No More Junior Salesforce Developers? How AI Will Impact the Job Market

We’ve all heard it a thousand times – AI is reshaping the workplace. 

There is a breadth of opinion on how exactly this will happen, and whether the effects will be net positive or net negative. But one topic that has stirred up a considerable amount of discussion is how the upcoming wave of “digital labor” will change the job market for entry-level people. 

A “Quiet Erosion”?

Senior Industry Analyst at SalesforceDevops.net, Vernon Keenan, wrote in December that virtual employees (VEs) threaten to “quietly erode the entry-level and mid-tier positions that have traditionally served as foundational on-ramps to career advancement and professional status”.

He warns that the workforce is at risk of being “hollowed out” by the rise of VE economics, with fewer people having opportunities to organically build skills and climb the career ladder, amid a commoditization of cognitive tasks. 

Salesforce Ben has previously written about comments from Marc Benioff and Mark Zuckerberg, who both discussed how AI tools are affecting the job markets at Salesforce and Meta respectively. 

Tech CEOs tend to talk up their own AI products, of course, while experienced developers are just as quick to point out that AI tools, while helpful, are still very much in need of experienced human supervision to, frankly, filter out the garbage they sometimes produce. 

Vernon Keenan outlines how, in his view, this “quiet erosion” of entry-level positions will take place amid a rise of VEs. 

Entry-level jobs going away will lead to a decline in foundational roles, a loss of mentorship opportunities, and barriers to skill development, he says. 

Then, shifting workplace demographics will lead to a polarization of job categories, resulting in a reduction in economic mobility. 

It’s a bleak picture for someone looking to get their career started, who may have to compete not just with other inexperienced and ambitious workers – and with AI programs that can (in theory, at least) do everything they can, quicker, cheaper, and without annual leave. 

A Bleak Picture… But is it Accurate? 

A quick search on careers.salesforce.com on February 18, using the filters “Software Engineering” and “New Grads”, revealed only one position.

A search simply for “New Grads” with nothing else specified brings up 16 roles.

Interestingly, and perhaps unsurprisingly for some in the ecosystem who have been paying attention, 14 of these are for the Sales team. The one role left over is in the Operations team. It is for an “Associate Analyst, Sales Strategy”. 

Marc Benioff has said that Salesforce would not be hiring any further software engineers in 2025 amid an AI productivity boost, but that the cloud giant would be looking to expand its sales team in a bid to push Agentforce. 

This certainly seems to be reflected on the company’s job listings page. 

Neither the Sales nor the Software Engineering teams are taking on any interns, according to careers.salesforce.com.

“It’s a Long Way Off” 

Robert Baillie, a Senior Salesforce Developer, says he is not entirely convinced by the concept of an entirely autonomous VE at this stage. 

“I’m sure that’s something that’s coming, but I think it’s a long way off,” he told Salesforce Ben. 

When asked whether VEs pose a threat to entry-level and mid-tier employees – particularly Salesforce Developers – Robert said, “Absolutely”, adding that, in his opinion, there are two paths that organizations are likely to take. 

The first is senior developer support. Robert said: “Rather than have seniors define work for juniors, you instead focus seniors on using AI to generate code, creating a gap where juniors would find their first and second jobs and have mentors and reviewers helping them improve their skills.”

From a management perspective, the senior’s role does not change significantly, other than they appear to have a much-shortened feedback loop from definition to implementation, he added. 

From the senior’s perspective, the role does change – instead of occasionally reviewing code, the role becomes almost exclusively that.

“The feedback loop is so short that as soon as you ask an agent to perform a task, it has effectively done it and the output needs to be considered,” Robert said. 

The second path is junior developer enablement.

“Instead of hollowing out at the junior level, AI can be seen as a replacement for the senior developer level and be used to guide the junior, rather than support the senior,” Robert said.

“This would be an even more worrying trend, but unsurprising since AI can very much tempt management tiers into thinking that outputting large amounts of code is the same as having very productive developers. 

“By making juniors appear more productive, and by presenting very quick turnarounds – from simple ideas to working standalone prototypes – it’s understandable that people might imagine that this can be translated into quickly changing production quality systems. 

“Prototyping ideas and production development on existing codebases are very different things and require very different approaches.”

He added that, in both scenarios, it is clear that there’s a risk to employees. 

“It’s rare that an organization sees an increase in productivity and doesn’t also see an opportunity to cut costs – especially in those organizations where software development is not the primary purpose of the business,” Robert said. 

“At No Point Have You Required The Developer To Understand Anything Other Than How To Ask The AI”

We asked Robert for his thoughts on whether the proliferation of VEs will affect skill development if they are being relied upon to do more work. 

He said: “Historically, when you first started to code you started with a blank screen and you needed to understand the fundamentals of the language before you got going – with AI code generation, this simply isn’t true anymore. It’s easy for you to generate code that does something very quickly, and believes that you understand what it does without actually knowing very much at all.”

This will be seductive at first, giving devs immense power to build something that appears to work, Robert said. 

“You may even get to the point where you have fully working apps that perform significant tasks that work well for a small number of transactions and small numbers of users – but at no point have you required the developer to understand anything other than how to ask the AI to produce the output they need,” he added. 

Once this happens, two forces will begin to compete with each other, Robert said. 

The first is the apparent ability to output functionality easily which pushes devs to output even more. 

The second is the rapidly growing codebase and embedded knowledge that you might expect the developer to have gathered, but which “is simply not there”. 

“That gap just gets wider and wider until the software fails, and the developer’s skills haven’t kept up with the code that’s been produced,” Robert said. 

Final Thoughts 

AI tools have proven to be extremely useful for developers, handling tedious and mundane tasks, with the seasoned dev simply having to look over the work once it’s ready. 

But tech industry leaders and junior developers alike may come to regret relying on these tools, which can certainly increase output – a tempting offer for both – but deprive the up-and-coming dev of much-needed experience and understanding.