Business

Starbucks building AI-powered internal software to replace Microsoft and IBM tools

Ten months ago, in September 2025, Starbucks walked into 11,000 stores with NomadGo's inventory Artificial Intelligence (AI) tool that promised 99% accuracy and counted up to eight times faster than a human.

After 9 months, in May 2026, according to Reuters, an internal newsletter retired the entire program. Baristas went back to counting milk by hand.

The NomadGo system confused milk varieties, missed products on shelves, and slowed employees down instead of helping them. CEO Brian Niccol had bet on it as part of his turnaround plan. It did not work.

Now, Starbucks (SBUX) is trying AI at scale again, but with a meaningfully different approach.

According to Bloomberg, citing an internal company presentation, the coffee chain is using AI-assisted coding to build its own internal software, specifically targeting tools that currently run on Microsoft and IBM infrastructure, in a move that CTO Anand Varadarajan framed earlier this year as a chance to cut the company's $400 million annual software bill.

SBUX closed July 9 at $106.41, up 2.45% on the session. The stock is up 27.92% year-to-date, according to Yahoo Finance.

Also Read: Starbucks Corporation Latest News and Stories

What Starbucks is actually building this time and why it is different

What matters now is this. The distinction between what failed and what is being attempted now.

The NomadGo inventory tool was a third-party product deployed to consumer-facing operations at the store level. Its failure was operational. Why? Computer vision that could not reliably distinguish between similar products in real retail environments.

Yes, that is a hard technical problem that the vendor did not solve before deployment.

What Bloomberg reported is a different category of AI initiative. Starbucks is using AI-assisted coding to build internal enterprise software that replaces vendor platforms, specifically IBM's maintenance management system and Microsoft's inventory management infrastructure.

Internally developed applications are expected to roll out by the end of 2027, pending testing, according to the Bloomberg report.

More Starbucks:

The AI-assisted coding angle is the structural shift that makes this bet more credible than NomadGo's. Building enterprise software from scratch was historically prohibitive for non-technology companies because of the cost, time, and specialized engineering required.

Now, generative AI coding tools have meaningfully lowered those barriers. Starbucks has been pushing its tech workers to use AI coding tools aggressively, reportedly factoring AI usage into employee bonuses, Bloomberg reports.

The company is also on track to reduce its enterprise technology budget by approximately $30 million in the fiscal year ending in late September, including $10 million in software savings and $13 million from reducing external contractors in favor of internal staff.

$400M software bill, $2B turnaround target, and a Starbucks business that is recovering

Varadarajan told employees that Starbucks spends approximately $400 million annually on software and sees "clear opportunities to reduce the spend," according to Bloomberg. The company is reviewing every technology contract and, in some cases, building its own alternatives for applications that engineers already heavily customize.

Starbucks has also been working for several years on a point-of-sale system designed to replace Oracle's Simphony platform. That's a separate and longer-running internal development effort that predates the current AI initiative.

Related: Starbucks taps childhood nostalgia with 5 new drinks

The technology cost reduction effort is part of a broader $2 billion turnaround program under Niccol, CFO Cathy Smith told CNBC's Kate Rogers. That context is important for you as an investor reading this story.

The AI software initiative is not a vanity technology project. It is a cost reduction lever in a company-wide financial restructuring.

The underlying business is showing genuine recovery. Q2 fiscal 2026 results, reported April 28, showed consolidated net revenues up 9% to $9.5 billion, with global comparable store sales up 6.2%, according to Starbucks' earnings release.

North America comparable store sales grew 7.1%, driven by higher transactions and average ticket. Non-GAAP EPS of $0.50 expanded 22% year over year.

This is the Starbucks our customers deserve.

These were the words Niccol said in the Q2 earnings release.

 In September 2025, Starbucks walked into 11,000 stores with NomadGo's inventory  AI tool that promised 99% accuracy and counted up to eight times faster than a human.
In September 2025, Starbucks walked into 11,000 stores with NomadGo's inventory AI tool that promised 99% accuracy and counted up to eight times faster than a human.

Universal Images Group via Getty Images

My read: the failure history makes this worth watching closely, not dismissing

I want to be direct about what I think is happening here.

The NomadGo failure was instructive, but it should not define how investors evaluate this current initiative. Deploying computer vision in a consumer environment to count physical products in real time is a genuinely hard problem.

Building internal enterprise software using AI coding tools is a different problem, and one that large technology companies have been demonstrating is solvable.

Related: Starbucks has new plan to beat Dutch Bros, 7 Brew

What concerns me is execution discipline. Starbucks has 41,129 stores, Q2 fiscal year 2026 results, and a history of deploying technology initiatives that do not scale cleanly.

The IBM and Microsoft systems being replaced are not trivial: maintenance management and inventory management are operationally critical. Replacing them with internally built software that was developed faster and cheaper using AI coding tools carries real integration risk if the testing process is rushed.

The timeline of "end of next year, pending testing" is the phrase I keep returning to. That qualifier exists for a reason. Starbucks has earned some skepticism here, given what happened in May 2026, and the stock's 27.92% year-to-date gain suggests the market is already giving Niccol meaningful credit for the turnaround.

Related: Starbucks, Dunkin', and Luckin embrace non-coffee products

The Arena Media Brands, LLC THESTREET is a registered trademark of TheStreet, Inc.

This story was originally published July 10, 2026 at 1:07 PM.

Get unlimited digital access
#ReadLocal

Try 1 month for $1

CLAIM OFFER