Machines and artificial intelligence (AI) have been part of popular discussion since Samuel Butler authored Erewhon (1872), where his satirical utopian society explored the morality of conscious machines as a natural evolution of the Industrial Revolution. One-hundred and fifty years later, OpenAI released GPT-4, introducing the world to AI that can respond like a human (and do so much more!), causing us to revisit many of the same ideas.
While GPT has had several iterations (hence the -4), the latest advances have made a tremendous impact on our world. From the classroom to government, AI is front of mind in conversations across all domains. As a baseline, generative pre-trained transformers (GPT) are a family of neural network models that use transformer architecture to allow bots to produce text and image content and engage in conversations much like a human would. In its simplest sense, GPT-4, more than any other AI tool prior, has the ability to convincingly mimic human language and conversational patterns, marking a major leap forward in its potential use cases.
Need to catch up on AI before we begin? Check out Be an AI-Minded Salesforce Admin.
And you can check out my presentation on prompt writing below.
For Salesforce professionals, ChatGPT and its peers offer a variety of advantages regardless of role. There are a number of use cases by discipline where GPT technology can offer assistance.
The above only scratches the surface of what Salesforce professionals can accomplish with AI, and these deep capabilities have to be evaluated through the lens of company policy, risks, and responsible use. There are a number of concerns and considerations when using AI as part of your job, including local rules, ethics, security, privacy, and erroneous information.
It may be too simple of a take, but I’ll go for it… Before using GPT or other AI technologies to assist with your job functions, be sure to inquire about acceptable use and policies at your company. That is to say, if you’re using these tools currently without checking local policies, stop and seek guidance.
While the power and efficiency of AI tools is certainly tempting, it’s important to operate within the local governance, risk, and control frameworks of your organization. Absent a formal policy, this presents an opportunity for you to be a thought leader, helping to define accepted use cases and frame the conversation around AI in your workplace. Where policies exist, you’ll have a formal rulebook to follow to limit your personal and company risk when using AI.
Here’s a framework for shaping a policy for responsible use of generative AI at your organization.
I asked ChatGPT, “If you anonymize and protect data, why would you recommend a company set limitations and establish security measures for ChatGPT usage?” Its answer was pretty remarkable, especially the first bullet point.
It can be tempting to take a utilitarian view of ethics regarding AI usage—here’s an example. An admin is tasked with creating end user documentation for managing certain tasks in Salesforce; typically, this may be an all-day task and they have 8 hours planned for its completion. Using AI, the admin completes the task in 2 hours, so they step out for the rest of the day and submit the completed work in the morning as expected. No harm, no foul—the outcome was achieved within the time allowed, right? Looked at through the lens of freelancing, let’s say a consultant uses AI to complete a task budgeted for 8 hours of work at $200/hr. Even though they complete the task in half the time budgeted, the consultant logs 8 billable hours against the project. Still no harm, no foul?
In both scenarios, the admin and consultant are being deceptive with customers and colleagues. Viewed through ends-based ethical frameworks (utilitarianism), this may be justified, but it’s absolutely a violation of more deontological ethical thought (duty based) that is centered on what people do, not the outcome of their actions. In both cases, they were paid to create the content, and even if the requirements were met, claiming additional effort that was completed by AI is unethical and was not the expectation.
This isn’t meant to devolve into a philosophical debate (queue the trolley problem!) but rather to stress that we should approach AI usage in the workplace with the highest concern for ethical practice, with an emphasis on honesty, fairness, and transparency to encourage adoption.
Say anything with conviction and authority and an uninformed audience will generally believe your claims. Remember: Conversational AI is built to respond with human language patterns and convincing authority, not necessarily correctly. In my conversations with conversational AI, I’ve found many instances where it provided incorrect answers, both factually incorrect and those that violate Salesforce best practices. Here are a few examples.
Conversational AI is not a substitute for our knowledge as Salesforce professionals. It’s a tool to assist us, not replace us. It’s critical to continue in the pursuit of deep platform knowledge to be able to safely and appropriately use AI as part of our solutions and configurations.
I appreciate your indulgence in entertaining the previous section. I’m enamored with the potential AI has to transform how we work, but I believe we must ensure AI adoption is done responsibly, honestly, and in a manner that encourages its adoption vs. making companies enforce strict measures against it.
With that, I’ll descend the several flights of steps from my soapbox and get to the fun stuff: writing prompts that can help you in your day-to-day role in the Salesforce world.
When writing prompts, here are a few tips to help get detailed and accurate responses.
Below are examples of prompts and conversations I’ve had with ChatGPT, grouped by their purpose: gaining knowledge, solving problems, and helping with design.
After a dozen years on the platform, I still run into products, features, and concepts where my knowledge is not as deep, I don’t have my own words to accurately explain them, or I’m looking to learn new information about something I know well. I consider these knowledge-building prompts, as they don’t solve a particular challenge or assist with a specific design, but instead provide a quick assist in finding information to shore up my Salesforce foundation. Some examples are:
ChatGPT does a remarkable job at providing a starting point for troubleshooting in Salesforce. Given a specific error message, formula, or request around how to configure a specific requirement, it provides quick insights and suggestions to help start your effort. It’s also tremendously helpful in translating the intent of existing formulas, validation rules, and Apex classes when encountering others’ work that may not be well documented. Here are a few examples.
The prompts above are just a few scenarios I’ve used, largely with great results. But, as noted above, responses may not always follow acknowledged best practices, so generative AI should be considered a job aide vs. a tool to replace understanding and effort on our behalf.
I’ve found generative AI to be very helpful in the solution design process to quickly put into words a solution to a specific requirement. As with the first example below, while I know the answer heading in, it allows me to quickly put the structure and words into the tech spec for a specific requirement, once I’ve validated the provided answer follows best practices and meets all aspects of the requirement. For others, it provides a starting point for deciding an approach to sharing an automation design that I can use to inform my final solution. The examples below demonstrate how architects can quickly express existing knowledge, defend design decisions, and outline approaches to make our work more efficient.
AI technology is bringing a swath of new opportunities, and the conversation is just beginning. As the Salesforce ecosystem expands its adoption of AI, we’ll continue to help Trailblazers explore and understand it. To dig in deeper, check out the resources below.
While its impact is widespread, one of the most exciting aspects of artificial intelligence (AI) is its ability to create conversational interactions that generate personalized experiences, supercharging productivity and efficiency. In this blog post, we’ll explore how the implementation of large language models on mobile devices is reshaping the enterprise mobile landscape and how Salesforce […]
Understanding the fundamentals of artificial intelligence (AI) is critical to being an #AwesomeAdmin. This technology has opened up broadly and both people and businesses are finding new and inventive ways to weave it into products and our lives. There’s no better time to get up to speed on AI and how you will be able […]
Einstein Bots interact with your customers quickly and accurately with automation and artificial intelligence (AI) capability. In terms of AI initiatives, the biggest challenge is measuring business impact. One reason most AI projects fail is that people are looking at “model performance” instead of business value, such as how much money, in either additional revenue […]
