Published on the 08/11/2023 | Written by Heather Wright

Build Code or Code Builder…

SAP and Salesforce have turned their focus on generative AI to aid developer efficiency, with both vendors adding the technology to coding assistants in the past week.

Generative AI’s code writing abilities were an early focus of the ChatGPT hype.

“Admins can get started easily with an out-of-the box experience, and experienced developers can build faster with a full-featured IDE.”

Early reports showed potential big benefits for software developers using an AI coding assistant. One study found that software developers using Microsoft’s GitHub Copilot completed tasks 56 percent faster than those not using the tool, with a June McKinsey report saying AI-based tools were delivering impressive speed gains for many common developer tasks, including writing new code and code refactoring, or optimising existing code. 

A September Indeed report suggested software development was the most like job to become augmented by the technology, 

Now, SAP says it has integrated GenAI capabilities into its low-code tools to deliver a turn-key environment for coding, testing, integrations and application lifecycle management. SAP Build Code will use SAP’s new AI copilot, Joule, to embed code generation capabilities for data model, application logic and test script creation.

The company has been taking a slow and steady approach to generative AI. Earlier this year it rolled out Joule, a chatbot which responds to natural language requests, providing answers based on your business data and helping employees find information or complete tasks.

Joule will be embedded in SAP applications from HR to finance, supply chain, procurement and customer experience as well as into the SAP Business Technology Platform (BTP) to provide insights across SAP solutions and third party sources, but only SuccessFactors, the BTP regulatory change manager and the SAP Start homepage for cloud apps are expected to get Joule this year. The public cloud edition of S/4HANA is expected to follow next year. 

Joule will be the coding copilot in Build Code solutions, which are built on SAP BTP and optimised for Java and JavaScript development. 

SAP says Build Code will streamline cooperation with business experts using the low-code SAP Build solutions, launched at SAP’s TechEd conference in 2022. 

Salesforce, meanwhile, announced general availability of its web-based integrated development environment, Code Builder, to help development teams customise CRM and code faster using GenAI to generate, test, debug and refactor code. The offering is pre-configured and optimised with Salesforce languages and frameworks.

The company’s MuleSoft Anypoint Code Builder launched into general availability last month.

Salesforce says the Salesforce Code Builder is used to customise an organisation’s CRM system. MuleSoft Anypoint Code Builder, now generally available for desktop, is used by MuleSoft developers to build APIs and integrations between third-party systems and Salesforce. Both are built on the same infrastructure, but pre-loaded with different tooling. Other extensions are also available for both offerings. 

Dan Fernandez, Salesforce Developer Services vice president of product management, says Code Builder is enhancing developer productivity while democratising access to a consistent tooling experience so any trailblazer can build on the platform.

“Admins can get started easily with an out-of-the box experience, complete with Salesforce tooling, and experienced developers can build faster with a full-featured IDE customised with their favourite extensions from the Open VSX marketplace, all from their browser,” Fernandez says.

The latest releases come as IT departments report struggling to meet business demands, and skills shortages continue to bite. Salesforce’s State of IT report found nearly two-thirds of IT departments are having trouble meeting business demands. 

GenAI upgrades to popular tools such as GitHub’s AI pair programmer Copilot, have been touted for their benefits. Early this year, the popular offering was credited with an average of 46 percent of developers code in files where it was enabled, with developers completing tasks 55 percent faster with Copilot. 

But they’ve also raised concerns. Samsung found out the hard way when proprietary code uploaded to ChatGPT was leaked. It quickly banned employee use of the tools. 

Apple, too, reportedly restricted use of Copilot over concerns about data leakage.

Copyright issues and the regurgitation of bad code has also been a concern, along with the risk that bad code could be feed into models by bad actors looking to introduce vulnerabilities that they can then exploit. In a survey of 500 US developers, 61 percent admitted to using untested code generated by ChatGPT. More than a quarter of them said they did so regularly. 

That highlights a key factor with generative AI – the ongoing need for human involvement.

 The Indeed research noted that the human element required in many jobs, including software development, remains irreplaceable. 

“A GenAI software developer can write a line of code, but can’t place it within the broader, complex software architecture envisioned by a human developer,” Indeed says. “There are certain work skills unique to humans that simply can’t be automated.”