Customer Service

How AI is helping fix retail’s $1.7T inventory problem

When retailers get inventory right, the impact goes beyond reducing waste. They free up margin to invest in better experiences, build resilience into their operations and give associates the space to do what they came into retail to do: understand customers, curate great products and bring shop floors to life.

Adobe Stock

July 9, 2026 by Nicola Bond — Co-Founder and CEO, Merchmix

Summarize with

On any day of the week, retailers can watch one of their biggest problems walk out the door. A shopper enters the shop to buy a trending dress they saw on Instagram last night, only to find their size is already out of stock. The shop is full of other designs and sizes, but what's in demand is unavailable.

A few miles away, the same product sits unsold in another store, bought in the wrong quantity for the wrong customer.

Behind the scenes, a retail buyer works from a confusing and outdated stock spreadsheet, trying to predict demand with tools that don't belong in 2026.

Inventory distortion problem

This is what sits beneath retail's $1.7 trillion inventory distortion problem: too much stock where it's not needed and too little where customers are ready to buy. McKinsey puts unsold U.S. goods at around $740 billion in 2023, with another $1.2 trillion in potential sales lost because products simply weren't on the shelf when shoppers wanted them.

For retailers who have taken a panicked call from a store about emptied shelves at the start of a viral trend, or walked a markdown-heavy shop floor thinking of margin leakage, this isn't an abstract number. It is a daily drain on cash, confidence, and customer trust.

Retailers now have an opportunity not only to reduce waste, but to reimagine how human judgement and AI can join forces so that every SKU bought, moved, and placed in store has the best chance of being in the right place at the right time — reducing reactive firefighting.

That shift depends on turning insight into action, faster, and operating from a single source of truth across the business.

But what does that actually look like on the shop floor?

Where agentic AI comes in

"AI in retail" has until now mostly lived in corporate pitch decks and conference keynotes. Shop-floor reality has looked very different: slow planning cycles, fragmented systems, and buyers relying on gut feel supported by last year's run-rate.

Agentic AI is the first wave of intelligent technology with the potential to genuinely change that reality.

Agentic systems don't simply produce a forecast and stop there. They behave more like a tireless back-room colleague, continuously reading demand signals such as what's selling in a city-centre store, what's spiking online after a viral trend, and what's stalling elsewhere, and acting within agreed, human-led guardrails.

When too much stock is ordered, they spot slow movers early and help reroute or reprice before markdowns build. When too little is bought, they surface emerging demand quickly enough to replenish, not just analyse the miss after the event.

The impact is simple: less capital trapped in the wrong stock, fewer disappointed customers, and far fewer late-night spreadsheet sessions.

The challenge is that most inventory systems were built for an unhurried world, with long seasons, slower trends, and forecasts that could afford to be wrong for a while. But legacy systems don't work in a TikTok world.

The gap cost not small

Today, trends are born and die on social media, supply chains are more fragile, and customers expect the same immediacy in-store that they get online. Yet many retailers are still managing this with static forecasts, siloed systems, and reports that tell you what happened weeks ago rather than what's happening right now.

The cost of that gap is significant. The 2024 Worldwide Inventory Distortion Study by IHL Group puts it at 6.8% of global retail sales, driven by both lost sales from out-of-stocks and markdowns on excess inventory. Increasingly, understocking is proving even more damaging than overstocking.

Inside organisations, talent is not the problem. Teams are forced to choose between reacting to yesterday's crisis and planning for next season, often without a single shared view of the truth. Systems don't talk to each other, and inventory distortion stops being "just" an operational issue and becomes a strategic risk.

At its core, this is no longer a forecasting problem. It is a decision-speed problem.

The most important shift now is moving from one-off forecasting to always-on execution.

Agentic inventory operating systems connect insight directly to action. They ingest signals from across the business in real time and make continuous adjustments within clear rules, rebalancing stock, refining buy quantities, and recommending price moves. People can then step in for the decisions that truly need judgement, while the system handles the rest.

For buyers and planners, the job begins to change. Instead of reconciling competing reports and debating whose spreadsheet is right, they can focus on what they are best at: curating products, shaping assortments, and making strategic calls on risk and innovation.

The shift taking place

It is a shift from a planning mindset to an execution mindset. Rather than waiting for the next meeting, decisions can be made continuously, in the background, at SKU and store level.

While true agentic AI is still in its early days within retail, there is now enough evidence to move beyond theory and start proving what happens when it is applied well. McKinsey reports inventory reductions of 10–20% for companies using autonomous planning systems, while maintaining or improving availability. Danone has seen lost sales fall by around 30% after introducing AI-powered demand modelling, and a 2023 Deloitte survey found that 62% of retailers using external data to enhance analytics reported measurable performance improvements.

On the ground, the difference is becoming visible. Store teams receive stock that better matches what their customers are actually asking for, rather than shipments determined by historic averages and hope. There are fewer rail-to-rail markdowns at the end of the season and less of the familiar frustration of missed demand.

Adoption, however, is still cautious. IHL Group's research in 2025 suggested that only around one in four retailers had successfully deployed AI or machine learning in the areas most affected by inventory distortion.

The hesitation is understandable. Every transformation looks neat on a slide and messy in reality. But AI in inventory is not about replacing the buyer's instinct or the planner's expertise. It is about giving them a clearer, faster, more connected view of reality, while taking repetitive, low-value work off their plate.

When retailers get inventory right, the impact goes beyond reducing waste. They free up margin to invest in better experiences, build resilience into their operations, and give their people the space to do what they came into retail to do: understand customers, curate great products, and bring shop floors to life.

About Nicola Bond

Nicola Bond is the Co-Founder and CEO of Merchmix. She has over 20 years of expertise spanning retail and tech; scaling and launching Marks and Spencer’s International Division whilst migrating from legacy Range Planners to SAP, and continuing onto pivotal roles at ASOS.com, Debenhams, Best and Less, and SABA in head of product roles.

Connect with Nicola: