On the Azure platform, Microsoft offers 31 services under the umbrella of ‘AI and machine learning’. There’s a lot of promotion of AI generally and machine learning specifically as meeting a variety of business needs but it’s often not clear what service should be used and what the sensible use-cases are. This adds another layer of complexity to managing your cloud costs.
Azure’s AI and machine learning services can be divided into a few categories:
- Platform services that are offered as a utility (pre-built machine learning models via API, for example)
- Platform services that offer the environment for building custom solutions (for example, Data Science Virtual Machines)
- A mixture of pre-built and custom capabilities (templates and ML models that can be tailored to your needs)
You can download a spreadsheet listing all the available services (as of this writing) here.
Many of these services are highly specialized or created for providers who are offering their own AI or machine learning service to customers and use Azure as the compute and storage engine. Imagine, for example, a company – let’s call it BillSight – whose business model is providing the medical billing industry with a form processing service. Billsight can build their backend on Azure using Azure Forms Recognizer; For Billsight, there’s a clear connection between the revenue stream and their effective use of a cloud-based text analysis platform.
For most organizations however, it’s not often clear which, if any, of the many AI and machine learning services are worthwhile. The combination of marketing hype and technical staff eagerness to try new technology can lead to expensive deployments of solutions that aren’t fit for an organization’s business requirements and are either wildly inappropriate or too much for the use-case. There’s also a bit of confusion about the difference between situations needing machine learning and those that benefit from good old fashioned business intelligence and analytics.
Consider this decision model:
This shows a simplified view of some of the key factors you should have in mind when deciding if you’ll use any of the Azure AI/ML services and if so, what’s appropriate. The key idea is that you need a doctrine to guide you.
In the next post, we’ll talk about pricing and cost efficiency.
Azure FinOps 