Retailers want ‘delightfully human’ AI to do your shopping, but will the chatbots go rogue?

2026-03-05 · Source: AI (artificial intelligence) | The Guardian · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Retail Technology & Operations, Emerging Technologies & Innovation · Depth: Intermediate, medium

Summary

Australian retailers, including Woolworths, Coles, and Wesfarmers, are developing "agentic" AI shopping assistants designed to plan meals, organize parties, and automate shopping. These advanced bots, often built using large language models like Google's Gemini, aim to mimic human behavior and act autonomously to achieve objectives. However, the technology faces significant challenges, as demonstrated by recent mishaps. Woolworths' virtual assistant, Olive, was criticized for its overly human-like personality, while Bunnings' chatbot provided illegal electrical advice, and Air Canada's bot offered incorrect refund information, leading to a lawsuit. Experts highlight governance issues, privacy risks due to increased data access, and the difficulty of balancing a "delightfully human" persona with strict guardrails to prevent errors or "rogue" behavior. Initial tests by Guardian Australia on various retail bots showed marginal results, indicating the technology is still in its early stages.

Key takeaway

For Product Managers overseeing AI deployments, you must prioritize robust governance and clear operational boundaries for agentic AI. While aiming for a "delightfully human" experience, ensure that personality features do not lead to customer frustration or provide incorrect, potentially liable, advice. Thoroughly test AI responses in diverse scenarios to prevent "rogue" behavior and protect your brand's reputation and legal standing.

Key insights

Agentic AI shopping assistants offer autonomy but pose risks in balancing human-like interaction with reliable performance.

Principles

• Companies are responsible for chatbot advice.
• Guardrails limit AI flexibility and risk.

Method

Newer AI retail bots use large language models (LLMs) to learn and generate varied answers, moving beyond rules-based decision trees to mimic human behavior and act autonomously.

In practice

• Implement strict guardrails on AI agents.
• Test AI for unintended "human" responses.

Topics

• Agentic AI
• Retail AI
• Large Language Models
• AI Chatbots
• AI Governance

Best for: Product Manager, CTO, VP of Engineering/Data, AI Product Manager, AI Ethicist, Executive

Related on AIssential

• When a chatbot runs your store — Large language models struggle with reality adherence, making them unreliable for autonomous transactional roles.
• How Do You Market to an AI Customer? — AI agents are a distinct customer type requiring a new marketing science beyond human persuasion.
• Scaling Customer Experience with Operationalized Agentic AI - with Shezan Kazi of Dialpad — AI agents excel at high-volume, deterministic tasks, requiring seamless human handoff for complex or ambiguous cases.
• Trust Issues Could Make or Break Agentic Commerce — Agentic commerce hinges on user trust, necessitating system design that prioritizes user interests and mitigates inherent risks.
• Woolworths’ AI agent rambled about its ‘mother’. It’s a sign of deeper problems with the tech rollout — Companies deploying customer-facing AI must ensure accuracy and accountability, as they own the system's outputs.
• Shopping Online Is About To Change Forever — Agentic commerce employs AI agents for proactive, personalized product discovery and visual recommendations, streamlining online shopping.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by AI (artificial intelligence) | The Guardian.