116 | "bro wtf?!?!"
Getting started with ChatGPT at a small nonprofit
I’ve been connected with dozens of nonprofit organizations over the past couple of years, which has been a wonderful opportunity in so many ways—from lots related to technology (my general wheelhouse) to the myriad challenges that come with chairing a new organization’s board. Through this StrefaTECH initiative, I’ve particularly had the opportunity to learn a lot about the possibilities of AI to help the perpetually stretched-thin staff … as well as the many barriers to capitalizing on those possibilities.
Last spring, I began an association with a nonprofit that’s staffed with more than two dozen incredible people. They’re all knowledge workers working remotely, so they spend most of their day in front of a computer. There’s lots of communication needed internally and externally, and as a fairly new, growing organization, there’s also a great need to find and share information. Sounds like a great place for AI to be of huge help!
Alas, they’re all also very, very busy striving to meet the organization’s ambitious goals and to keep up with the ever-changing understanding of what contributes to success. Hitting weekly/monthly/quarterly metrics takes a lot of effort. And because of the entrepreneurial, innovative spirit of leadership, processes are evolving and changing regularly. Sounds like the kind of place where there will never be the right time to try AI.
And indeed, both observations are spot-on … AI could be of huge help, but there never will be a time when investing in upskilling, training, piloting, and planning to bring in a new tech initiative will make sense.
So, we’re starting slow. We signed up for a set of ChatGPT Teams licenses and are in a pilot phase where everybody can “do what they want” with the tool. Some on the team already have used ChatGPT a bit or have joined webinars or training about generative AI. Others have no idea what it can do.
After a few weeks, a handful of staff members had logged into their accounts (though most hadn’t tried … yet!). Of those who’ve used it, there’s also been a mix of experiences. The title of this issue is a verbatim prompt from one staffer: “bro wtf?” It’s the inspiration for a few comments on how you can best approach AI to get the most from it!
1. Have a conversation!
Many people who are new to using AI through an app like ChatGPT approach like “a better Google search”—asking a question that has a defined answer and wanting a great response on the first try. Spoiler alert! This is not the magic that generative AI brings. Yes, sometimes you actually will get a great answer, and yes you can use ChatGPT in place of Google for many things … but the magic comes in the conversation you can have with AI.
Here’s a quick example: I’d like to understand how chaos theory might apply to a program for autistic children, to help inform a grant application I’m working on.
Starting search string (Google) / prompt (ChatGPT): “chaos theory as relates to autistic children”
Google search: links (and short snippets) for various articles, some relevant, some not so much
ChatGPT response: a few paragraphs summarizing research with links to sources
The AI response itself (whether from Google’s AI Overview, ChatGPT, or another AI tool) is often much more helpful in getting to what you want quickly than a traditional Google search. But the magic comes in what you can do next—have a conversation!
Follow-up: “Tell me more about the over-regulation research”
Google search: provided links to a bunch of articles, but none of them had anything to do with chaos theory or autistic children. All Google knew about was the specific request for more about over-regulation research
ChatGPT: gave a few more paragraphs specifically diving deeper into the over-regulation research related to autism mentioned in the first part of the conversation
2. If you’re not getting what you want, say so!
In your conversation with AI, if you’re frustrated that it’s not giving you quite what you want, tell it! You don’t have to worry about hurting its feelings (unlike a colleague, intern, spouse, or otherwise wonderful human being who wants to help you but just isn’t quite “getting it” yet!). And the fastest path to getting what you want and moving on is to be brutally honest with your conversation partner.
For our example, I might prompt next: “I want to know specifically about preschool children with autism. Stick to that topic.”
This will guide the rest of the conversation—the AI tools keep track of what you’ve prompted—so you may need to give updated guidance later in your chat if you no longer want it to stick with your correction!
3. Stuck? Ask it to guide you
A highly under-utilized part of conversing with AI is asking for help. Many experts are great at keeping conversations going by telling you things you didn’t know enough to ask, and AI can do this too!
In our example, one strategy suggested in the response for working with preschool children was “Sensory Tools: Incorporating sensory tools, such as fidget toys or weighted blankets, can provide the necessary sensory input or reduction, helping children maintain a regulated state.” You might follow up with a prompt something like this: '“I like the idea of incorporating sensory tools. What else should I know in order to do this effectively?”
Or even something more general, like:
“What else should I be asking?”
The more you’ve told your AI conversation partner about what you want to know—and what you already know—the better it will be at guiding you in a helpful direction.
4. Alas, you may end up with “bro wtf?”
All of this sounds great, but there’s a truth with generative AI … it sometimes hallucinates, and also it may just not “get it.” This might lead to your own personal “bro wtf?” moment, where you’re wondering if your conversation partner has gone completely off the rails or just hasn’t been paying attention.
As it turns out, you can simply incorporate this reaction into your AI conversation. The ChatGPT exchange that inspired the title of this article was from an analyst at the nonprofit organization who was trying to get ChatGPT to create a compelling graph of data he had exported from their CRM. ChatGPT did OK at grasping the general idea of the graph—a barchart with good judgment for labels and general structure. But what it presented was visually rather, well, unappealing. After a number of prompts trying to get ChatGPT to regenerate the chart with colors and more vibrancy, but getting variations on the same drab chart, our intrepid analyst was clearly exasperated when he typed that simple prompt: “bro wtf?”
Did ChatGPT take offense? Stomp off in a huff? Turn the energy around and make accusations about the analyst’s times of being clueless himself? No! It responded as it’s been trained—to be supportive and always helpful! I just typed that into my chat with the example above, and the response was:
Not the response you were expecting? Let me know what’s up, and I’ll adjust!
Dang … if only I was consistently so nice when my (conversation) partner raises an eyebrow in a very human version of that short phrase!
The Bottom Line: Use It!
There is so much you can do better and faster with the help of AI … and the more that you use it, having back-and-forth conversations as I’ve suggested above, the better the results you’ll get. You’ll develop instincts and habits that will short-circuit the annoyances more quickly and will help you and your conversation partner hone in on the wisdom you seek.
And as always, as you have exchanges with AI, remember that whatever you do with the results is your responsibility. So if you decide to share info you’ve gotten from your chat buddy, think about what you need to independently verify and…
Make Good Choices