The problem comes first – how Linen House decides where AI belongs

Last Updated on 29 July 2026

Does a microwave really need AI? Tez Osman recently bought one that has it. For the CTO / Director at Linen House, the three-plus-decade-old Australian bedlinen and homewares brand, the appliance says everything about the state of the market – the AI label is turning up on everything and its presence tells you almost nothing about whether a product will help your business.

At recent industry events we have heard the same sentiment again and again – leaders are overwhelmed by the volume of AI being pushed at them, unsure where to start and worried they are already behind on adoption.

Treat AI like any other technology

Linen House is a large and multi-faceted operation. The family-owned business runs more than 20 stores nationally, a direct-to-consumer web business, a wholesale arm and an extensive drop shipping operation with the country’s largest retailers, all built on cloud ERP NetSuite as the source of truth for close to a decade. Osman has watched plenty of technology cycles come and go from where he sits.

His method for AI strips away the novelty entirely. Osman describes treating each new tool like any other piece of software – what is the offering, what does it actually do, and how does that help the business. Then the standard due diligence follows: the overhead and the true cost of ownership, then evaluation and testing.

The Linen House team never starts with an impressive tool and hunts for somewhere to plug it in. It starts with the areas of the business that take too much time or produce too little, then asks whether an AI tool can improve them. If the answer is yes, the tool gets evaluated on its merits. If the answer is no, the shiny object goes back on the shelf regardless of how it is marketed.

What adoption looks like in practice

That discipline has still produced deep adoption. Linen House connected its business Claude account to cloud ERP NetSuite through an MCP connection, and the result has spread well beyond the IT function. Managers, directors, supervisors and admin staff now run their own reports, formatted the way they want them, on schedules they set themselves.

The eCommerce team has gone further. Shopify offers its own connection to Claude, and pairing the two means data can be pulled from one system, collated and applied to page builds and updates in the other. The same setup cleans product data and replaces the tedious spreadsheet work that used to soak up hours. The team also uses it to analyse Linen House’s own websites alongside competitor sites.

The company came to this via an earlier phase of workflow automation tooling, which it began exploring a couple of years ago. Newer agentic tools have since replaced much of what those early experiments were reaching for – itself a lesson in how quickly this category moves.

Where AI is the wrong tool

The most useful part of Osman’s account is his honesty about the failures. His team has repeatedly gone down the path of automating a task with AI, confirmed that the technology can technically do it, then discovered that conventional automation is a better fit. Deterministic automation does the job once, does it the same way every time and stays done.

He gives a specific example. One report ran correctly four times in a row, then failed on the fifth. Each run already took twice as long as the staff member who normally produced it, and every failure meant supervision and reruns and consumed tokens. Osman’s assessment was straightforward – yes, the tool could technically do it, but no, not as well as regular automation. The person kept the job and checks the output in five minutes anyway.

There is a human boundary in the strategy too. Customer service phone lines at Linen House remain staffed by people because person-to-person connection is central to how the brand wants to treat its customers. Osman does not rule anything out permanently. He does insist that any change clear the same bar as everything else.

What other retailers can take from this

Linen House’s story has two important takeaways. The first is the sequence – problem first, then tool. Adoption pressure is real, and the retailers navigating it best are the ones who let their own friction points set the agenda, then evaluate AI against ordinary software standards of cost and reliability.

The second is the foundation underneath it. Every practical win Osman describes depends on structured, trustworthy data sitting in a system of record that AI tools can safely reach. An assistant connected to a clean ERP can generate reports, reconcile channels and answer real operational questions. The same assistant pointed at fragmented systems produces fragmented answers. The unglamorous work of getting the data layer right is what turns AI from a demo into a daily tool.

Osman shared the full story – from the original ERP decision through the platform consolidation and his assessment of AI in retail – at a NORA webinar on 14 July alongside panellists from NetSuite and Annexa. The session is available on-demand here.

Frequently asked questions