Cloud Developer Services provided by the cloud provider you use as your infrastructure provider is certainly one way to incorporate AI into your application. It does usually require a significant investment in developer and data engineering expertise.

Another approach, if the AI you want to incorporate is standard functionality, is using the AI APIs from your cloud provider. However, if you want more specialized functionality than voice to text, text to voice, language translation, image recognition and video recognition then this approach may not work for you. Although AWS does have an AI forecasting API it may not meet your needs and other business functions aren't usually provided by the major cloud providers. Plus if you truly want to add value in your SaaS application, especially if it is a vertical SaaS solution, you probably want to solve some difficult and unique problems for your particular vertical market.

Low Code AI solutions provide another alternative.There are some low code solutions from the cloud providers, AWS Sagemaker Canvas as an example, but they still require significant technical expertise for developing, deploying and tuning machine learning models. In addition Amazon does not mention ISVs using AWS Sagemaker Canvas. Although I have not looked at it closely Google's AppSheet looks straightforward but Google doesn't mention it as a tool for ISVs either.

Alternatively you can look at some of the major low code providers who also have AI offerings. Although this article was written by an open-source vendor, Planet Crust, who provides a low code open source platform, their insights into the tradeoffs of using a low code solution as a SaaS ISV is helpful. They have listed 8 low code platforms suitable for ISVs. Some names you may recognize in their list are Salesforce Lightning and Microsoft Power Apps.

One low code vendor providing AI low code solutions is Levity, based in Germany. I've found their explanations of the technology and provider landscape to be very useful. Even though they don't focus specifically on SaaS ISVs, these ideas of how to add AI to your application are useful.

I believe the above gives you some good ideas of how to implement AI functionality in your SaaS solutjon.