In Peak 2026, AI won’t fix a delivery promise you shouldn’t have made

Retailers are investing in AI to make ecommerce delivery faster, cheaper and more efficient, but the real opportunity lies in knowing what to promise customers in the first place, argues Gavin Murphy, CMO, Scurri.

With Peak in full flow, we can see that AI has for many moved from experimentation into operations in a whole bunch of different areas; how can AI automate carrier allocation, predict failed deliveries, manage integrations, monitor performance and reduce fulfilment costs.

My worry is that, in the mad dash to automate delivery, fashion retailers might overlook something very important; are they improving the experience customers actually want, or simply becoming more efficient at delivering an experience of their own choosing?

Two recent research studies commissioned by Scurri, examining the attitudes of UK retailers and consumers, showed up a potentially costly disconnect between ambitions for the technology and customer expectations.

Among retailers, 58% say they are likely to use an automated AI system to manage ecommerce delivery strategy and carrier operations. Their expectations are as always high; some 34% identify automated carrier allocation as a leading opportunity for AI, while 33% point to predicting and preventing failed deliveries, and 31% to monitoring delivery issues and carrier performance.

This is to be expected, particularly during Peak when a huge spike in volume puts a ton of pressure on fulfilment operations. But consumer research suggests that delivery innovation needs to be thought about against a rather different set of expectations.

Customers are not demanding a race to the doorstep; for them, affordability almost always matters more than outright delivery speed. While 61% regard free delivery as important, only 29% prioritise next day delivery and just 12% same day delivery. More importantly, 67% say they would definitely or probably accept longer delivery times during Peak in exchange for free or cheaper delivery.

This challenges the assumption that faster delivery is necessarily better delivery. Retailers have spent years competing on speed, conditioning customers to expect increasingly ambitious delivery promises, yet many shoppers appear willing to exchange speed for value, provided they understand when their purchases will arrive.

This must have an effect on how retailers deploy AI. An automated system that relentlessly selects the fastest available carrier may meet operational benchmarks but not improve customer satisfaction or protect margin. An intelligent system that considers the customer's preference, delivery economics and likelihood of successful fulfilment could create considerably more value. So, the objective should be to make every delivery promise more appropriate to customer preferences.

And consider, there is a cost to promising too much. During Peak, a delivery promise is only as reliable as the network supporting it. Carrier capacity fluctuates, fulfilment centres become congested and available inventory may be distributed across multiple locations. Promising next day delivery indiscriminately creates exposure to expensive carrier services, operational exceptions and ultimately disappointed customers.

Fortunately, AI can make those decisions more intelligently. By bringing together inventory location, carrier capacity and historical performance, delivery destination, service costs and customer preferences, AI can help retailers determine which delivery options are commercially rational and operationally achievable for each order. And, crucially, choices that reflect the realities of the network.

For instance, a shopper who values affordability could receive a lower cost delivery option with a longer but still guaranteed timeframe. Someone purchasing an urgent gift might be offered a faster premium service, provided the retailer has sufficient confidence that the promise can be met.

How these offers are pitched depends on getting communications right. After all, even the most sophisticated delivery operation cannot eliminate disruption. Peak inevitably creates exceptions, from missed collections and routing problems to weather-related delays. However, here comes the opportunity; 46% of consumers say they would accept a delayed delivery if warned in advance. When something goes wrong, 42% want to be informed before the expected delivery time and 40% want a revised delivery estimate. Retailers can mitigate these problems using carrier performance and shipment data and so identify emerging problems before customers become aware of them. This capability has commercial value; after all, almost half of retailers surveyed, 48%, identify 'Where Is My Order?' enquiries as a leading delivery exception issue.

There are other reasons retailers should not pat themselves on the back just for getting AI implemented. 31% of consumers do not believe AI would improve their delivery experience. This scepticism reflects the fact that customers judge delivery by results, not the sophistication of the systems behind them.

Retailers should ideally deploy AI selectively to enable decisions that influence outcomes so they can match orders to suitable carrier services, identify potential failures, maintain visibility across delivery networks and communicate intelligently when circumstances change.

This also relies on dependable data, integrated carrier systems and sufficient operational control to act on AI-generated insights. Automation cannot compensate for fragmented information or processes that cannot respond to changing conditions.

During Peak 2026 we are certain to see further advances in AI-powered delivery management but the lesson learned may well be that faster fulfilment is not necessarily the same as better fulfilment.