can happen proactively instead of only after product goes out of stock.

6. Optimize Slotting
Where traditional slotting is often reactive and based mostly on velocity, AI-driven slotting is predictive and accounts for more subtleties. “You don’t want to change your layout frequently because it’s a chore,” Quinn said. “But looking at maybe your top 100 items once a month and adjusting based on seasonality and those sorts of things. Instinctively you think you’re going to sell more air conditioners in spring and summer and more heaters in winter. But it’s more nuanced than that. AI can pick up patterns that humans can’t.”

MLMs can be used to predict the best slotting location for new products by analyzing factors such as dimensions, weight, and similarities to existing SKUs. They’re able to highlight correlations between products that could be missed with manual analysis. If 75% of customers who order a specific oven also order the same cleaning accessory, for example, the system would recommend placing them closer together to reduce pick-path distance and cross-zone congestion.

Using that same pattern recognition, AI models can also analyze error history and identify products that are prone to picking errors, such as situations where two SKUs have similar packaging, naming, or item numbers. In those cases, the AI may recommend physically separating the products to reduce the chance that employees pull the wrong item from the shelf.

7. Align Labor with Volume
Labor challenges are often attributed to headcount, but sometimes the real challenge is timing. AI labor forecasting can help supervisors align staffing levels with projected inbound and outbound volume to reduce congestion and improve throughput.

For instance, a midweek inbound surge can be anticipated rather than being reacted to. Armed with that foresight, distributors may choose to shift replenishment resources to receiving in advance and expand picking coverage ahead of shipping cutoffs.

8. Prevent Equipment Downtime
AI can use telemetry from conveyors, forklifts, and sortation systems to automatically predict equipment failures before they disrupt operations. Transitioning maintenance from reactive to predictive is a powerful way to not only minimize maintenance costs and downtime but also ensure peak day-to-day performance.

As a proof of concept, in 2025, DHL conducted a test of portable predictive maintenance tools with AI. What the company called a “doctor in a box” was a package with advanced sensors and cameras inside. The package was placed on the belt so its sensors could detect vibration, orientation, sound, distance, and light levels as it moved through the sorting system, while simultaneously recording video and audio for analysis. Though it provided early indications rather than outright diagnostics, it quickly proved useful in identifying issues that would otherwise go unnoticed, such as critical vibration, unnecessary strain on packages and equipment, and damage risks at higher sorting speeds.

9. Maximize ERP and WMS Systems
Beyond operational planning and execution, AI is also beginning to reshape the core systems distributors rely on to run their businesses.

“You’re seeing enterprise resource planning (ERP) and warehouse management system (WMS) providers embedding AI and machine learning into their systems,” Quinn said. “Some generative AI applications within an ERP are answering questions like ‘How do I do [blank]?’ or ‘How do I create a purchase order?’ So, it’s creating a better help function for an ERP, which can be extremely difficult to figure out. But if you’ve got it hooked up to an LLM (large language model) that can research that and give very step-by-step directions.”

Integrating AI into a WMS, meanwhile, unlocks optimization opportunities such as dynamically rerouting picking paths based on congestion and order priority, or updating wave plans automatically as carrier cutoffs approach.

10. Assess Results
Candidly evaluating the impact of any data or AI initiative is crucial to long-term success. When distributors achieve their targeted results, they can build on it by asking how to take it to the next level. “Ensure there is domain ownership within each function, with clear KPIs to measure and report progress on performance,” Oca advised. “Realign KPIs to reflect the additional horsepower that AI brings to the business and be transparent about how this will be tracked and reported.”

If results are lackluster, it’s an opportunity to learn and pivot. “You’re going to fail, but you want to start,” Quinn said. “The first time you tied your shoes, you weren’t very good at it. But you got good.”

What Lies Ahead
With so many novel opportunities in such a rapidly evolving landscape, determining if and where to invest can be daunting. Quinn recommended the Applied AI for Distributors conference — held June 23-25 at the Marriott Chicago O’Hare by Distribution Strategy Group — as a great place to start. Dedicated to real-world AI execution in wholesale distribution, the conference aims to connect attendees with technology providers whose platforms are already operating inside distributor environments while also offering workshops and early-access sessions. The Tech Talks portion of the 2026 FEDA Annual Executive Leadership Conference, Sept. 15-18 in Park City, Utah, is another place where distributors can learn more about how these kinds of technologies are being applied to real-world scenarios.

AI adoption is quickly shifting from experimentation to operational necessity. Distributors can now find a technological solution to help uncover and manage nearly any key performance indicator they can think of, from B2B returns to predictive maintenance. Still, Quinn offered a word of caution for those considering investing in AI solutions. “The biggest thing is to not chase shiny toys. Any tool you employ has to fit your strategy as a business. You don’t want your business to fit the technology; you want the technology to fit your business.”