Software as a Service, or SaaS, is a concept that is familiar to many. Long-time Photoshop users will recall when Adobe stopped selling its product and instead shifted to a subscriber model. Netflix and Disney+ are essentially Movies as a Service, particularly at a time when ownership of physical media is losing ground to media streaming. Artificial Intelligence as a Service (AIaaS) has been growing in market adoption in recent years, but the uninitiated might be asking: what exactly is it?

What is AIaaS?

In a nutshell, AIaaS is what happens when a company develops and licenses use of an AI to another company, most often to solve a very specific problem. For example, Bill owns a company that sells hotdogs through his e-commerce site. While Bill offers a free returns policy for dissatisfied customers, he lacks the time to provide decent customer support, and rarely replies to emails. Separately, a software developer has created a chatbot that can handle most customer inquiries using natural language processing, and often solve the issue or answer a question before human intervention is even required. For a monthly fee, the chatbot is licensed to the hotdog vendor, and implemented on his website. Now, the bot is solving 80% of customer issues, leaving Bill with the time to respond to the remaining 20%. But Bill is still too preoccupied making hotdogs, so he subscribes to a service like Flowrite, that uses AI to intelligently write his emails on the fly.

AI is also being put in service to analyze large sets of data and make predictions, streamline information storage, or even detect fraudulent activity. Amazon’s personal recommendation engine, an AI powered by machine learning, is now available as a licensed service to other retailers, video stream platforms, and even the finance industry. Google’s suite of AI services range from natural language processing, handwriting recognition, to real-time captioning and translation. IBM’s groundbreaking AI, Watson, is now being deployed to fight financial crimes, target advertisements based on real-time weather analysis, and analyze data to help hospitals make treatment judgements.

Benefits of AIaaS

• Price: Purchasing a sophisticated piece of software can be expensive, while subscriptions can defray costs over a long period of time — making products more accessible to small businesses. Furthermore, many companies lack the resources to develop software in house and may find it cheaper to license existing software.
• Up to Date: Services mean you always have access to the latest version, as opposed to making an annual purchase of the newest iteration of expensive software. Since many AI platforms are in a state of constant development and tweaking, staying up to date is a big benefit.
• Infrastructure: AIaaS vendors can handle the data center and processing needs, offloading those responsibilities from the customer.
• Supports Development: A steady stream of paying customers enables the service provider to put more funding toward development, adding new features, eliminating bugs, and continuing to improve the product.
• Bigger is Sometimes Better: The three AIaaS giants — Google, Amazon, and IBM — command the clout to hire the best developers in the field, often resulting in the best solutions (though not always).

Challenges of AIaaS

• Closed Source: Fans of the transparency that comes with open source software will often be disappointed that AIaaS products are largely closed source.
• You Don’t Own It: Companies have been known to unilaterally pull the plug on their services, leaving their clients scrambling to find alternatives. And a client may find that a product isn’t perfectly serving their needs, wanting instead to take development in a different direction. Ownership has its advantages.
• Price: Over the course of time, a service will have been paid for many times over, and it would have been better to pay a lump sum up front. In many cases, purchasing isn’t even an option – a clear indicator that service providers find better profitability in a rental agreement.
• Not Always the Best Solution: Surprisingly, quite a few people don’t mind engaging with chatbots. Unsurprisingly, most people prefer dealing with their own kind. Depending on a business’ clientele, chatbots may be off-putting.
• Has its Failures: Companies rushing to embrace bleeding-edge technology are often left disappointed. Google’s medical AI made headlines in 2020 for failing to translate success in a lab to success in the real world, causing analyses of retinal scans to take more time rather than less. IBM’s Watson has had similar blunders, also from hospital implementations.

What’s Next for AIaaS?

Machine learning AIs improve with time, usage, and development. Some, like YouTube’s recommendation engine, have become so sophisticated that it sometimes feels like we have entire television stations tailored perfectly to our interests. Others, like language model AI GPT-3, produce entire volumes of text that are nearly indistinguishable from an authentic human source.

Microsoft has even put GPT-3 to use to translate conversational language into a working computer code, potentially opening up a new frontier in how software can be written in the future, and giving coding novices a fighting chance. Microsoft has also partnered with NVIDIA to create a new natural language generation model, three times as powerful as GPT-3. Improvements in language recognition and generation have obvious carryover benefits for the future development of chatbots, home assistants, and document generation as well.

Industrial giant Siemens has announced they are integrating Google’s AIaaS solutions to streamline and analyze data, and predict, for instance, the rate of wear-and-tear of machinery on their factory floor. This could reduce maintenance costs, improve the scheduling of routine inspections, and prevent unexpected equipment failures.

AIaaS is a rapidly growing field, and there will be many more niches discovered that it can fill for years to come.