The world of technology development is chaotic, and AI seems to be the most complicated part. Gartner predicts that by 2026, 75% of organizations that focus on trust-focused AI will be successful. That the other 40% will fail. Anyone who follows AI news in 2022 knows how high the percentage of AI project failures is. According to news sources and analysis, AI projects have a failure rate of 60-80%.
Just such a percentage seems worrisome. This is even more painful when compared to the failure rate of IT projects as a whole. fails to deliver on time, exceeds budget, or fails to meet user expectations Why is the percentage of failures of AI projects so high? So why is it different from the percentage of the IT landscape? What exactly is the cause of this catastrophic failure?
Earlier, we discussed the business philosophy of a successful company to highlight the benefits of using AI, which, of course, has to be useful. But applying the ministry to a typical organization can be challenging. In this article, we will present API-led and other technical perspectives for AI applications, such as how to use API-led to identify AI to help reduce the failure rate of these projects.
Why Integration Strategy for AI?
Before diving into AI and API-led, if you’re new to how API-led works, it’s a good idea to study API-led Connectivity first.
What do integration and AI have in common? It must first be said that the implementation of AI is an integration problem, and the high failure rate is due to the failure of integration around AI itself. Therefore, a quality integration architecture is required to make a project. AI Can Succeed
Comparing AI and Integration to Plumbing Integration’s main function is to maintain the flow of information, while AI relies on it to provide insights for the organization. But without a good way to convey information from one AI to another, the implementation of AI is impossible. Integration architectures can predict whether an AI project will succeed or fail.
AI still hits a snag must solve problems that have already been solved in Integration Space as well. The problem is that this Monolithic architecture contains large clusters of information. Whereas AI is usually designed to solve only one problem. As a result, the system requires many connections just for one use.
Within the Integration Space, this causes computational issues. and delay development time Like AI systems, monolithic models take too much time to develop. and cannot be reused in other projects in the future It also lacks efficiency in calculating the amount of data.
What is API-led AI?
API-led Artificial Intelligence is a simple, standardized way of looking at AI. The goal is to make small, reusable blocks. to be applied throughout the organization To readers who are already knowledgeable about AI, this might sound impossible. Or is it a bad advice? How do you break an AI model into smaller pieces? Some models may not be able to break apart. But some systems can do this.
Let’s look at an example from our fast food restaurant recommendation system. Let’s say you want to introduce a menu to a customer while he or she is ordering. And you want to apply this information to the drive-thru system. dining system and ordering via mobile phone The old way was to create three systems to use separately at a time. The drive-thru build team had to test and deploy the system. When finished, start doing the same thing again for other systems. The customer information received from the drive-thru system may have some duplicates. But some information in the drive-thru system is not available in the dine-in system or order via mobile phone at all.
So what does API-led look like? First, it looks at features related to all three systems, such as what’s currently on offer. time of day and location of the restaurant With that information, we will create an AI that can solve all 3 systems at the same time. Although it can’t work by itself. But we can increase the results from this system in every Times used with additional models that are smaller than before. The model can be reused on the other two systems without the need for any adjustments, greatly shortening the time to market for setting up other related systems.
Build a smart organization in a better way
API-led AI allows you to reap the full benefits of API-led for AI. Through iteration and abstraction, API-led provides a future-critical foundation to accelerate system development. API-led is proven to be the solution. that is effective against the problems of artificial intelligence
Learn more about the benefits of API-led in MuleSoft’s catalog ofIntegration case studies or contact us
