Artificial intelligence tools are increasingly integrated into government operations, offering solutions for tasks ranging from document summarization to fraud detection. Key technologies such as ChatGPT, Claude, Gemini, and Microsoft Copilot are shaping how public sector employees approach productivity and data analysis.
Key Insights
- Text-based AI tools like ChatGPT, Claude, and Gemini assist with drafting, summarizing, and analyzing textual data, making them valuable for tasks such as resume screening and legislative review.
- Image generation AIs such as DALL·E and Midjourney specialize in creating visuals from text prompts, while productivity-focused tools like Microsoft Copilot and Google Workspace AI automate routine office functions.
- Specialized AI models are used across government sectors for functions like fraud detection, policy research, and public service delivery, including chatbots and document translation tools that enhance efficiency and accessibility.
This lesson is a preview from our AI Fundamentals for Government Employees Course. Enroll in a course for detailed lessons, live instructor support, and project-based training.
There are dozens of tools out there when we talk about AI technologies, but here are a couple that you'll likely encounter in your government work. ChatGPT is one that we've looked at before and used as an example. This is by OpenAI, and it's also used by Microsoft in its co-pilot, which we'll talk about in just a moment.
This is a good general-purpose text-based AI. Claude by Anthropic is another widely used LLM. You may have also encountered Gemini by Google.
That's kind of Google's answer to ChatGPT. You may be using that at your agency. And also, of course, Microsoft's co-pilot, which is based on OpenAI's ChatGPT, but it's particularly handy because it's embedded inside its office tools like Word, PowerPoint, Excel, those sorts of things.
So there are many, many more, but here are a few that you probably have access to or have maybe even dabbled with already. These are kind of the big players at the moment. But the field is really, really large.
There are a lot of AI technologies available to us, and new ones are popping up and going away every week, if not every day. So just to give you a sense of the field out there, I've divided AI technologies into these specific categories. There are text-based AI tech, so that's like the ChatGPTs, the Claudes, the Geminis.
And again, those text-based ones are particularly good at summarizing reports, drafting responses, helping you solve problems, analyzing data, anything that has to do with text; they're particularly good at. And some of these models or some of these tools are beginning to edge into some of the other categories, but for the purposes of fundamentals, we'll keep them in the text category for now. There are also some image generator AI technologies.
You may have heard of DALI, Firefly, or Midjourney. This is AI text that has been specifically trained to create visuals. So there's a whole family out there of image-based AI tech.
So if you're not necessarily getting what you want from the text-based one you're using, let's say like ChatGPT, know that there are others out there, and it's worth exploring. Another classification of this technology I'd call productivity. So I immediately think of Microsoft Copilot there, because again, that's built right within your tools.
Google Workspace has begun to introduce AI into some of its different tools. And that's AI, that's, as the name suggests, or as the title suggests, trying to make you more productive, trying to recommend ways to automate some of the more routine office tasks that you may be doing every single day. And then, of course, there's the other category, right? There's the super specialized AI that's being built, because like we saw with our example of a cat detector, right? You can build AI to do anything you want it to do.
So some AI has been built. There are models out there that are specifically looking at fraud detection, or medical imaging, or specific to translation, in ways that some of the other models that are built to do multiple things just maybe aren't as good at, because these are applied to specific missions. We are starting to see use cases of AI in government in all types of positions.
For example, in human resources, a lot of federal and local government HR folks whom I've interacted with are starting to use AI to help them draft job postings or help them summarize resumes. In grants and finance, AI is being used to help detect fraud. Really large data sets that may take humans hours and hours and hours to do may really benefit from having an AI assistant find things that maybe they just missed, or, oh, I didn't notice that, because humans can get fatigued.
It's natural, and it happens. AI is also being used for policy and research to summarize legislation or scan large data sets. This can be really useful for us as government users of AI.
When new PDFs come out, and they're 80, 100 pages, normally that'd be a maybe multiple coffee endeavor to get through that large PDF and to try to understand it. But now with AI assistants, they can help you summarize and scan those larger documents and data sets, and things. Certainly, for public services, we used the example earlier when we were talking about small language models of chatbots.
Imagine all the phone calls and things that have been saved by deploying this AI technology. Not everybody has to reach a human necessarily, because you could just interact with the chatbot and find out when their offices are open. Or document translation may not necessarily be perfect, whereas otherwise, you may be looking at something in a different language and have no idea what it was talking about.
Well, AI can be employed to give you a roughly 100% accurate translation that you don't need to actually wait till the office opens to get assistance with.
