Retail Leaders Reveal AI Adoption Is a People Problem, Not a Tech One

Eight retail leaders recently gathered in London for a UiPath advisory board, and their most consistent finding was that AI failure is rarely a technology issue. According to the UiPath Blog, the dominant theme across every session was adoption and change management, not model capability or platform features. The core insight for operations teams is that the technology is ready, but the human and data infrastructure often is not.

What happened

The discussion took place under the Chatham House Rule, meaning no specific quotes are attributed to individuals. However, the UiPath Blog reports that every retailer present identified adoption as a significant challenge. Many are dealing with processes 20 to 30 years old, shaped by isolated teams, with some now employing dedicated change management staff just to keep initiatives moving. A quieter barrier described was the tendency for teams to focus on what AI cannot do rather than what it can, leading to a gap where modern planning tools exist but real work still happens in spreadsheets.

Data integrity emerged as a critical blocker. UiPath’s "Reaction Gap" survey of 500 UK retail leaders, cited in the blog, found that while 98% of leaders describe their systems as connected during peak trading, only 25% have the real-time visibility to act on issues as they happen. The blog notes that the hardest problem is not a lack of data, but getting numbers to match across back-end systems where departments manipulate the same data differently. Poor or siloed data blocks agentic use cases before they even start.

Regarding workflow design, the leaders described a model where agents handle the routine 90% of decisions, leaving humans to approve the 10% requiring judgment. However, this requires the system to explain its reasoning and risk. A significant tension was raised about training: if agents remove the manual pre-work (like building trade packs), how will new graduates develop the judgment needed for high-level decisions? Furthermore, while 97% of UK retailers have implemented AI in some form, 47% are still waiting to see commercial impact, partly due to the long lead times required to prove ROI.

Why it matters for small operations teams

For small and mid-sized companies, this validates a practical, low-risk entry point into AI automation. The retail leaders’ advice to "start small" and target the "middle band" of work—tasks that are too tedious for humans but too complex for simple macros—is directly applicable to your operations. You do not need to overhaul your entire ERP or supply chain to see value. Instead, focus on contained, measurable wins like non-stock procure-to-pay processes. The blog highlights that supplier onboarding can take up to eight weeks, creating exceptions and manual chase-downs. Automating this specific bottleneck provides a visible, quick win that builds internal trust. The most important metric is not immediate ROI, but internal demand: when your team says, "I need this to work," you have proven the value. This approach mitigates the risk of change management fatigue by making the benefit tangible to the employees doing the work.

What to do this week

1. Identify one "middle band" process in your operations that involves high-volume, low-judgment tasks, such as supplier onboarding or invoice reconciliation, and map out where manual handoffs occur.
2. Select a single, contained data set for this process and clean it to ensure a single source of truth, addressing the data silo issue before introducing any AI tool.
3. Pilot an AI agent on this specific workflow for one month, measuring success not just by time saved, but by whether the team members explicitly request to keep the tool after the pilot ends.

Sources

• We asked eight retailers to be honest about AI — UiPath Blog