Salesforce's new Code Builder programming tool shows both the upsides and downsides of using AI to write software code. The potential benefits are clear: having an ‘AI assistant’ generate basic code by translating plain English into programs could really boost programmer productivity. It would save tons of time on repetitive coding tasks.

But there are also risks if programmers start to rely too much on the AI black box to do the work. Without fully understanding the AI's logic, the code it writes may have quality or security issues that humans miss. And developers could lose skills by getting lazy and just letting the AI handle everything.

The best path forward is to use Code Builder as an aid, not a replacement for human programmers. Developers need to review the AI-generated code, check it for flaws, and tweak it when needed.

The key is integrating the AI thoughtfully to combine the best of human expertise and machine efficiency. Code Builder's speed on routine coding allows developers to focus their conceptual skills on harder problems.

Used strategically, Code Builder can make programmers better at their jobs, not obsolete. But only if they understand its limits and don't become over-reliant. AI should empower people, not replace them.

Salesforce is smart to roll out Code Builder gradually while collecting user feedback. This careful approach will reveal the best ways to apply the AI capabilities - and avoid the pitfalls.

With disciplined use, tools like Code Builder can help humans achieve more than they could alone. But if used carelessly, AI risks deskilling rather than empowering. Striking the right balance will allow innovators like Salesforce to boost potential while minimizing peril.