Exploring the Application of Microsoft CoPilot and Azure OpenAI.

Although Microsoft's CoPilot may not be launched yet, the technology propelling it has been in existence since November 2022. CoPilot isn't exactly the same as OpenAI. It sits as a layer above OpenAI and, by extension, the Azure Open AI LLM (Large Language Model) engine, using Azure Open AI as its central processing unit. It employs the Microsoft Graph API to access user and tenant data, and is integrated with various, increasingly available, Microsoft tools.

Please note, we only talk about M365 copilot, "CoPilot" is a brand and Microsoft intends to release many versions of it, such as Security Copilot and PowerPlatform CoPilot.

We, at Automastery are eager to demonstrate the advantages of GPT technologies through customer demonstrations, underscoring how GPT is a game changer for knowledge workers across organizations.

Data remains the backbone of many businesses. Despite decades of IT departments grappling with data organization, a substantial part of this valuable resource remains as unstructured, hard-to-access data. Think about the volume of knowledge hidden in employee email inboxes that in many cases is untouched.

In the past, data entry clerks have spent countless hours inputting data, often from printed papers or random emails with sensitive details like credit card numbers. People are costly, making data input expensive, especially when trying to structure an increasing amount of unstructured data. This is where GPT tools like CoPilot and OpenAI come into play, revolutionizing the approach for organizations that are ready to embrace the future.

Here are some specific automations we're developing on Azure and M365 for customers and our own internal use:

Sentiment Analysis of Viva Engage, or *cough* Yammer Posts

At Nasstar, we appreciate employees who go the extra mile, which we express through recognition tags in Yammer posts. Our automation scans Viva engage feeds for such recognition posts, employing AzureAI to examine the text content and determine the relevant category of recognition, such as teamwork, customer excellence, or leadership. 

Once a Yammer post is made and the Power Platform detects the message, we use PowerAutomate connectors to collect the message and send the text to the LLM for categorization. The Azure Open AI model is instructed to specifically use a structured JSON reply for easy category identification.

Following this, we look up the praised user's manager using the Graph connector and send an alert via Teams to them, highlighting the recognition. The user is then added to a SharePoint list, and an approval workflow for a recognition gift is triggered. 

With approval, we return to the LLM to generate a congratulatory message attached to the post. This small automation requires minimal effort from the user or the manager, other than approving the workflow.

A Win/Win situation that builds staff satisfaction and retention.

Expert Finder from a Random Thought

The traditional way to match needs with a list of options is to use drop-down menus. However, in this modern world, data should be delivered to the user rather than making them search for it.

For example, it can be frustrating to locate the latest accounts payable Excel workbook or the skills matrix. Microsoft Syntex aims to solve this file hunting problem, but it doesn't address the issue of data surfacing in response to quick questions.

That's where Co-Pilot steps in. With Azure Open AI and Power Virtual Agent, we've developed a PVA bot that integrates with Teams, a website, or even Alexa. The bot fields user inquiries and connects them with the relevant experts.

Instead of maintaining a hefty Excel sheet with technology synonyms, we leverage AI to find the match. Our bot surfaces expert information quickly and easily, and can also conveniently schedule meetings between the experts and the user.

CLI Assistants

The film "Her" - Warner Bros copyright

Our team has used Azure Open AI to create "Tai," our command line assistant. Built in Powershell, Tai offers data manipulation assistance or knowledge support at the command line. "Tai" is an affectionate name that is an abbreviation of the powershell cmdlet “Talk to AI”.

Our consultants are experts in building Microsoft Teams deployments, but ‘tai’ brings some extra hands just when it is needed. We can ask Tai -

 "> Tell us which cities this customer has sites in, and from that data give us a set of telephone number transformation rules for each of those sites". 

with the returned data we can very quickly deploy both Teams direct routing and telephony equipment in minutes rather than days.

Tai is great at sorting the unsorted. For example, how many times do you get a CSV file that you really want in another format because the application needs it in XML, for example. - Tai comes in handy again.

"Tai, take this data [data] and refactor it into XML that salesforce can understand"

I love the quote from Arthur C. Clark - "Any sufficiently advanced technology is indistinguishable from magic." It very much feels like we have arrived at that point.

Future of CoPilot

We're thrilled to explore the upcoming features of Microsoft CoPilot. The ability to generate a presentation from a customer's recent communications will be phenomenal, but we're looking beyond that. We're keen on developing Graph extensions to expand CoPilot's already extensive capabilities.

The challenge for customers will be deciding who within their organization could benefit most from CoPilot. Is it the Excel expert or the one who barely navigates basic text formatting? Stay tuned for a future blog post on persona mapping and maximizing your CoPilot investment.

If you are looking to deploy PowerPlatform or looking for advise around implmenting CoPilot and AzureAI in your own tenant, our consultants can help just give us a shout for an easy discussion.