AIDigest/2026/07/26/2026-07-26-06-starbucks-ai-software-cost-cutting

Starbucks builds AI-generated internal software to replace vendor systems

Source: Forbes — 2026-07-12

Summary

Starbucks is using AI-assisted coding to build in-house replacements for two enterprise systems it currently licenses: a Microsoft Dynamics 365-based tool for inventory tracking and IBM's TRIRIGA platform for maintenance management. The effort, first reported by Bloomberg on 2026-07-09 from an internal presentation, sits inside a broader $2 billion cost-cutting push aimed at Starbucks' roughly $400 million annual software budget. New EVP and CTO Anand Varadarajan, who joined in January 2026 after nearly 19 years at Amazon, is driving the initiative under CEO Brian Niccol, with some replacements targeted to go live by the end of 2027 pending testing.

Key Takeaways

Two named systems are on the chopping block: Microsoft Dynamics 365 (inventory tracking) and IBM TRIRIGA (maintenance management) — both to be replaced by AI-built internal software.
The move is funded and framed by a $2 billion cost-reduction plan targeting Starbucks' ~$400 million/year software spend, not a standalone IT experiment.
Timeline: some AI-generated replacements could go live by end of 2027, contingent on testing outcomes.
Leadership: new CTO Anand Varadarajan (ex-Amazon, joined January 2026) is running the effort under CEO Brian Niccol, and the strategy stresses fixing underlying business processes before layering on AI — a lesson from Starbucks' earlier AI missteps.
Market reaction was immediate: IBM shares fell roughly 3%, ServiceNow about 3.5%, and Salesforce around 4% on the report, signaling investor concern that AI-assisted coding lowers the barrier for large enterprises to build over buy.

Reel Script

Hook (~18s, ~40 words): Starbucks just told Microsoft and IBM they might not need them anymore. Not a pilot, not a slide deck — a real plan to rip out two paid enterprise systems and replace them with software its own engineers built using AI. IBM stock dropped on the news.

Core Concept (~70s, ~160 words): For decades, the build-versus-buy math favored buying. Licensing something like an inventory system from Microsoft or a maintenance platform from IBM was cheaper than paying a team of engineers to build and maintain it yourself. AI-assisted coding changes that math. Think of it like the difference between hiring a contractor to build custom cabinets versus buying flat-pack furniture — flat-pack was cheaper because custom labor was expensive. Now imagine a robot that can build custom cabinets almost as fast as flat-pack ships. Suddenly custom looks a lot more attractive, especially if the flat-pack furniture doesn't quite fit your kitchen. That's what's happening inside large enterprises right now: AI coding tools are shrinking the labor cost of building internal software, so companies like Starbucks are asking whether it's still worth paying vendor licensing fees for tools they could now build in-house and tailor exactly to how they operate.

Hands-On (~75s, ~190 words): Here's specifically what's on the table. Starbucks currently runs its inventory tracking on a Microsoft Dynamics 365-based system — that's the software that tells a store manager how much oat milk or cold brew concentrate is sitting in the back room. It also runs maintenance management on IBM's TRIRIGA platform, which schedules repairs on espresso machines and walk-in coolers across thousands of locations. Both are getting AI-built in-house replacements, with some targeted to go live by the end of 2027 if testing holds up. This isn't happening in a vacuum — it's one piece of a $2 billion company-wide cost-cutting plan, and software licensing is a specific target: Starbucks spends around $400 million a year on software, and every dollar it can move from a vendor's invoice to its own engineering budget is a dollar it controls. Notably, the plan leads with fixing the underlying business process first, then applying AI — Starbucks has stumbled with AI rollouts before, and this time leadership under new CTO Anand Varadarajan is being explicit about that order of operations.

Takeaway (~25s, ~55 words): This is a warning shot for every enterprise software vendor charging recurring fees for functionality AI can now approximate in-house. Watch which companies follow Starbucks' lead over the next year — that'll tell you how real the build-over-buy shift actually is. Follow for more signal, less hype.

Discussion

(No questions yet — ask follow-ups via a Claude Code chat session on this repo; answers get appended here.)