.jpg?width=300&name=ChatAI-FeaturedImage(Header).jpg)
By now, it’s likely that you’ve heard of ChatGPT.
A curious Google search will land you on the OpenAI main page where a call to action beckons “Try ChatGPT for yourself” — Subtext: it’s not just for geniuses. It’s accessible to you, don’t be scared.
It starts innocently enough if you’re like me – you just wanted it to tell a story. Not a good story, just a small example of what it’s capable of.
“Write a children’s story about a child that meets a wizard & is granted a wish” …. [3/10 stars, would not read again]
Within the first few examples, it was easy to be impressed, then maybe you tried cranking up the weirdness in your prompts:
“Write a 500-word essay about the benefits of scurvy in the style of a pirate”. [7/10 stars, my story ended with the quote, “Yo ho ho and a bottle of vitamin C!”]
It has no problem doing that. In fact, it welcomes the ridiculous ideas you throw at it & immediately starts generating a viable response. As if to say, “This was weird [yawn]. What’s next?”
It does that, too. In fact, it’s GOOD at it. This brings up the underlying theme throughout the ChatGPT experience: It does what you tell it to do.
Maybe that’s obvious that ‘it does what you tell it to,’ but this was important to keep in mind as news about ChatGPT started popping up in peripheral feeds more and more with increasingly alarming titles.
“Schools look to ban ChatGPT; students use it anyway.”
“Japanese universities become latest to restrict the use of ChatGPT”
“ChatGPT banned in Italy over privacy concerns” [this decision was back in March & has since been reversed]
Fear over ChatGPT was escalating quickly. Combined with the notion that AI is one of the only things Elon Musk openly fears, people began asking themselves: should I be worried or excited?
Side note, if Hollywood has taught us anything about this moment, it’s going to be important to get on the sentient AI’s good side. Maybe even apologize to Alexa for all the stupid requests over the years, just for good measure.
Weeks pass.
I wasn’t worried yet. What’s to be worried about? Catfishing? AI-inspired fraud attacks? The internet is already a scary place. I continue to use it in random, fruitless ways. For a moment, I even found myself using it as a substitute for Google’s search engine.
**Generate a recipe for a cake using only these ingredients: [things I have available]** [2/10 – this was gross]
A month passes, still, no robocalypse —- apart from mashups of local news pundits allowing it to write their opening lines & TikTok get-rich-quick schemes, nothing big really happened, society remains un-crumbled, and we all just got used to a world with Chat GPT, big whoop, make it talk like a pirate, who cares?
**What stocks should I buy? [revised: What Kind of stocks should I buy in 2023?] -> [further-revised: What is the best way to make money in 2023?]**
Then companies were starting to incorporate it into more & more tools (this should be expected – the company isn’t called Closed AI). New apps, new ways to leverage the AI, new pirate essays. Momentum was increasing, but I didn’t really care yet. It wasn’t going to get me rich quickly & easily. The recipes it gave me weren’t great.
Lots of people are using it to the point where it’s frequently saying, “It can’t handle the volume of requests.” When that happens, I like to imagine an international digital night club and the bouncer just calmly clicks the line closed and says, “We’re at capacity right now.”
More of a reason not to care – it won’t even LET me ask it my stupid questions.
Then it was going to replace Salesforce admins. THEN IT WAS GOING TO REPLACE SALESFORCE ADMINS!?
I’m a Salesforce admin!
**Make this list of accomplishments sound more important! ***
**Write a cover letter to a hiring manager explaining why I am better than you! **
Suddenly, I care & I care more than ever.
I was already convinced it would replace lots of jobs (how else will the machines rise?), but I never thought MY JOB was going to be taken over by AI…
After all, my job is hard, right? I use as much regular intelligence as I can muster each day & there’s no way an artificial substitute will do, right? You can’t replace Steve with Stevia, right? (bow)
I started fantasizing about how it would happen & felt betrayed by that little Einstein avatar.
I’ve been a huge proponent of Einstein Activity Capture since its release & I’ve even saved the “Einstein Insights” bubbles as my chef’s kiss at the end of an implementation.
We all knew SFDC was venturing into new spaces (see Blockchain) & that ChatGPT could do some impressive things, so the idea that the language-processing capabilities of ChatGPT & the Salesforce-object-oriented connections of Einstein could combine forces into an all-knowing-super-Admin was a little frightening.
I read a few articles and the salesforce release content & marketing materials, all of which were vague enough to leave me with more questions than answers.
So, is it Doomsday for admins or not?
With the utmost confidence (that one can have in this environment of rapid & exponential change) at this moment, I can say to you:
Nope. It is not Doomsday for admins.
Although Chat GPT is great at doing lots of different tasks, Salesforce administration is not yet one of them. It’s a strong crutch to lean on, but it’s not taking the lead anytime soon. How can you feel certain? In the words of the mischievous folks at OpenAI, “Try ChatGPT for yourself!”
I began feeding it some of the questions I was coming across daily to see what kind of support my AI sidekick could offer, and the results were mixed.
- This was a pleasant surprise, although there’s a big gap between “a technically sound idea” and “what will actually work.” The initial response starts with apex code, which is a complicated starting point, but I liked the idea.
- The first response ended with a caveat of “You’ll need to make an apex controller if you want this to work,” so then I asked it my favorite type of question, “Do the thing you’re describing.”
- It does it…it writes the controller.
- Then, in an almost petty fashion, I asked, “Make this easier” – NOT because I need it to be simpler (ahem) but because writing an apex class that tries to relate a contact to a new case automatically, every time, always…seems incredibly problematic.
- What happens if they add a validation rule to the case over time?
- What happens if we have duplicate contact records?
- What if the contact is already related to the Case?
- What if Service Rep A only handles cases with contacts from Accounts ABC, and Service Rep B only deals with Accounts XYZ
- What if…. LIFE happens?
- Then, it gave me an easier method, which combined Process Builder with Flow. This was the direction I hoped it would take, but umm…Process Builder? If you’ve been keeping up with the times, you’ll know that building a Process Builder in today’s Salesforce world is like insulating with asbestos – things have changed, and this solution will soon become obsolete.
Result: The AI spat out a block of code that will potentially solve my salesforce problem
Feedback: This solution is technically sound, it can be implemented with the right resources & will temporarily help make a situation “seem more automated”, but the human element is so absent from this solution (understandably) that it makes me wonder if this would end up causing more issues than it solves. Either way…
Next Step: Hire a salesforce admin.
I tried another prompt:
This one incorporates a common tool (Slack) now owned by Salesforce and has hundreds of ways to send and receive information from your CRM. The solution provided is not one that I would ever suggest for these reasons:
- Slack is a Salesforce App, so they play nicely together – see “Salesforce for Slack” & creating a custom webhook to alert a Slack channel when a record is created is complete overkill.
- Create a Workflow Rule? Remember the asbestos thing? Both Workflow Rules & Process Builders are being deprovisioned by Salesforce soon, so incorporating them in new automation is kind of like installing a tape deck into a Tesla.
Result: The AI spat out a block of code that will potentially solve my salesforce problem
Feedback: This solution will probably work, but WF Rules will be retired eventually & sending an outbound message to an endpoint URL is a legacy approach that will take more time to implement (by who?) & it will not age well. It’s the opposite of a flexible solution, and Slack Channels change, requirements change, and businesses change. Either way, let’s say we go with the robot’s advice!
Next step: Hire a salesforce admin to correct, implement, and manage your new solution.
At this point, maybe you’re thinking: “Look, Steve, all of that seems a little too complicated. I can figure out how to connect the two. There’s a button right here that says ‘Connect!”
: clicks button :
There’s more where this came from, and I realize this whole article may seem a little biased against ChatGPT, so let’s recap: I compared it to robot-overlords, I mentioned robocalypse at one point, and I’ve clearly got a vested interest in Salesforce Admins remaining an important part of the CRM ecosystem. I will own up to all of that!
BUT – try it for yourself. Set up a free sandbox via trailhead, try to copy/paste some of ChatGPT’s solutions into it & see the results. In a world where businesses + AI = Success, it’s going to be increasingly important for PEOPLE to live in the middle of that equation.
For now, it doesn’t matter how intelligent the platform/system/talisman is – it still requires a person to wield it.
Effective Salesforce Administration requires attention to detail, knowledge of an industry, and understanding of the nuances of a business and the culture of its people.
Although ChatGPT is the beginning of something incredible – it only does what you tell it to & it’s not the end of Salesforce Admins.