There’s been a lot of discussion lately about whether AI is fully ready to take on everything we imagine, or if it still needs time to mature, while those debates continue, I’d like to shift the focus and share some of the amazing, practical features we can already take advantage of with AI today.
When I first came across NLP models, I thought they were just knowledge bases trained to know everything. But my academic side kicked in, and I realized GPT and similar models are more like smart language calculators, NOT actual databases of knowledge.
They don’t “understand” things the way we might think. Instead, they process patterns based on their training. This got me thinking—how can we use this for business? That curiosity led me to explore practical ways to leverage these AI models for solving real-world problems.
Business Case
So, in my mind, the model should to be like below.
%%{init: {'themeVariables': { 'fontSize': '18pt'}}}%%
flowchart TD
B["fa:fa-users Users"]
B--> | fa:fa-comment-dots Interact with| C((fa:fa-gears GPT))
C--> | fa:fa-spinner Communicates to| D(fa:fa-server Service)
D-.-> | fa:fa-code Provides data| C
C-.-> | fa:fa-table Responses back| B
A few things should be clarified to make the above model working.
%%{init: {'themeVariables': { 'fontSize': '16px'}}}%%
mindmap
root((COPILOT))
GPTFacts["__GPTs are__
Language Calculator
__NOT__ a Database!"]
ent["__Enterprises Demand__
Data Privacy
Applicable Use-Cases"]
gptcapb["`__Extended GPT Capabilities__
API Interactions
DB Interactions
Interaction with Private Resources`"]
* I know you’re bored of that much theory but here it comes the funny part 😃
We can think about many use cases but I’ll focus only on one in this article.
CASE: Call Centre AHT (Average Handling Time)
- Benefit
👉 Cut costs by reducing agents’ screen time. - Tangibles
👉 Agents can use GPT on the go to access the Knowledge Base.
👉 GPTs create case records, with agents approving or enriching.
CASE: Management and Sales Enquiries
- Benefits
👉 Instant omnichannel (SMS, Teams, speech) access to mission critical information on the go - Tangibles
👉 Informed decision making
👉 Proactive problem solving
👉 Data-Driven and factual
👉 Competitive advantage
Of course you can adapt or even add extra cases depending on your business but above are just a few options of what you can do with this kind of integration.
Functional Perspective
Thanks to Danielle Sutton who has inspired me to check Copilot Studio in her presentation in UK Dynamics 365 & Power Platform User Groups at #Birmingham, I thought it would be good to check the options in Copillot whilst doing things in #Azure AI Studio.
Well, not too much surprisingly I found that Microsoft has already added the feature we discussed at the beginning.
Topics
Topics are simple workflows that you can orchestrate Copilot activities.
Knowledge
When you create a Copilot Topic, you can add Knowledge sources. Copilot is smart enough to understand the user and follow your instructions to pull info from sources like Dataverse, DevOps, Salesforce, SharePoint, SAP, Oracle, SQL Server, and even your custom Knowledge base.
Actions
When you design your Topic, you can connect with Power Automate, Excel, SharePoint, and others using Actions, giving it more flexibility and almost limitless extensibility.
Channels
It wouldn’t be enough if we couldn’t escalate conversations outside of Copilot for direct user interaction. Check out the Channels box for more options to escalate conversations beyond what’s shown in the image.
Technical Considerations
Assumptions: Reader knows how to create a Copilot Topic
Well, technically it’s totally Low-Code and even No-Code 😃
Below is a sample Copilot screen
It’s almost impossible to demonstrate the configuration for all of the features in here but I just want to touch on one of the Knowledge sources which is Dataverse.
So first of all, if you want your conversation to gather information from one of your sources then it should be defined as Knowledge source like in below. As you can see, I added a Dataverse as knowledge source.
You can follow below steps to do that
* sorry for mobile users but this animation needs to be seen in desktop browser.
Here comes the most exciting part, testing Copilot if it’s really capable to understand and provide the data from our Dataverse?
* sorry for mobile users but this animation needs to be seen in desktop browser.
Conclusion
Copilot Studio is now a container designer for Power Virtual Agents and seems like Copilot itself will gain more and more attention in near future and taking place in our lives more oftenly than before.
This one in here to share one of my experiences as well as the applicable cases that inspired me.
Thank you so much for your patience if you’ve come accros this line. 👍 😃