your123 - stock.adobe.com

Tip

No- and low-code AI's role in the enterprise

Low-code tools enable noncoders to build and deploy AI applications. The growing list of these tools ensures that AI development is no longer confined to experts.

Kashyap Kompella
By
Published: 09 Mar 2023

No-code and low-code AI tools are increasing in popularity among enterprises. Contrary to popular belief, they are not meant to replace expert developers and programmers, but instead empower business users to create useful applications more easily. When used well, such tools make enterprises more nimble and responsive to changing customer needs.

No- and low-code software development platforms use pre-made visual interfaces, drag-and-drop components and pre-built modules so users can create applications without writing code. Both no- and low-code development enable business users with little or no programming experience to build apps, automate processes and create custom solutions. Low-code platforms also support users who write code for custom or complex functionality and can integrate with existing systems and data. Examples of low-code tools with such capabilities include Appian, Google AppSheet and Looker, Mendix, Microsoft Power Apps and Salesforce Lightning.

No- and low-code AI tools aim to reduce the barriers of expertise required to build enterprise AI applications. Citizen developers -- often everyday workers, such as project managers, business analysts and marketers -- use such tools. There are only about 25 million software programmers in the world, but about 1 billion knowledge workers. No- and low-code tools can significantly broaden software development capabilities by involving the workforce without programming skills.

Enterprises in need of AI-based applications -- but without an AI development workforce -- should know these tools' current breadth of uses, as well as related developments, like automated machine learning (AutoML) and AI coding assistants.

Use cases for no-code AI and AutoML tools

Different types of tools target specific use cases: website builders, mobile app builders, chatbot builders, workflow automation, business analytics and database management, for example. In each use case, low- and no-code tools reduce development time and effort, decrease business user dependency on IT teams and reduce bottlenecks.

While there is a shortage of software programmers in general, it is even more severe in the case of data scientists and AI specialists, particularly those with expertise in the end-to-end MLOps lifecycle. There are products to help data scientists and AI engineers build and deploy ML models. These AutoML tools automate end-to-end ML workflows to the greatest extent possible, encompassing data pre-processing, feature engineering, model evaluation and selection, hyperparameter tuning, deployment and monitoring.

Typically, enterprise AI teams include experts in hyperspecific areas, such as modeling, data engineering or integrations and deployment. Since AI model development is iterative in nature, no- and low-code tools suit this type of quick, iterative development. Overall, they help increase individual and team productivity for AI applications with a faster time to market.

AI coding assistants as intelligent low-code tools

AI coding assistants are also emerging and are related to no- and low-code. GitHub Copilot, Amazon CodeWhisperer, Tabnine and OpenAI Codex are examples of AI tools that act as coding assistants for software developers building general software. These tools provide code suggestions in integrated development environment tools and even autocomplete code. The AI-generated code may still need to be debugged, tested and manually tweaked to meet requirements. AI coding assistants don't provide the same level of abstraction and automation as no-code tools. They can help achieve some of the same goals, such as reduced development time and increased developer productivity.

Lastly, underneath the hype of generative AI tool ChatGPT is the ability to generate code for specific tasks designated by text inputs, called prompts. Although ChatGPT can generate seemingly accurate but incorrect code, a workaround is to feed the incorrect code back to ChatGPT, which gives a prompt to debug it. Despite limitations, these interactive conversational and dialogue features enable pair programming of sorts, where a programmer works in tandem with an AI coding assistant. As evidenced here, intelligent low-code tools offer a fresh approach to empower both nontechnical users and expert programmers alike.

A variety of no- and low-code tools more efficiently bring to life different software apps that a modern enterprise needs, but these tools still require oversight and expertise from technical professionals. In the future, generative AI tools, like ChatGPT, may reach the same goals but use different approaches.

Next Steps

The rise of the non-developer

Is ChatGPT better for coding than low-code platforms?

Dig Deeper on AI technologies