Much of the world only knows AI from what they see in movies. For decades, Hollywood has portrayed artificial intelligence (AI) in contradictory ways, simultaneously depicting it in a positive and negative light: AI is either going to save humanity, or it will be the biggest threat we’ve ever faced. While these portrayals are often stories of fantasy, our future is destined to be heavily impacted by artificial intelligence.
AI is already part of daily life for some today. When you open your phone with face ID, open your spam-free email inbox and use automatic responses, find a ride-sharing trip to the office, or when you pay at your favorite barista during a work-from-home coffee break. AI is hard at work behind the scenes.
The business opportunity for artificial intelligence
While there are instances of AI today, the opportunity is in front of us. It can be used to do things in smarter, faster, or in completely new ways. There are three types of AI capabilities commonly grouped together:
Process automation: Augment or automate manual processes or tasks
- Example: Transfer employee onboarding information from paper forms or email attachments into a HR system.
- Business value: Process efficiencies and time saved that can be allocated to other high value tasks, reducing employee workload while providing a faster, smoother experience for the new hire.
Analytics: Identify hidden patterns and trends in large datasets
- Example: Detecting wear of parts from sensor data in airplanes.
- Business value: Optimizing maintenance and replacement processes based on actual ware and tear, reducing non-necessary maintenance efforts and costs.
Engagement: Create automated interactions with employees, customers, or suppliers
- Example: Provide employees with 24/7 IT service management via chatbots.
- Business value: Service staff can prioritize their time-solving complex, high-value issues rather than answering frequently asked questions while reducing time to resolution for the employee, having their issues resolved much quicker.
The role of data in AI
If these are some of the AI use cases and value associated, where do APIs fit in? Organizations hire data scientists and developers to develop the AI or machine learning algorithms. Developers will go through a number of steps as they develop the algorithms or they leverage and compose publicly available AI capabilities into their solutions.
Developers will go through an iterative design and development process. This starts with data collection and cleansing, then analyzing, model development, and testing. Data is used to both develop the AI capabilities as well as run it to gain the desired insights or process execution once in operation.
Therefore, tapping into the potential of AI starts with accessing and consuming data. If you consider that AI has its own hierarchy of needs, similar to Maslow’s pyramid – data access and consumption sits as the foundation of what artificial intelligence needs to function. You need access to data across the organization, systems, and devices to get started with any AI initiative. When data is accessible, you can consume it to train and run AI.
The role of APIs in AI
The role of APIs is focused on two areas: unlocking internal data for data scientists to develop, train, and productize AI models; and speeding up access to external services and accessing best of breed capabilities from leading AI providers.
Unlocking internal data for data scientists to develop, train, and productize AI models
Whether an organization wants to automate a back-office process, predict machine part wear and tear, or deploy a chatbot to support its IT service desk, an AI solution will be composed using a mix of capabilities. Regardless of whether these capabilities are developed in-house or composed of external services, real data is needed to train the AI. For instance, an effective chatbot needs to read historic chat transcripts to analyze employee inquiries and understand how to solve them.
Speeding up access to external services and accessing best of breed capabilities from leading AI providers
Many AI capabilities are available “as a service”. These services come from established technology companies as well as start-ups that target specific industries or use cases. Developers can often access and consume these AI capabilities through the vendor’s APIs.
In the example of chatbots as a self-service channel, rather than building their own solution, developers can tap into one of the many natural language processing services that are accessible through APIs. If that chatbot also needs to understand images that the user uploads, the developer might leverage an image processing API to bring in that capability.
Securely discover, access, and use data
AI teams realize that the key to their and their organization’s success is unlocking the data wherever it exists (on-premises, in the cloud, legacy infrastructure) in a way that helps them deliver innovative AI solutions while maintaining governance, visibility, and data security.
Traditional approaches to data connectivity in the enterprise can’t keep up. Often, legacy infrastructure is a barrier, and data is siloed. System landscapes are highly fragmented with data living in hundreds of different systems and apps. Add to that regulatory requirements. Access to and quality of data is one of the three biggest barriers to AI adoption.
MuleSoft’s vision on connectivity of systems, data, and devices is based on composability. Composability means to package data and assets across the enterprise in modern, secure APIs and publishing them in a marketplace for others to consume. These composable building blocks are easy to find and use. By enforcing security and compliance standards every step of the way, anyone across the organization, including AI developers, can self-serve and consume those approved assets to deliver their own projects.
If various components are wrapped in APIs that can be easily discovered, understood, consumed, and secured, they enable different teams across the organization to access data and digital capabilities in a way never before possible. This also gives IT the tools to manage and secure them at scale.
This approach fundamentally changes the definition of integration. It reimagines integration from a system-to-system approach to productizing data for consumption for specific processes and audiences. We call this API-led connectivity, and it allows AI developers to quickly and easily access data and build out new composable data assets while adhering to data governance and without being held back by inaccessible data.
