The trend has been clear for almost a decade now: machine learning is here to stay, and is becoming an ever-more important part of our lives. The world of enterprise software is perhaps the most affected of all. AI and machine learning touches everything, from deep-learning powered analytics to machine-learning backed chatbots for customer service.

In no small part, this is a change that’s driven by the cloud. Cloud technologies make it easier, cheaper, and more accessible than ever to integrate machine learning into SaaS offerings. As a result, a whole new world of intelligent SaaS is opening up: SaaS offerings that include machine learning as a core component. In this article, we’ll look at the cloud changes driving the intelligent SaaS shift, and what you can do to be positioned for the coming shift.

Machine Learning In The Cloud

The biggest driver of the intelligent SaaS revolution is the breadth of offerings from every cloud provider. While some parts of the cloud are almost fully commodities (storage is the same at every provider) machine learning and analytics offerings have become how cloud providers differentiate themselves.

Google pushed into an early lead with their AI Hub, making a name for themselves by providing accessible, powerful ML algorithms that can be easily integrated into any product. Amazon countered with AWS SageMaker, which offers a similar level of sophisticated machine learning. Azure now has an ML hub, and even smaller cloud players such as IBM and Oracle are opening up their internal machine learning tools on their cloud

For all these offerings, there is a simple API that lets you integrate their algorithmic tools into your product offering without doing too much heavy lifting. Instead of requiring a large AI team to design and test the algorithm, with an API all you need is a smaller set of DevOps and Development resources. And because these offerings are designed around a single hub or portal, the results are no longer opaque to management and higher-ups. Instead, you can see a report dashboard that lets you understand how AI is driving your business goals.

On-Demand Compute

It’s not just the development experience that has changed. Behind all this innovation is a huge change in the hardware technology powering the cloud. Google is the true trailblazer here, with their TPU technology. TPUs (Tensor Processing Units) are similar to CPUs or GPUs, but are optimized for machine learning workloads. This specialized hardware makes it both cheaper and faster to develop your own models.

While Google is designing their own hardware, every cloud platform is making it easier and cheaper to build AI tools in-house. Training models used to be time consuming and expensive, requiring scarce hardware to be reserved for long training sessions. Now, providers are filling their data centers with specialized machines, and the price has plummeted. Data privacy tools and integrations with other open-source value-added tools are getting better as well.

All this adds up to a world of intelligent SaaS where building AI models in-house is within grasp for many enterprises that used to rule them out. Just as cloud providers are competing with their AI offerings, so too many SaaS products are starting to compete on the power of their AI, and differentiate themselves with the quality of their predictive models.

The Data Revolution

The last piece of the puzzle is the growing power of the cloud in storing data so it can be used for analytics and machine learning. Offering unlimited on-demand storage was how the cloud got started, but storing formatted data in a useful way had always been a bit of a challenge. Data for AI typically requires storing structured, relational data.

In the past few years, we’ve seen an explosion of petabtye-scale databases like Amazon RedShift and Google BigTable that have stepped up to solve this problem. For just about any data scale within reach of modern organizations, there is a fast, scalable database that can ingest data and deploy it for machine learning applications. Often these databases tie directly into provider AI hubs, and can be integrated with provider tools. And innovation in this space isn’t slowing down, with new tools for data lakes, ultra-high speed data ingestion, and more.

All this goes to show, an enterprise looking to be well positioned for the coming world of intelligent SaaS has a lot of options. Of course, data is the most important element here, so starting a data lake early is imperative. You can use that data as the “ground truth” as you experiment with different models and off-the-shelf offerings from providers. Start to ask what you could to with better analytics, and see what the AI dashboard has to offer.

After you’re dipped your toe in the water, you can assess how things are working with regards to your business goals. You can use your provider’s dashboard to assess performance and understand what tools are offering value and what are not. You can also start to develop in-house expertise and level up the cloud tools you’re using, by digging deeper into provider offerings or spinning up some TPUs and training your own model.