The Prompt is the new Computer Interface, Again! - The new world of Microsoft Copilots!



Do you remember a time before the GUI when you were the world's greatest prompt engineer? Well, guess what?, the prompt is back and it is king ..... again!


Back to the Future

As someone from a generation where computers were seen as a niche for the "geeky", I recall a time when they lacked graphical interfaces, my weapon of choice was an Amstrad CPC 464 with a green screen. At the time, every operation required a deep understanding of specific commands to get the computer to do what you wanted - even playing games from a cassette required a specific command.

The intrigue of movies like "War Games" (1983) and "Electric Dreams" (1984) captivated me - how cool was it to be a Matthew Broderick taking on the world from my bedroom. Despite being just 8 or 9, I reveled in the thrill of typing commands into a computer and watching it respond. This era was a testament to the early marvels and complexities of computing. 

Looking back at those two films now, little did I know that the AI aspect of those films might have driven my AI obsession today. Although I will say, I am no Joaquin Phoenix in "Her" (2013) either.

The evolution to Graphical User Interfaces (GUIs) marked a significant leap. I remember encountering the mouse for the first time and being awestruck by the new possibilities it heralded. GUIs transformed the computing landscape, turning what was once an alien concept of pointers and windows into a mainstream phenomenon. This shift brought computers closer to a broader audience, no longer exclusive to the tech-savvy.

Parallel to my journey and career with Windows environments, the power of the Linux command line continued to intrigue me. The ability of Linux enthusiasts to automate mundane tasks was both enviable and inspiring. The speed to automate from a simple text prompt was a paradox; it was a skill from a previous era but was a precursor to the future of human-computer interaction of the future.


The Evolution of Human-Computer Interaction


The core challenge in human-computer interaction has remained constant: making computers intuitive and user-friendly. Despite the advancements in GUIs, there is still a learning curve involved for end-users. The idea that computers should intuitively align with human needs, rather than the other way around, has been a long-standing vision in the tech world.

Voice input technologies, like Alexa, have been groundbreaking, simplifying daily tasks through voice commands. Yet, they often struggle with understanding the nuances of natural human speech. The real breakthrough came with AI and the researchers at companies like OpenAI and Microsoft making strides in AI advancements, which have significantly improved digital assistant capabilities to what seems like black magic today.

Microsoft's recent developments, as highlighted in the 2023 Ignite sessions, have been particularly noteworthy. My predictions late last year about the integration of AI within the Power Platform, particularly the Power Virtual Agents (PVA), are coming to fruition. The rapid adoption of these technologies, as indicated by OpenAI's impressive user statistics, underscores their transformative impact.


At Nasstar and Kaizenova, we've been at the forefront of exploring and leveraging these AI capabilities. Our project to develop an AI-powered "Expert Finder" within PVA was an exciting venture, despite the initial challenges in integrating it with a broader Generative AI solution. Our cmdline AI assistant, named "TAI", was an experiment in using Generative AI to build complex Microsoft Teams routing rules from a simple user-generated prompt. TAI could take a request such as "Build me the routing tables for Mexico City and apply them to the customer tenant". And within seconds it had the Teams/SBC environment setup.




TAI was built well before OpenAI introduced function calling, or the new Microsoft Copilot Studio integrating Power Automate flows. In building TAI, we learned a lot about how Generative AI isn't the answer to all automation problems. We called it "fuzzy" responses and it took a lot of prompt engineering to teach it how to speak JSON in a consistent but meaningful way. Think training a 5 year old and the response is "But why??". 

The pace of innovation means that those early challenges of integrating the Teams PowerShell module, external API's and Generative AI together, have now been made significantly easier with modern tools, if there is even a concept of modern AI tooling yet.


Microsoft Copilots: A New Era of Computing

The transition of PVA into Microsoft Copilot Studio is a game-changer. This platform, along with Azure AI Studio, is set to redefine the way we build applications for our users to work at their pace and on their terms. Microsoft Copilot's demonstrations have showcased its ability to understand and execute complex tasks based on simple prompts.

The integration of AI into everyday computing tasks, as exemplified by Copilot in Windows, is paving the way for a more intuitive and responsive computing experience. For instance, the AI's ability to suggest enabling dark mode in response to a user's discomfort is a small yet significant step towards user-focused automation.

Imagine a world, in the not to distant future, a conversation between a user and an AI assistant like this.


User: "I'm struggling to manage my inbox. I've heard about creating rules in Outlook to organize emails. Can you set that up for me?"

Windows Copilot: "Certainly! I can create a rule for you. What criteria would you like to use for organizing your emails?"

User: "I want emails with 'Project Update' in the subject to go into a specific folder."

Windows Copilot: "Understood. I can create a rule to move emails with 'Project Update' in the subject to a designated folder. Do you already have a folder for this, or would you like me to create one?"

User: "Please create a new folder named 'Project Updates'."

Windows Copilot: "I'm creating the folder 'Project Updates' and setting up the rule now. Just a moment..."

[A brief pause]

Windows Copilot: "The folder 'Project Updates' has been created, and all emails with 'Project Update' in the subject will automatically move to this folder. Is there anything else you'd like to set up?"

User: "That's perfect, thanks! Can you also make sure I get a notification for these emails?"

Windows Copilot: "Certainly, I'll adjust the settings so you receive a notification for each email that goes into the 'Project Updates' folder. This will be active in just a moment."

[Another brief pause]

Windows Copilot: "Done! You'll now receive notifications for emails in the 'Project Updates' folder. If you need more adjustments or have other tasks, feel free to ask."


Microsoft's vision is that this will become a common way of working. Think of it like Captain Picard's "Make it So!", As a user, you know what you want, you shouldn't need to know how to do it.

This example is almost here with the M365 copilot. But what happens if you need something more specific that the copilot doesn't know how to achieve. This is where Microsoft AI consultants come in and help "train" your M365 copilot. These are YOUR version of copilots, grounded in your data, shown how to use your existing applications API's, and trained in your ways of working!.

Your copilot will know how to apply for a holiday on your behalf, will understand how to create a Microsoft Teams resource for your upcoming flight, and many more specific requirements that only businesses themselves will know that bring value to their own business.


AI's Role in Business and Beyond

In the business realm, Microsoft Teams is becoming a central hub for AI integration. The potential for AI copilots to understand and execute tasks based on natural language requests within a business context is immense. Our work at Nasstar and Kaizenova involves extending these capabilities to create "ChatApps" - applications that leverage chat interfaces to intuitively understand and fulfill user requests. At Nasstar we have already built tools where our teams can ask "start a conversation with the expert in SIEM technologies" and the AI will begin a Teams chat with our Microsoft Sentinel experts. We haven't told the AI that Sentinel is a SIEM product, nor told it how to correct for spelling mistakes, but the result is that the ChatApp knows how to map those experts together with the user.

Our goal is to take the ChatApp out of trial from our department into the rest of the business such as HR and Legal.

Looking ahead, the way we interact with technology is set to undergo profound changes. Just as Generation Z finds traditional shared drives outdated, future generations might view GUIs and manual document creation as relics of the past. Maybe even the concept of a 'document' itself may evolve, but that is for a future blog post.


Envisioning the Future with Generative AI and Microsoft Copilots

A glimpse into this future can be seen with OpenAI's ChatGPT. The ability of this AI to adapt and modify images based on user feedback is a testament to the potential of AI in creative and practical applications. If you haven't tried DALL-E 3 in chatGPT, it is like black magic.

As an IT consultant, I am excited about the possibilities these advancements bring. The concept of generating comprehensive documents for clients with a simple prompt in a Chat application like Microsoft Teams is no longer a distant dream but an emerging reality.

The prompt really is the new interface, and maybe a world envisioned in the film Her is not so far away! Less the love interest between computers and humans, of course.



Chris Burns is the Automation and AI capability lead for Nasstar, and an evangelist for all things AI in the industry. He has spoken in many Azure events in his native North East.



Comments