An enterprise retailer has made an eight-figure investment automation. State-of-the-art goods-to-person systems, autonomous mobile robots (AMRs) in warehouses and distribution centres (DC).

The system is live, but it’s producing errors.

Inventory accuracy is just 60%. Warehouse management systems (WMS) and last-mile delivery platforms don’t talk. Pick-and-pack processes weren’t formally documented before the robots arrived.

And somewhere upstream, there’s a ‘process’ that only Susan in the warehouse knows about, because she’s been quietly handling it for a decade with a spreadsheet on her desktop.

This scenario might be fictitious, but it’s indicative of a broader pattern playing out across enterprise retail. After years of hype, automation has reached its correction phase in 2026.

Today, boards aren’t asking ‘what are we automating?’, they’re asking ‘what’s the ROI?’ and ‘can we solve the problem through other means?’.

This week, Delivered highlights what happens when you automate before you’re ready. According to Leonie McCarthy - whose consultancy 6R Retail helps retailers implement major technology and systems change - it’s more common than most retailers realise.

TL;DR

• Automation scales what already exists (the good, the bad and the ugly); if your processes are broken, it scales the problem.
• Retailers that get the most from automation won’t be those that moved fastest; it’ll be those that were most honest about their operational realities first.
• Before investing in robotics, ask if your data, processes and systems integration would survive the transition.

The audit era begins

After years of investment, 2026 marks a turning point for automation. Many retailers are realising that major automation isn’t an automatic win if it’s built on broken processes, fragmented inventory, or poorly-designed operations.

Automation scales inefficiency, not just efficiency

The core problem with automating a broken warehouse or DC isn’t that the technology fails. It’s that the technology succeeds, but at the wrong things.

Automation doesn’t fix poor slotting, inaccurate inventory, or inconsistent processes, it amplifies them. And it does so faster, at higher volume, with less scope for human correction.

“Things take a lot longer and cost a lot more than anyone’s expecting,” Leonie continues.

“You get a certain way through, realise it’s not working, and then you have to go back to the beginning to fix the data. Or you go live and have an awful customer experience, and then you have to go back and fix the data.”

Our State of Shipping Report found that 16% of retailers said AI and automation was their top investment priority in 2025, with 10% prioritising inventory accuracy and visibility.

Accurate data is a non-negotiable first step, before automation. Everything that automation will eventually be asked to act on (product attributes, warehouse locations, carrier allocation rules) must be clean before the system can commence.

“It’s the foundation work,” Leonie says. “If you don’t have accurate data, you’re building on sand.”

As more retailers activate ship-from-store, the process requirements are identical. Without clean data and documented workflows, automation doesn’t improve the experience, it exposes every gap in it.

The Amazon lesson most retailers skip

For the first time in its history, Amazon’s logistics engine is sustainable.

Only once, in 2018, were Amazon’s shipping costs lower than its retail sales. But that’s been the norm since 2022. Investments in automation and robotics are now delivering operational leverage.

Leaked documents, reported by the New York Times, suggest that Amazon plans to replace half a million jobs with robots.  Morgan Stanley estimates that if 30-40% of Amazon’s US orders are fulfilled through its growing network (40 by the end of 2026) next-generation warehouses by 2030, the company could save $10 billion a year.

Amazon is the automation benchmark every enterprise retailer holds itself against. What’s less frequently discussed is the sequence of how it got there. Years of obsessive process standardisation came before robotics.

Amazon made its processes ‘boring’, then used automation and robotics to scale an efficient, profitable, repeatable machine.

“Amazon had years of not posting profit. I don’t know how many Australian retailers would be allowed to function like that,” Leonie adds.

“Every organisation has what I call ‘under-the-desk’ processes. Things that are invisible to most people because someone is quietly handling them. When you try to automate, you find all these little gaps you didn’t even know were there.

“You have to be honest about what your actual processes are, not your aspirational ones. The difference between those two things is often quite a chasm.”

Then there’s Ocado, whose model works because the entire operation was designed around automation from day one - not retrofitted into an existing warehouse.

Most Australian enterprise retailers are doing the harder thing, and understanding why these two benchmarks succeeded makes failures easier to diagnose and harder to excuse.

Why investments underperform in the first 24 months

• Poor data foundations: Inaccurate inventory, incomplete product attributes and location data, means the system has nothing reliable to act on.
• Undocumented process gaps: The ‘Susan’ problem. Processes that exist in one person’s head don’t survive automation. They become expensive exceptions.
• Automating the wrong thing first: Customer-facing features before back-office basics. “How long does it take to get label printing right? It’s one of the biggest, consistently underestimated, pain points in every warehouse implementation.”
• No clarity: Misaligned definitions, expectations and success metrics between stakeholders.

“The number of times I’ve had an entire conversation about a word definition,” she continues.

“Like what is ‘pack’? Is it a pack inside a plastic bag? Is that plastic bag inside an inner carton? Is that inner carton inside an outer carton? I have literally had to draw diagrams and go to the warehouse floor to establish what a ‘pack’ actually means because the client assumed the vendor understood, and the vendor assumed the client had aligned internally. Neither was true.”

She also points to undertesting as a persistent issue. “I’ve had standoffs with organisations where I say we need four weeks of testing and they say two will be fine. You usually end up needing six. That costs the client more, but not as much as releasing something that hasn’t been tested rigorously enough.”

For retailers expanding into ship-from-store, the risk compounds across a larger, more complex footprint. Real-time inventory visibility across a sprawling store network is table stakes for the model to function.

The failure patterns are fixable; the harder problem is cultural.

Thinking like a technology company

Amazon and Ocado aren’t just retailers that use automation well, they’ve adopted a ‘we’re technology companies that operate retail infrastructure’ mindset. Their success is a product of that identity. Teams and roadmaps are built to think ahead, adapt quickly, and evolve as the technology does.

“More retailers need to adopt and adapt to those skills. We need to be constantly thinking about where the technology is taking us now,” Leonie adds.

“The question I’d encourage retailers to ask isn’t ‘what should we automate?’. It’s ‘where do humans create value that can’t be automated? And how do we design the systems that support that?’”

Before you automate: a readiness checklist

Before automating, the real questions must be answered.

1. Data integrity

• Is inventory accuracy above 95% across locations, SKUs and product attributes?
• Is product master data (dimensions, weights, barcodes) complete and validated?
• Are delivery zones, carrier options and allocation rules validated across all fulfilment locations?

2. Process documentation

• Are actual, not aspirational, processes and standards (including informal ones) documented?
• Are exceptions to these rules clear?

3. Systems integration

• Do WMS and last-mile platforms share data in real time, with end-to-end visibility?
• Are label printing and dispatch workflows stable and tested?

4. Scope and language alignment

• Are key terms (pack, SKU, order, location etc) defined and agreed upon?
• Is scope documented with a clear change process?

5. Testing and change management

• Are testing timelines realistic, with buffers built in proactively?
• Is there a named owner for each process, a change management plan, and a contingency if go-live thresholds aren’t met?

The retailers who will get the most from automation in the next few years won’t be the ones who moved fastest. They’ll be those that were the most honest about their operational realities and fastidious in their preparation before they started.

The technology is ready. The question is whether the warehouse, DC or store network underneath it is.