The CEO of Salesforce appeared in an IG reel posted by Bloomberg, where he says that Salesforce has seen productivity increases of 30-50% in key engineering functions, and that AI is doing 30-50% of the work. He also said that he won’t hire any more coders, which implies that productivity gains from AI will allow Salesforce to keep up with growth with the same headcount.
I’m somewhat skeptical of the second claim, but I can see how the first claim could be plausible.
I’m generally skeptical when CEOs of public companies extoll the virtues of Generative AI. I think there’s an incentive structure for CEOs to overstate the utility of generative AI. Publicly traded companies are under pressure from Wall Street to be more productive, and I imagine much of that is based on perception. I imagine Wall Street analysts, who are generally not software engineers, may view Generative AI as a transformative technology, and may reward CEOs who promote it with more favorable valuations. CEO compensation is often tied to stock valuation, so there’s an additional incentive to promote it.
Benioff doesn’t explain how they measure productivity, but I imagine it’s something like 30-50% more pull requests, more completed support tickets. I think this is pretty plausible – I’ve used Generative AI in some of my workflows and I can appreciate how it’s sped up my work. I could see it being particularly useful for things like support, which involves trawling through sometimes dense documentation to find an answer.
But I’m skeptical that Salesforce will no longer hire coders. I wrote my undergrad thesis on automation, and one thing I remember reading is that productivity enhancements as a result of automation (of which AI is just the latest round) actually tends to increase demand for professions that become more productive.
For example, there are more bank tellers now than there are before the introduction of the ATM. ATMs reduced the cost of opening new branches, so banks opened many more branches.
Similarly, I could imagine that AI productivity enhancements actually increases, rather than decreases the demand for new coders. Let’s assume that it takes 5 coders to support 1 custom integration for a large Salesforce client. That’s very expensive. But let’s say AI lets you support the same integration with 2 coders. That makes it cheaper for Salesforce to pursue more custom integrations, which means they may pursue more of them the same way banks opened more branches.
If you can still charge clients the same amount of money for the integration, then if anything it supports the case for hiring more coders, since the revenue per coder is now much higher.
Granted, I don’t understand software productivity super well and I am completely unsure if software industry staffing ratios work like this. But I don’t think it’s at all certain that Salesforce will forego hiring more coders, especially after Generative AI (and its limitations) becomes better understood.