After an early lead with AI integrations and investments, Microsoft continues to advance its position in the AI world. Following the Copilot app (formerly known as Bing Chat) and other AI integrations across its products (e.g., Office suite), its latest move is to make it easier to build your own chatbot.
The company’s new Copilot GPT Builder lets users build and deploy customized chat assistants, built on Chat GPT’s underlying functionality and focused on specific tasks. This sits alongside purpose-built GPTs that Microsoft has configured for specific functions within Microsoft’s Copilot Pro app.
The new build-you-own function now joins those pre-ordained templates, letting end users tinker with customized versions in a low-code way. For example, an HR professional may want to build a Copilot GPT agent to help them quickly find information within a specific set of company policy documents.
Similarly, an equities researcher or hedge fund manager may want to quickly pull up recent information or announcements about a given company to support trend analysis and investment inputs. The point is that use cases and end-points are broad, giving users the ability to choose their own adventure.
Under the Hood
Going under the hood, Microsoft pulls this off in a low-code way by letting users train their customized chatbots by linking them to a data source. So sticking with the above examples, data inputs could include things like a Google Docs folder, a premium research library, or any other text-based source.
After doing this, there’s a setup process to configure the chatbot further, including name, description, etc. Part of this process also includes uploading a file for retrieval augmented generation (RAG) – a technique often used with large language models to ingest data that wasn’t in their initial training sets.
But to be fair, we should acknowledge that Microsoft isn’t alone, nor first, with this functionality. A similar offer exists from none other than OpenAI, which can be accessed within its Chat GPT Plus paid tier. It similarly lets users train and build their own versions of ChatGPT for purpose-built functions.
In fact, Microsoft’s latest move could indicate yet another way that it’s competitive with OpenAI. This could get complex as Microsoft’s financial investment in OpenAI so far has given it an inside track on integrations, including its early-mover advantage with Bing Chat… which went on to become Copilot.
Copilot GPT Builder is now available to all Copilot Pro subscribers. Copilot GPTs are custom versions of Copilot with specific instructions & docs for grounding (RAG) chat responses. Build one interactively in the Creative tab guided by Copilot, or directly in the Configure tab. pic.twitter.com/DWgxdkkjPe
— Jordi Ribas (@JordiRib1) March 11, 2024
Ah-ha Moments
Backing up, one of the outputs of Microsoft’s latest move could be demonstrations of chatbot versatility. It can then utilize those sort of like Microsoft Office templates (presentations, resumes, research reports, etc.). They could also reveal, and provide inspiration around, innovative endpoints for custom chatbots.
Sticking with that last part, the world is still trying to wrap its collective brain around the ways that AI can apply to, or unlock value within, myriad functional areas. Otherwise known as the al-dente test, this will be an experimental and iterative process that will take years, and include unforeseen results…
Or at least that’s what we’ve seen in past tech revolutions, such as the smartphone. It took time for developers to gain their native footing. The results included things that no one had imagined initially, including everything from Uber to TikTok. These apps seem obvious now, but certainly weren’t in 2007.
That’s the exciting part about where we now sit with AI…. there are lots of ah-ha moments to come. Meanwhile, platforms that democratize that learning curve – such as Copilot’s no-code approach – will accelerate things. Meanwhile, anyone can get started with this process in CoPilot’s $30/month Pro tier.
