Our latest Salesforce Ben Developer Survey Results have identified some very interesting trends in how developers are making use of AI in 2025. While some things are no shock at all – like the overall increase in AI usage and involvement in AI projects – other metrics seem to tell a conflicting story regarding how Salesforce Developers are working with AI.
Fear not, I’m here to dig into these metrics and make sense of the noise. Join me as we uncover the hidden story behind these insights!
The Metrics
Through our surveying of the Salesforce ecosystem, we discovered two metrics that were quite intriguing. Firstly, we found that a very high number of developers were leveraging AI in their day-to-day operations. In fact, the data showed us that almost 90% of respondents who were Salesforce Developers were using AI in their current role. This number was split 50-50 between using it occasionally and using it daily or regularly.
The survey results also showed some interesting data around the kinds of projects developers were working on. Despite the heavy adoption of the tools developers were using, those very same respondents told us that only around 30% of them were working on AI projects, with another 20% either planning to do so or hoping to in the future.
This is quite the juxtaposition – almost every developer we spoke to was using AI frequently, but just short of one third of them were working on AI-focused projects. This shows us a few interesting things about how AI is being used, prioritized, and where the most value is being realized.
What Are AI Projects?
We need to be careful when we talk about the gap between these two metrics, and we do this by first defining what we mean. Developers using AI in their daily tasks are using tools like Agentforce Vibes, GitHub Copilot, Claude Code, Cursor, and ChatGPT to assist with development. While the developer themself sits in the captain’s chair, they will leverage these tools to perform tasks under their supervision.
This reduces the amount of time developers spend on repetitive tasks that don’t take a lot of thinking and allows them to free up time for more strategic decisions that they’re far more suited for (and, quite frankly, these tasks are more engaging).
AI projects are quite a different beast. While AI tools are already built and allow you to immediately start seeing value and saving times, AI projects are where there is an idea or plan for where a business may be able to leverage AI. The AI functionality does not yet exist to take advantage of, and this is why the developer is involved – to make it happen!
Projects that developers are likely involved in are implementations of various Agentforce features within Salesforce or creating Agent Actions that Agentforce can take advantage of using Apex. It may also be that there are custom AI builds that integrate with Salesforce, or that Salesforce data feeds into an external tool to feed another third-party AI engine.
AI Tools Are Not the Only Use of AI
We’ve all become quite familiar with the different AI tools that we can use in our personal and professional lives. These tools allow us to be more of ourselves in all realms while helping out with some of the more monotonous and mundane tasks that take up our time.
That being said, these are not the only uses of AI when you think about it. I’m not just talking about the aforementioned AI projects, but the AI that we’ve been using for years and taking for granted. Think about features that Salesforce Einstein brought to us back in 2016: Predictive Lead Scoring in Sales Cloud, Recommended Case Classification in Service Cloud, and Recommended Experts, Articles, and Topics in Experience Cloud. Although they’re not one of the more modern generative AI tools that we’ve all been focusing on, they are indeed still forms of artificial intelligence.
ChatGPT started a new wave of excitement around artificial intelligence because it showed us what large data could do in terms of creating new things, rather than just showing us patterns in our own data (i.e. predictive/analytic AI). We saw firsthand the scale to which AI could seemingly create something new for us. As the years have gone by since the release of ChatGPT, we’ve seen enterprises build their own AI tools that leverage LLM technology. Most relevant to us as Salesforce professionals is Agentforce, which uses LLM generative AI to inform actions that should be taken against our business and customer data.
AI has come an extremely long way in a very short period of time, accelerating at a very impressive rate, especially since 2022. With that said, you must keep in mind that artificial intelligence is nothing new. Generative AI is finally becoming mainstream, but this is simply a new category of artificial intelligence. It isn’t the first, and it won’t be the last (I’m looking at you, AGI!).
The Impact on Productivity
The biggest promise that any of these AI tools or projects bring with them is enhanced productivity for all. Whether you are looking at the AI tool usage in daily operations or the AI projects that developers are working on, the ultimate goal of these is to see more output from less human input.
Developers utilizing tools such as Agentforce Vibes, Claude Code, or GitHub Copilot will be able to hand off their repetitive tasks to an LLM to free up their time to focus on the more important and more complex tasks.
Why AI Projects May Be Lagging Behind
There’s been an immense amount of speculation as to why AI adoption in business seems to be weaker than anticipated, but I don’t want to rehash anything I’ve heard before in this article. When you think about it, developers have had a lot more exposure to LLM AI than the average person – let’s look at one of the most common examples of LLM AI that we know of, and that is ChatGPT by OpenAI.
Did you know that ChatGPT was not the first iteration of the tool? In fact, ChatGPT was simply a friendlier UI that was layered on top of a powerful transformer called GPT (Generative Pre-Trained Transformer). GPT is the underlying engine behind ChatGPT, and while the chat interface was introduced in November 2022, the first version of the GPT engine (GPT-1) came out in June 2018. The only thing was that to access it, you needed to know how to call it via API. This is something that only developers can do, as there was no user-friendly interface layered on top for us everyday folk quite yet.
Additionally, there were other tools that were targeted for developers, such as GitHub Copilot, which came out in October 2021 following a short preview. Think of GitHub Copilot as an AI pair-programmer that developers could use to help speed up their work before ChatGPT existed. Developers have had access to tools like ChatGPT long before ChatGPT, Gemini, Grok, or Perplexity existed for the rest of us.
Something else that I wanted to call out, which specifically compares the 90% adoption of AI tools by developers to the 30% figure for AI-focused projects, is this: tangibility.
Developers are leveraging tools that they have seen in action already, are learning about based on examples that already exist, and are seeing their colleagues and others in the industry benefit from. It’s said that there’s nothing quite as powerful as a personal referral!
One of the core reasons that I believe AI projects may have a slower uptake is quite simply because the results are not quite as easily demonstrable as an off-the-shelf product. Every business is nuanced, and every business will have a different set of use cases that they wish to see some of the available AI tools (like Agentforce) applied to before they feel confident in investing heavily into an implementation.
The reality and driving force behind the lack of tangible results is that there is inadvertently a lack of proven business impact, at least at the scale that business decision makers considering a major investment would want to see. Decision makers want to see that their financial investment is going to deliver some revenue increases – otherwise, why would they front the money? Yes, AI is cool, but if the results it delivers are not making an impact on the bottom line, then what’s the point?
I suspect that we will see a larger number of AI projects taken on once we start to see more successful AI projects delivered, with the results speaking louder and the competition having an edge that is unique to the newer technology. Don’t let the 30% figure fool you – the time to master AI is now.
Summary
Although the numbers show an extremely high adoption rate of artificial intelligence by Salesforce Developers, this does not contradict the other discovery about the moderate amount of AI projects they are involved in. Developers are finding value in using AI tools in all sorts of projects, and right now only a third of those projects happen to be AI-focused themselves.
What are your thoughts on the speed of development in the AI space compared to the uptick of AI tooling? Let us know in the comments below!
Want to learn more? Make sure to download our SF Ben Salesforce Developer Survey Results 2025.
