Everybody seems to be talking about ChatGPT. Over a million people signed up in the first 5 days of launch, and there is no shortage of LinkedIn posts talking about its awesome power and the potential for the future. The company behind it is OpenAI, which has investment from Microsoft and there are rumors that it will take a 49% stake for a jaw-dropping $10 billion. They probably see that this is a way to leap from the hold that Google has on the search market. Google has had AI technology for several years but has been unwilling to launch as it is concerned about the accuracy of the results which would impact its reputation. As impressive as the results that ChatGPT delivers, they do need to be tempered with some skepticism. IIt can be compared with a consultant; the “advice” that ChatGPT gives is eloquent, and convincing, but not necessarily 100% accurate.
The CEO of OpenAI responded to the initial concerns about the accuracy of the results with this tweet, where he is clearly trying to manage expectations.
What is ChatGPT
I asked ChatGPT – “What is ChatGPT” and here was the answer:
ChatGPT is a pre-trained language model developed by OpenAI. It is built on the GPT-3 architecture and is fine-tuned for conversational and language generation tasks. It can be fine-tuned on specific tasks or used as a general-purpose language model to generate human-like text. It is designed to understand and respond to natural language input, and can be used in a variety of applications such as chatbots, language translation, and text summarization.
The future – coming soon
I think this is going to be a massive trend, and very large businesses will get built with this as the interface, and more generally [I think] that these very powerful models will be one of the genuine new technological platforms, which we haven’t really had since mobile. And there’s always an explosion of new companies right after, so that’ll be cool.
Hype or a glimpse of the future (for Salesforce professionals)
What is clear is that ChatGPT cannot replace experience or skills. You cannot ask for something you don’t understand, because you cannot phrase the question to get the best result. But more importantly, you cannot validate that the result is correct. Users have also found that ChatGPT ‘lies’ about a wide range of issues, from making up historical and biographical data, to justifying false and dangerous claims.
But it can provide a huge increase in productivity – currently – in tightly scoped areas. It is a very fast, accurate assistant when tasks are narrowly scoped and accurately delegated. So this means that certain Salesforce roles will benefit more than others. Let’s look at each in turn.
Experienced Salesforce Admins should not fear ChatGPT as it will make them more effective and productive. If they learn to phrase the prompts accurately it will create a Validation Rule, Formula, Apex Class, Lightning Web Component (LWC), or a Unit Test for a LWC in seconds. However, it cannot build a flow or any declarative result as the ChatGPT results are plain text.
BTW Do not be tempted to use ChatGPT to answer Trailblazer questions or help you pass a certification. If you cannot answer the questions yourself, you cannot check if ChatGPT is making stuff up.
As ChatGPT has been taught by looking at millions of lines of code, it produces pretty good code. Again, the more accurate the prompt, the better the answer. So even if the code may not be 100% correct, it is a productivity enhancer for an experienced developer. At a minimum, it can create the core classes which is boring, non-value-added work. It’s possible that those at greatest risk are junior developers. Junior developers need to understand how to use ChatGPT to help improve their coding skills.
ChatGPT is never in doubt, but not always right. So it is very poor at giving advice or forward thinking. If you are a consultant, ChatCPT could write a proposal for you by setting up the core content. But you will need to take it from there. It is an assistant that doesn’t have contextual thinking, is not able to think outside the box, or does not pick up on social cues. So be very wary of learning on it too much. For example, ask it about the process for merging orgs. It will give you an answer that, at first glance, appears very credible. But we discovered that it is not accurate enough. ChatGPT doesn’t understand the nuances. Simply regurgitating a ChatGPT answer is a quick route to losing the trust of any client.
Salesforce Business Analyst
If you are a business analysts ChatGPT could be used to suggest process steps that can be used to sense-check any work you have done. So, it is useful to prepare before a mapping workshop or validate that you haven’t missed anything. It is great at formating textual documentation. Maybe even writing training content.
The architect certification is so hard to achieve because it is experience applied in context. Of all the roles, the architect is probably the least likely to be able to leverage ChatGPT, but I am sure that there are narrow opportunities.
Salesforce Prompt Engineer (new job role?)
Maybe this is a role in the future. Possibly, this is a skill we all need. A Prompt Engineer is someone expert at asking ChatPGT questions – a “prompt”. Changing the odd word or the order of words in the prompt makes the results dramatically better. We can all work together to share and improve the ‘prompts’ that we put into ChatGPT, which will in turn improve the results we receive. To kick-start the collaboration we created a site called http://SalesforcePrompts.com, where you can can share prompts and results. You can save their prompts privately (so it’s a great place to store them), but we’d prefer you make them public so we can all learn together.
ChatGPT will not be replacing Salesforce professionals anytime soon. But it will make experts 10x more effective and productive because they can exploit ChatGPT to deliver routine work in a fraction of the time. However, a skill that needs to be mastered is asking the right questions. Very much like being a Business Analyst.