Restaurants Start Developing Their Own Software: Will AI Change SaaS Costs?

The technology budget for restaurant businesses has been growing steadily in recent years.

Buying separate subscriptions for POS, scheduling, training, inventory, checklists, maintenance management, CRM, loyalty, delivery, and reporting has become a common model for multi-location businesses.

AI-powered coding tools are beginning to change this equation.

The 'SaaSpocalypse' discussion highlighted in Nation’s Restaurant News’ Restaurant Daily on August 24 revolves around a simple question:

If a restaurant can produce its own software much cheaper than before, is there a need for a separate SaaS subscription for each operational function?

Eight-store chain developed some SaaS tools in-house

One notable example of this discussion is Keva Juice.

After requesting price reductions from the software companies it used, Gary Thomas, the CEO of the eight-store juice chain, decided to replace some subscriptions with systems he developed himself.

Keva developed its own operational platform called UseFroot using Anthropic’s Claude tool.

The system includes functions such as:

onboarding and training,
daily task lists,
inventory management and forecasting,
employee communication.

According to the Keva CEO, the company aims to save approximately $30,000 in subscription costs annually. This figure is not based on the company's own accounting or an independent financial audit.

The scale of the company is more important than the figure itself.

In the past, developing such software in-house could require a software engineer, product manager, and a continuous IT budget.

AI-assisted coding is making it easier for smaller businesses to experiment with simple operational tools.

The 'build vs. buy' calculation is changing

The classic decision in restaurant technology was:

Will you build the software yourself or buy it ready-made?

For most restaurants, the answer has been ready-made SaaS.

The reason was clear.

The software company managed:

hosting,
security,
updates,
support,
backups,
integration.

While the restaurant only paid a subscription.

When AI coding tools lower the cost of application development, the calculation changes.

Especially for very simple workflow tools, businesses may start asking themselves more frequently:

“Is this task I’m paying hundreds or thousands of dollars a month for really that complex?”

Applications such as checklists, internal training portals, simple maintenance requests, or in-store task tracking could be the first candidates.

The same question is being asked at Starbucks scale

This trend is not limited to small chains.

According to an internal presentation reviewed by Bloomberg, Starbucks is developing in-house alternatives for some functions of the inventory tracking system provided by Microsoft and the maintenance management tool from IBM.

According to information provided to employees by Starbucks CTO Anand Varadarajan, the company’s annual software spending is around $400 million, and management sees areas where this budget can be reduced.

Starbucks’ goal is not to eliminate all third-party software.

The company continues to use a number of external technologies, including Microsoft infrastructure. The discussion is more about whether it is economically viable to develop in-house some applications that manage highly customized or relatively standard workflows.

This distinction is important.

It is still too early to say "SaaS will end."

However, the pricing power of software providers may change.

Restaurants are re-evaluating their software budgets

The technology overload in the restaurant sector post-pandemic has been a long-standing problem.

NRN previously reported that businesses rapidly added various technology tools during 2020, resulting in fragmented systems and a technology burden.

In a chain, there can be separate vendors for:

POS,
online ordering,
delivery aggregation,
loyalty,
labor scheduling,
training,
food safety,
inventory,
maintenance.

While each system may seem reasonably priced on its own, when multiplied by the number of stores, the total technology expense grows rapidly.

This is compounded by integration fees and the cost of employees learning different systems.

The trend of developing one's own applications with AI is also driven by this "subscription fatigue."

However, developing your own software is not free

The biggest misconception here is thinking that when code is produced with AI, the software cost disappears.

Creating the first version of an application is just the beginning of the work.

In real restaurant operations, the software must:

handle thousands of transactions,
manage authorizations,
keep data secure,
perform backups,
communicate with other systems,
continue functioning after errors.

Additionally, when an employee says, "the application has an error," it must be determined who will resolve the issue.

When purchasing SaaS, the restaurant delegates a significant portion of this maintenance burden to the vendor.

In its own software, however, the same burden falls back on the business.

Starbucks’ inventory experiment is an important warning

This risk is not theoretical.

In May 2026, Starbucks phased out another AI-based automatic inventory counting tool used in its North American stores.

According to a report by Reuters, the system could misidentify dairy products and similar items or fail to count some products altogether. Starbucks later reverted to the standard manual counting approach in stores.

The tool in question was not one of the new systems developed from scratch by Starbucks; it used NomadGo technology.

However, the incident provides an important lesson:

Using AI does not automatically make operational software reliable.

A 95% accuracy rate for an inventory system may technically look good.

In a restaurant, however, an incorrect milk stock estimate could mean a customer is unable to order the product they see in the app.

Even a small error in operational software can be significant.

It is necessary to separate critical systems from simple workflow tools

It is not correct to evaluate every restaurant software in the same risk category.

For example, if an employee checklist application fails, the team can use a paper list for a few hours.

If the POS crashes, sales can stop.

Similarly, the loyalty system may experience temporary issues; however, an error in the payment system or food safety monitoring tool can have more serious consequences.

Therefore, in-house development with AI is more sensible initially in low-risk areas.

Businesses can roughly categorize their technology portfolio into three groups:

Strategic and critical systems: Areas requiring high reliability, such as POS, payment, and cybersecurity.

Operational systems: Tools with strong integration needs, such as inventory, scheduling, and maintenance.

Simple workflow tools: Checklists, training materials, internal communication, forms, and task tracking.

It is more likely that rapid development with AI will first become widespread in the third group.

Vendor lock-in may become a bargaining issue

This transformation is also significant for SaaS providers.

If a restaurant believes it can develop a simple version of the software it purchases in-house, it gains alternative power in contract negotiations.

Particularly, the following topics become more important:

who owns the data,
in what format the data can be exported,
whether integration APIs are open,
exit conditions from long contracts,
pricing per store.

A technology system from which a business cannot easily extract its own data can create not only software but also an expensive dependency.

In the age of AI, the capacity to separate from the vendor is becoming part of the technology strategy as much as "vendor selection."

The advantages of ready-made restaurant platforms do not disappear

On the other hand, technology providers specialized in the restaurant sector have significant advantages.

For example, MOD Pizza has had a technology relationship with PAR Technology for over ten years. According to PAR’s announcement in April 2026, MOD uses an integrated technology stack including POS, payment, operations, and Punchh loyalty, and the loyalty program surpassed 5 million members in 2025. This data is based on PAR’s company statement.

The advantage here is integration.

A restaurant can easily create its own checklist application.

However, getting the POS, loyalty, payments, and delivery systems to work simultaneously is a much more challenging engineering problem.

Therefore, rather than completely eliminating the SaaS world, it is more likely that low-value and easily replicable software layers will come under pressure.

What does this mean for restaurants in Turkey?

Multi-location restaurants and coffee businesses in Turkey are also increasingly using software subscriptions.

Therefore, the first step should not be to say, "let's write everything ourselves."

A more accurate starting point is to take stock of the technology inventory.

The business can ask the following questions:

How many separate SaaS subscriptions do we have?
Which tools are actually used every day?
Are there two systems doing the same job?
What is the total software cost per store?
If we change vendors, can we extract our data?
Which simple workflow can we develop in-house with low risk?

After this analysis, AI can be a real cost-saving tool.

Otherwise, the old "SaaS chaos" could turn into dozens of small in-house applications this time.

Stafflink’s Opinion

In our view, the upcoming change in restaurant technology is not that "SaaS will end," but rather businesses will now ask more rigorously for each subscription, 'why am I buying this?'

AI coding tools are shifting bargaining power to the restaurant side, especially in checklist, training, internal communication, and simple operational applications.

However, generalizing the initial cost advantage gained by small chains like Keva to the entire technology stack is risky.

Systems like POS, payment, loyalty, inventory, or maintenance require years of maintenance after the first line of code is written. Cybersecurity, integration, and employee support are also part of the cost.

Starbucks’ recent closure of another AI inventory tool due to accuracy issues serves as a strong reminder of this.

Therefore, the healthiest model may be not to “build everything,” but to “own what differentiates you, buy what must never fail.”

The new question in the technology strategy of restaurant operations is no longer just "which software should we buy?" but rather:

“Which software is truly worth buying?”