AI is transforming not only the methods of software development but also the logic behind corporate software procurement.
According to an internal presentation obtained by Bloomberg on July 9, Starbucks is accelerating its push for in-house software development, aiming to leverage AI to enhance development efficiency and gradually replace some long-term commercial software purchases. The targeted systems include Microsoft's inventory tracking system, IBM's equipment maintenance management tools, and the Oracle Simphony point-of-sale (POS) system, which has been under transition for years.
If these projects proceed as planned, some of the replacement software could complete testing and be deployed as early as the end of next year. This signifies that the changes driven by AI are evolving from merely boosting development efficiency to prompting large enterprises to reassess their software procurement models, thereby introducing new competitive pressures for traditional enterprise software vendors.
Since the beginning of this year, the stock performance of both Microsoft and IBM has trailed the S&P 500 index, as market concerns persist over whether AI will erode the moats of traditional software providers.
Annual $400 Million Software Spend Becomes a Key Cost-Cutting Target
The most immediate driver for Starbucks's in-house software initiative is cost reduction.
Earlier this year, the company's Chief Technology Officer, Anand Varadarajan, stated at an internal forum that the firm spends approximately $4 billion annually on software, noting "there is a clear opportunity for optimization here." He explicitly directed employees to aim for reduced reliance on external software.
Software costs are just one component of Starbucks's broader $2 billion cost-cutting plan. According to the internal presentation, the enterprise technology department is projected to save around $30 million in the fiscal year ending this September. Approximately $10 million of these savings are expected to come from reduced software procurement spending, with another $13 million stemming from decreased use of external consulting and professional services, as related work is shifted to internal teams.
Concurrently, Starbucks is adjusting its technology team structure, expanding technology centers in Nashville, USA, and India, while maintaining its research and development team at its Seattle headquarters. Since February of last year, the company has eliminated approximately 2,300 positions, with the technology department also affected.
AI Programming Lowers the Barrier to In-House Development
AI has become a significant catalyst for Starbucks's software self-development efforts.
Internal materials indicate that in developing the platform intended to replace IBM's equipment maintenance system, AI-assisted programming has taken on core development tasks, substantially increasing efficiency. Starbucks has been consistently promoting the internal use of generative AI in recent years. A previous Bloomberg report noted the company has even incorporated employee usage of AI tools into performance bonus evaluations to accelerate AI adoption within development teams.
An official company blog post earlier this year also stated that AI and other digital technologies will be crucial supports for long-term growth, helping baristas reduce back-office tasks and dedicate more time to customer service.
Beyond developing new systems, Starbucks is conducting a comprehensive review of its existing IT contracts. Internal documents show the company is reassessing "every contract, every service," with in-house development increasingly becoming the preferred option for software that originally required significant customization.
In-House Development Faces Persistent Challenges
However, developing software internally is not without its costs.
While self-development can reduce licensing fees, long-term maintenance, continuous upgrades, and the expansion of technical teams often translate to higher personnel and operational expenses. This is a classic trade-off long faced by large corporations. A recent attempt by Starbucks reflects this reality. An AI-driven inventory management system previously deployed by the company has been taken offline, with operations reverting to manual counts to address accuracy issues encountered in practical use.
Meanwhile, Starbucks continues to use third-party software products from various vendors, including Microsoft, and has not completely severed ties with external suppliers in the short term. More importantly, significant divergence remains in the market regarding the extent to which AI can autonomously handle complex enterprise software development.
For the broader software industry, Starbucks's experiment may be just the beginning. As AI continues to lower the cost of software development, more large enterprises with ample technical teams may start seriously reevaluating a question seldom debated in the past: whether to continue buying software or to build it themselves.
Comments