Skip to main content

As Microsoft is investing in OpenAI, and after several weeks of full-scale beta testing online, it seemed important to us to take a step back on the usage of ChatGPT. 

Yes we know, you are telling yourself: “Oh no, another article about chatGPT and its Salesforce applications, it’s the 3rd this week!”. Don’t worry, we just want you to take a breather through all this flow of information. And for that matter, this is what we are going to talk about, information. 

Before starting this article, we asked chatGPT, pell-mell, to write a formula with REGEX to validate a phone number, to write a training plan for the standard Sales Cloud features, and even write this article. But we promise, the one you are reading, was written by us. Here is an excerpt from what chatGPT wrote :  

Not bad huh? We also went further by asking it to adjust its answers, add a certain nuance, respect a certain length, etc., and it does it quite well! Here for our example, we have even asked it to correct its angle: 

We let you try it! There are many various applications. 

By the way, this article does not claim to cover all the issues related to ChatGPT. 

And be careful, we are not trying to sell it to you. We do not own any stock at OpenAI. 

ChatGPT does have its flaws. Its designers readily concede this. 

Today, we are not trying to challenge the role of AI in the workplace. This question is of course paramount, but it is very broad and could not be reduced to an article. 

We are also not looking to challenge ChatGPT or AI more generally in the field of creativity. This question is just as important as the previous one, but here we prefer to assume that despite all the efforts of AI developers, it cannot replace the human ability to grasp the nuances and complexities, which is the breeding ground for creativity. 

Ok, once we have said all this, doesn’t the rise of a potential working version of ChatGPT also question our relationship to information? After all, isn’t it a great search engine above all? 

We are overwhelmed with information today. It’s even cliché to say it, but it’s always good to remember it. 

The production of classic emails and documents may have slowed down (has it?), but the flow of information is still there. Email management, spreadsheet and other word processing softwares have been supplanted, at least in part, by messaging (Slack, Teams, Workplace from Meta) and online document management tools (Notion, OneNote, Monday, AirTable , etc.), and more specifically, by tools such as Salesforce. All the more sophisticated as they are often capable of automation. 

To these elements are now added social networks and applications, not to mention the more sophisticated format of video, which is taking up more and more space. We have even reached the stage of the “creator economy”! 

Everything has followed a trend of massification. 

When it comes to using this “raw” information to create content, several questions then arise: 

  • What makes good information? 
  • Are we using the tools we already have at our disposal correctly? 
  • Do we have the right methodologies and processes? 
  • Who do we talk to? How to personalize the content, bring my share of creativity to it? 

Asking the right questions is important. That’s not for nothing that we say that what matters with ChatGPT is to ask the right questions, and to be able to refine it (or at least give the right instructions).  

Take the example of writing a training plan for a standard tool like Sales Cloud. 

Of course, training plans are probably already available online. But let’s assume that we want to produce original content for our business: 

  • Has my experience on the platform led me in the past to properly identify Sales Cloud’s main components? Have I been able to formalize that? 
  • Have I, for example, saved the web pages that explain this or that part well, so that I understand what to prioritize? Am I able to find these pages? 
  • Has a colleague (or myself) not already carried out such training, and could communicate to me the information they have, even if I know that I will have to adjust it? It seems obvious, but would you be able to find such a document quickly?  
  • Perhaps I need to follow a training module myself to learn how to train and how to properly prepare for training? 
  • What are the specific needs of my client? Was I able to collect them? Was I able to formalize them? 
  • What added value do I want to bring? Is the objective clear? Shouldn’t this added value be collective? How to consider different sensitivities? 
  • Who is it destined to? 

Eventually, talking about the use of ChatGPT is talking about the ability of each organization to manage its knowledge. And in that regard, the designers of ChatGPT didn’t invent anything, it’s called Knowledge Management (and by the way Salesforce has a very interesting module which allows to write and distribute articles to help Customer Service, it’s Lightning Knowledge). 

Our example above is potentially very generic: but if you need to use internal information relevant to your project, obviously confidential, and present only in your databases, you will surely wonder if ChatGPT is reliable enough a tool, that you want to entrust with this data. 

Besides, whether you want to use an “internal ChatGPT” tool or not, questioning yourself about your relationship to information is essential: if such a tool exists, it’s because it can rely on data which is relatively structured and clear instructions. Your “internal ChatGPT” could not work with unstructured, scattered, unformalized data. Why not tackle this right now? 

Note that Salesforce already offers AI features through its Einstein product family  (Machine Learning, Natural Language Processing). And if Salesforce can offer it, it’s because Einstein relies on a structured database, your Salesforce org. 

Whether we adhere to ChatGPT or not, it seems essential to ask ourselves the question of the structuring of our organizational knowledge. 

Article written by Jean-Pierre Rizzi, project manager certified at Cloudity.