By Barley Laing, the UK Managing Director at Melissa

The retail world is in the midst of an agentic commerce revolution, as fast evolving proactive AI agents autonomously research, compare and complete purchases on behalf of consumers.

It is no surprise that agentic commerce is beginning to transform the retail industry, like many others, by automating operations, enabling real-time decision making and personalising shopping journeys.

In fact, it is estimated by Bain that by 2030 autonomous AI agents could drive up to 25 per cent of ecommerce spending in the US.

Consumer adoption and engagement rates with AI agents is also soaring, with recent research by PayPal and Commerce revealing that 64 per cent of UK consumers want to use agentic AI for shopping.

Quality data powers AI

However, AI only performs well if it has a solid data foundation. This means access to accurate, up-to-date data, because inaccurate answers will be provided if retailers don’t anchor their AI initiatives in clean data.

In fact, a study from Gartner at the end of 2025 found that at least 50 per cent of generative AI projects were abandoned after proof of concept due to poor data quality, inadequate risk controls, escalating costs or unclear business value.

Retailers are finding that the greatest obstacle to deploying AI is not the technology itself, but the state of their data.

Access to clean data is essential for those attempting to train, deploy, scale and determine the return on investment (ROI) from their AI initiatives.

Inaccurate data causes failed AI rollouts and delayed ROI due to unreliable automation, with ineffective personalisation and inaccurate recommendations.

This leads to a rise in cart abandonment, increase in costly manual fixes, and an erosion of customer trust, along with a surge in complaints and refunds.

Recurring data problems

Common data issues faced by retailers include outdated and duplicate data. Consequently, the majority of data teams spend more than half their time preparing data, leaving less time for generating actionable insights.

Another common challenge is inconsistent or incomplete data, which can reduce the accuracy of AI recommendations and increase the risk of bias.

Also, there’s issues with legacy and fragmented data that can hinder retailers’ ability to define and maintain consistent business metrics across the organisation.

Finally, poorly structured and non-machine-readable data remains a major obstacle, with many retailers spending hours each month manually reconciling disparate datasets to identify and correct inconsistencies.

Retailers can build AI-ready data in five steps:

1. Cleanse and enhance customer records by verifying names, addresses, emails and phone numbers in batch and in real-time as data is captured at the onboarding stage.
   As part of this process, parse and structure the data into a consistent, usable format to help prevent biased results.
2. Match and merge duplicate records using advanced fuzzy matching to create a single, accurate and trusted customer profile.
   Data duplication is a common issue with duplication rates of 10 to 30 per cent on many customer databases.
   It occurs when errors in contact data collection take place at different touchpoints, two departments merge their data, or when combining datasets after a business acquisition.
3. Enrich customer data with demographic, firmographic, geographic, social media and property attributes, as well as missing email addresses and phone numbers, to enable AI to deliver effective analytics, personalisation and omnichannel marketing.
4. Continuously monitor your data across the entire data lifecycle to prevent inaccurate data from entering your database and ensure it remains clean and reliable over time.
5. Retailers should have well labelled (machine-readable) data to ensure accuracy across mission critical AI applications, because high quality data labelling, along with maintaining data accuracy, helps to maintain consistency, and trust as systems grow.
   It’s particularly important when any errors can have financial, operational, legal, or safety consequences.

In summary

Agentic AI is set to be transformational in retail.

However, for it to operate effectively it’s smart to put robust processes in place to deliver clean, trusted and unbiased data. This will power successful agentic commerce operations which will provide a strong customer experience that supports growth.

Those who don’t have quality data run the risk of delivering AI hallucinations and bad outcomes from their agentic activity.

Published 22/07/2026