Retailers need a system which connects the front end of the journey to the operation behind it to meet customers’ expectations and keep ahead of demand, says Nishith Rastogi, Founder and CEO, Locus.
ChatGPT and Google Gemini are overtaking social media as a starting point for product discovery and agentic AI is turning that discovery into real conversion as OpenAI and Google develop protocols that allow customers to make purchases within AI platforms. Locus research found that 37% of consumers are more likely to try a new brand or product after using AI. For retailers, this is a positive opportunity, introducing customers to businesses they would never have found through conventional search.
The friction occurs at the next stage, fulfilment. When an AI recommendation leads a new customer to an unfamiliar retailer, that first delivery becomes the entire relationship. Get it right and the AI’s recommendation is validated. Get it wrong and the retailer rarely gets a second chance to prove the algorithm was right about them.
Fulfilment has been left behind
For someone buying from an unfamiliar retailer, that first delivery decides whether the recommendation feels well placed and whether they return. Locus research reveals only 7% of UK shoppers say fast delivery promises are consistently met. At 3pm, next-day delivery may still show as available online even when the warehouse has already hit the volume it can clear before collection.
This is where current fulfilment operations fail. An AI agent can pull up live pricing and availability to make a recommendation, but fulfilment cut-off times don’t always match this and the conversation between the cart and the warehouse is often fragmented, preventing things like date changes and courier capacity from being reflected in real time.
AI agents will intermediate more than US$15 trillion in B2B purchases by 2028, which means the same disconnect between what an AI system promises and what can actually be delivered will show up in procurement and supply chain relationships too. This is a bigger challenge for smaller and mid-sized retailers, which don’t have the logistics headcount and infrastructure of household names to keep up, but are now more visible thanks to AI discovery. Failure to fulfil will land disproportionately hard on the retailers least able to recover from it.
Fulfilment needs to become agentic
The solution isn’t to slow down AI-driven discovery, at this stage that would be impossible, with agentic commerce predicted to redirect as much as US$3–5 trillion in global retail in the coming years and ONS figures showing online share of UK retail sales climbing to 29.4% in June 2026, the highest proportion since April 2021.
Really, agentic commerce needs to be met by agentic fulfilment, so that what’s promised at the point of discovery is genuinely deliverable by the time an order reaches the warehouse. Warehouse management, transport management and checkout need a shared, live picture of the same factors.
That starts with capacity. A delivery promise generated at checkout is often based on a cut-off time rather than what the warehouse can actually clear that day or what carrier capacity is genuinely available for collection. If checkout systems had a live view of warehouse workload and carrier capacity and could adjust the available delivery window as that workload changes, retailers could stop making promises they know, in real time, they can’t keep.
For any retailer working through a shared logistics network or third-party provider, this matters even more. Capacity committed to one client affects what can be offered to the next, which means checkout needs visibility across the shared operation.
Even after a sale is made, elements of the delivery network are prone to change at the last minute whether that’s a carrier reducing capacity, a route disrupted by traffic or weather or a warehouse running behind schedule. Recovering from that depends on spotting the problem early enough to act on it.
A transport management system monitoring deliveries in real time can flag when a plan is becoming unworkable while there’s still time to do something about it, whether that’s reassigning an order to another carrier or adjusting the route. Anything that falls outside agreed cost, service or carrier rules should escalate to a human decision-maker with enough runway to act and the resulting change needs to impact the customer’s ETA rather than resulting in a parcel that never arrives.
A refreshed technology stack
For CIOs and technology leaders, this means the systems behind the checkout need to operate on the same timescale as the AI shaping the demand in front of it. Delivery capacity needs to be treated as live operational data. That means connecting inventory, warehouse workload, carrier availability and route conditions into a single live source that checkout, fulfilment and transport systems can all update in real time.
And increasingly, it means building in a safety net for disruption, with clear rules for when a human needs to step in.
We won’t replace judgement with automation, but we can close the gap between what’s promised and what’s possible at the speed AI is now setting new expectations. Bringing together agentic discovery and agentic fulfilment will turn AI-driven discovery into growth and retained customers.
AI can put a retailer in front of a customer who would never have found them otherwise. What happens after checkout will decide whether that customer comes back.
