Creating Custom GPTs Is A Click Away. Is That Good Or Bad?

If you like ChatGPT 3.5 or 4, you'll be surpirsed by how much OpenAI added to ChatGPT-4 Turbo. Turbo runs faster, handles more words, supports images and text-to-speech, and is trained up to April 2023. But get this - Turbo lets you create custom GPTs.

You heard that right. You can create your own GPT on the fly, for any use case, with no programming knowledge. For $20/month (as of today). You can even sell GPTs you create in a GPT Store, much like an app store.

What’s this mean to your business? Do you have the talent to take advantage of it? Should you create a GPT? More than one? Or just use GPT products and services offered by others? How will answers to any of these questions impact your business model?

If you don’t have time to even think about these questions, we’ll get you started. Follow this page for weekly insights, and contact [email protected] for more information.

??

A?Rose By Any Other Name…

My editor let me lead with the acronym GPT before spelling it out because even in government consulting-speak, it’s an awkward term: Generative Pre-Trained Transformer. The definition is no sweeter…

A type of language model that uses a transformer architecture

That helped, didn’t it? Let’s break it out:?

• Generative means capable of generating new content. This is different from software we use which returns search information, executes macros, or performs calculations?in ways we expect
• Pre-trained means the model is good at understanding language and can generate coherent responses because it was trained on a massive amount of text data
• Transformer is the architecture GPT uses, which happens to be particularly effective at understanding and generating sequences of sentences and paragraphs

Generative AI is a Rubik’s cube of terms, with seemingly every important term related to others. In just this short explanation you see training, model, and architecture. Read about those and you find more, and then again. But these three are the basics and all you really need to know, for now, is that GPTs are good at generating coherent and contextually relevant text in response to your inputs.

?

What’s Generative AI Good For?

Great question. Think of generative AI as an uber-smart colleague and you’re headed toward the answer.

Want to improve how you describe an approach, simplify complex content without losing meaning, get your page count under the limit, produce a first draft, avoid blank-page anxiety… and do any of these and other tasks in seconds or minutes?

But creating a GPT... what’s that good for? Here’s where it gets really interesting.

Anything you can do with a good ol’ fashioned canned GPT, you can do with one you create. But creating allows you to add special instructions, proprietary knowledge, and focus so your GPT does one thing very, very well. Need to do five related tasks or actions? Create as many GPTs as needed. Set up takes just minutes, and you can use and refine in under an hour.

Something Borrowed, Something Blue

You have many options in the govcon market, today, with more products and services coming online every day. How to choose is an important – and challenging – question.?

As with most things, the answer is “It depends.” Be clear on your use cases, including an answer to the question why, for each. Answers to the why question should be about a business need or objective. Ultimately, all our technology use should serve business objectives and the mission, and you should be able to draw the line of people/process/tool to a business outcome.

Be mindful of technical and organizational debt. Both exist in every organization, and your teams pay interest on them every day. The truth with both is we stop seeing them. Unfinished technical and organizational business become work-arounds, other duties as assigned, being double-hatted, cut corners and “good-enough for now.” Talk to your teams about whether and how GenAI use can retire debt, not compound it.

In time, a hybrid GenAI approach will probably make sense. It’s hard to predict without specifics, but you might use a dedicated, GenAI-assisted capture/proposal tool and add a custom GPT for resume or past performance writing. You might use an AI-assisted tool for pipeline management alongside that capture/proposal too, and a GPT you create to support gate reviews and decisions.

You might also recombine as you gain experience with the general capability, and with particular tools. Considering prices for today’s products and services, you might like the return on investment.

?

What Might Go Wrong?

If you’ve been following DWPA’s GenAI Discovery Project, you know we’re methodically trying to answer questions about our own GenAI use, plus determine the services we take to market to support clients’ capture and proposal GenAI use. Experience tells us we can go wrong by moving too fast, and moving too slow.

We could move fast by deciding on the basis of assumptions, without taking time to check them. And we could slow-roll decisions because choices are overwhelming and we don’t know where to start. Our GenAI Discovery Project is helping us thread that needle.

Follow DWPA’s company page for weekly discovery insights. To learn more or launch your own discovery project, contact [email protected].