Starbucks tests AI-built software to replace microsoft and IBM systems

61

TL;DR:

Starbucks is using AI-assisted coding to build inventory and maintenance software in-house, targeting a reported $400 million annual software bill. The move could reduce vendor dependence, but a recently abandoned AI inventory tool shows that reliability remains the decisive test.

Article:

Starbucks is using AI-assisted coding to build in-house inventory and maintenance software, potentially replacing tools bought from Microsoft and IBM. Some applications could reach stores by the end of 2027, subject to testing. The timing is financial: Starbucks reportedly spends about $400 million a year on software while pursuing $2 billion in broader cost reductions. For enterprise technology buyers, the project tests whether generative AI can reduce vendor dependence without pulling excessive maintenance, cybersecurity and reliability risk inside the company.

The internal plan targets a Microsoft system for inventory tracking and an IBM platform for maintenance workflows. Chief technology officer Anand Varadarajan told employees there were “clear opportunities to reduce the spend in software,” according to a recording reviewed by reporters. The technology team is reportedly aiming to cut its fiscal-year budget by about $30 million, including roughly $10 million in software spending. Starbucks declined to comment on the reported internal presentation, leaving rollout dates and savings targets subject to change.

This is a selective shift, not a wholesale break: Starbucks continues to use third-party software. Scale makes execution more important than coding speed. The company ended a separate AI inventory-counting programme in May after nine months across North American stores; at the end of its second fiscal quarter, it operated 41,129 stores worldwide. A faulty internal tool can therefore turn a modest licence saving into widespread operational friction.

Starbucks states its rule plainly: “If it strengthens the experience, we scale it. If not, we iterate on it.” That discipline will determine whether the initiative becomes a credible template for other retailers or another warning that cheaper code can still produce expensive systems. The first proof points should be deployment accuracy, reduced downtime and measurable licence savings, not launch announcements.