Crown Logo

Technology Support For San Diego Since 1996

Click here for 60 minute consultation858-483-8770

ChatGPT for Teams is a cost-effective way to boost your team’s documentation and policy skills.

As AI continues to grow as a software marketing term, some products really can take your team’s effectiveness to the next level. Copilot and ChatGPT are probably the two most popular AI platforms in use, while Google’s Gemini rebranding has a lot of buzz surrounding it as well. If you haven’t taken the plunge yet, there are some very potent business uses that you might want to consider, especially with OpenAI’s ChatGPT for Teams. Today, we’ll update our take on using ChatGPT to accomplish a broader range of business tasks.

Multiple GPTs

One of the most important features that you get with the current paid subscriptions to ChatGPT is the use of multiple GPTs. ChatGPT is a product that, by itself, can help answer research questions, create snippets of code, etc. With ChatGPT for Teams, you get the ability to make multiple GPTs that are more focused and pliable than the general algorithm.

How to multiply your GPT effectiveness

The idea is that you can give a more focused skillset to each of these GPTs, telling them how to talk to you or which information to seek out when searching the web. You can see a few of mine here, which gives you an idea of the things that I’ve played around with. You can also see that I have four more, mostly related to my role in the company, and even expert bots that can help me with internal datasets that have been uploaded for the GPT to refer to.

Again—and I must stress this, this blog cannot be written by a robot—but certain research functions can be conducted far faster by a Large Language Model (LLM) than by searching on Google. My “History Hunter” GPT helps me fact-check events in computing and software history, mostly as a confirmation that I correctly recall things that I already lived through. While Google’s Gemini is basically taking over the search results (shadowing Bing’s takeover by Copilot), the frontpage of a search engine is also full of content that is promoted and paid-for, not content that best suits my needs.

The other bots serve different functions and have different knowledge examples uploaded to them for reference. EOSbot, here, is all about business methodology and can read and report on a few specific, relevant books on the topic. When I have a question about roles, processes, and goalsetting in our organization, I don’t need to go back to the books—I can simply ask EOSbot.

Strength in Numbers

A cool feature here is that you can get synergy from your GPTs. If you’re chatting with any of your GPTs on a particular topic, using the “@” symbol calls another one of your bots into the chat. You can ask different bots to respond to one another, or work with each other’s answers in a way that draws on their particular skills and directives.

If this seems a little complex, it’s worth noting that the expert GPTs are made very simple to make. When you begin making one, GPT Builder (itself a GPT-expert bot) asks a few questions about what you need your GPT to do, how to respond, and if it should fill in the gaps or ask for clarifications. These questions are important to think about; if you need a GPT to be as factual as possible, you need to explicitly tell it to not fill in the gaps when answering.

I had a moment when building a specific GPT with GPT Builder where I realized that I needed help staying current on the research going on with effectively asking questions to AI chatbots. I thought: “wouldn’t it be nice if I had a GPT that could report on current best practices for prompt engineering?” It only took a few minutes for me to realize that I don’t need it to report on the current best practices… I could just have it write the prompts based on expert research on the matter. I built “MetaGPT,” a chatbot expert in prompt engineering—that is, a GPT expert that optimizes questions to ask GPTs. I was unsurprised a few days later when I saw this headline: “Prompt engineering is a task best left to AI.”

Getting Down to Business

If this is all sounding a little too rosy or like I’ve finally been converted, it’s worth reemphasizing that there are many ways to utilize AI inappropriately. Chatbots and language models are great at nearly solving, or even solving, some narrow use cases, but they need so much instruction for complex and specific documents that it’s still more effective to just write the document yourself.

An underdiscussed problem with this technology is that its existence does more to erode trust between author and reader than any previous technological advance. If you think about any novels that you’ve read and become attached to, you’ll probably find that without the confidence that a person made all the connections, constructed the space for you to imagine the story on your own terms, and achieved all this in a way that you can relate to, well… it’d cease to be literature altogether. AI turns music into mere sound, design into mere color.

Recent Advances

So, between these two concerns—one practical and one more lofty—what are some appropriate things to use AI for that don’t require that human direction or connection?

Some of my favorite use cases in the past few weeks have been coding (with ChatGPT 4, specifically), using it to conduct meta-studies on various types of policy, and supplying it with particular materials related to some of my roles (including international standards and frameworks related to it) to get up-to-date (but general) advice on specific business practices.


Coding with GPT 4 is much improved with its built-in code interpreter. Unlike in our previous look, GPT 4 recognizes when it has made a mistake, because it runs the code for itself, behind the scenes. When it creates an unusable script, it regenerates it after its own debugging, seemingly ensuring that the script makes sense. I started a recent program by giving it a sample input and an already-processed, desired output, then asked it to write a script that goes from A-to-B. It took some time to point out errors, but I didn’t have to get into the minutia of the code or algorithms at all. After a lot of back and forth, I ended up with a usable program that processes spreadsheet data into a usable report automatically.

Policy and Business Practices

If you’re a part of a small or medium business that has some less developed roles or processes, using GPTs to research the relationships between those processes can be a game changer. Remember, using multiple GPTs gives you the ability to create knowledge sources that can look at problems from different perspectives; if you’re interested in how two of these processes work, set up a GPT that thinks like someone in operations or finance and bounce some ideas off it.

While it isn’t a replacement for good internal communication, it can show you general bits of information to help you get a ballpark figure on something. A recent success in my experience is creating a “SalesBot” that thinks like a salesperson, simply so I could discuss how long it should take to go from a qualified lead to closing a somewhat complex deal. Supplying a little bit of detail gave me a good picture of how things go generally; since I’ve directed it to reputable industry sources,

I’m fairly confident in the answers that I get about business practices, but also want to stay flexible with my understanding and reach out to my team about their own experience instead of treating generated results as fact. On the other hand, GPTs can use reference materials from frameworks and guidance put out by government and standards agencies, and you can set up a GPT to cite its sources on anything that needs to be exact and factual. In the instances where I’ve given the GPTs books that I’ve also read, it does a good job of representing the materials.

Keeping these caveats and strategies in mind, ChatGPT does quite a lot of heavy lifting that you can’t really get from Copilot, but they’re very different products. Copilot has some very specific domains where it’s intended to work: inside Bing, inside of 365, inside of Teams; ChatGPT is tailored more to being expert and behaving a particular way. As they both mature, they may converge or diverge, but right now it seems that ChatGPT has a real edge on research and knowledge-based tasks that makes it irresistible for small teams.

-Written by Derek Jeppsen on Behalf of Sean Goss and Crown Computers Team