Microsoft Ignite 2024 – Highlights

Carina ClaessonPower Apps2 months ago4 Views

I just got back home from Chicago and a week full of adventure. In this post I’ll go though the the announcements that caught my attention the most and I’ll also let you in on a my thinking around the new features. Happy reading!

Summary

My perspective is Power Platform and a bit beyond and with that perspective, I want to highlight the following. Autonomous agents now in preview. New OOB agents in Power Apps and the possibility to create agent from an app. The Power Platform Admin Center has been re-imagined, along with new concepts Managed Security and Managed Operations. What we used to know as Managed Environment will transition into Managed Governance.

Azure AI Studio has been re-branded to Azure AI Foundry. Copilot Studio + Azure AI Foundry – easily use your Azure AI Foundry OOB or custom models in Copilot Studio. Create reusable parts with Copilot Studio Components. Copilot Studio going multi-modal. Microsoft Purview + Dataverse. Lastly, already announced at the Power Platform Community Conference, Power Apps – Start with a Plan. We did see a lot about it at Ignite though, so I wanted to include it. Let’s now talk about each one of these!

Building an agentic world

The communication around agents keep coming to us. From small to medium to autonomous agents. It has begun to sink in, agents are here to stay and change both the way makers build and how users interact with apps. There has been a lot of buzz around autonomous agents, and these are now in preview. Autonomous agents are operating in the background independently, and there are autonomous triggers even, so it does not have to start with a user opening a chat. Remember that custom copilots are now just agents. I learned a new word at keynote, agentic. It was shown at the keynote and it’s in the book of news. Building an agentic world!

New agent possibilities in Power Apps

At Microsoft Ignite demos were held showcasing the new possibilities to utilize OOB agents in model-driven apps as well as to generate an agent from a canvas app. It was mentioned a roll out during the next upcoming weeks for some of the OOB agents. The Data Entry Agent can take multiple sources to help fill out model-driven forms faster. Upload an e-mail or document, have the agent to fill out your form. The Power Apps data exploration agent will be utilized for searching, sorting, and filtering rows, using natural language.

In the session The Future of Power Platform: Intelligent Apps, which is now available on-demand, we got to see the bits and pieces of what builds up intelligent apps, including demos for each area.

From showcasing the Plan Designer (Power Apps Start with a Plan), to the new OOB agents and generating agent from app. To the new concepts Managed Security and Managed Operations, which were also highlighted in all other sessions describing the new Admin Center.

Below shows Generate agents from Power Apps from that same session. Public Preview December 2024.

From the Microsoft Ignite Book of News:

”The app-specific agents will leverage the logic, knowledge and actions already existing in the apps to execute tasks autonomously. The users of the apps will then be able to oversee the actions that the agents took in the app and act where agents hit roadblocks to complete.”

Interesting future indeed with all these agent and it makes me think of – in what scenarios will we create an agent for business logic when we build our custom apps? We use the low-code tools in model-driven apps today, in combination with Dataverse Plugins, Power Automate cloud flows and Azure Functions for more complex scenarios. I wonder in what scenario I will build an agent for business logic in the future… and not. Business critical functionality, maybe not an agent. Time will tell!

New concepts Managed Security, Managed Operations

Before we get to the new Power Platform Admin Center (PPAC), which got a lot of attention and totally deserved it, let’s look at an overview of related concepts. We got it explained at Ignite how we have had Managed Environments since 2022. Now Microsoft introduces Managed Security and Managed Operations. What we used to call Managed Environment will be Managed Governance in the future.

There are dedicated official Microsoft blog posts about Managed Security and Managed Operations. What about Managed Governance? Well, I assume we will see updates in the UI and official documentation over time.

There are also already new chapters in the official documentation, e.g. Managed operations in Power Platform. Managed environments… still is… Managed Environments. I expect our beloved official documentation to be restructured as these concepts are established and more detailed features are rolled out.

Notice below how we have Managed Environments in the middle and then Managed Governance in the middle. From that same session.

In the session Enterprise Scale: The Future of Power Platform Governance + Security we again saw these same concepts.

Power Platform Admin Center Re-imagined

Do you know what, the Power Platform Admin Center has been re-imagined! This is so much more than what I had hoped for, thinking about how Microsoft will go from the classic UI (legacy way to experience the settings) to managing it in PPAC. We got to see a lot (!) about the new Admin Center. Looks so good. Bye bye classic mode!

From a Power Platform perspective, this was a big thing during Microsoft Ignite. The new Admin Center has already been rolled out and there is a switch to turn it on in your PPAC.

Let’s go further into the details. The Admin Center is built up of what they call “Hubs”, e.g. “Security Hub”, “Licensing Hub”, even a “Copilot Hub”. I like it how we will be able to take the temperature of things, e.g. How are we doing with security in the ”Security Hub”, and what action do we need to take. The “Safeometer” it was called in one of the sessions.

I turned this on in my developer environment. Looks like this now:

Construction zone

Below is from theater session Leveraging rich visibility to secure AI driven business applications, we can see that barometer and we can also see that PPAC can go dark mode!

Let’s look at a few of the new settings. Certain settings are for Managed Environments only. I like it how it shows in the UI now if a certain setting is for managed environments only. In the past it has been a bit confusing to get an understanding of what features are for managed environments and what are for all environments. Ideally we want to have that clear, in UI and in the official documentation. Below we can see an icon next to what’s a “Managed Environment” setting.

About Administrator privileges, that is a good one. It will give an overview of the number of System Administrators per environment. Very good for clean-up and remove those that should not be System Administrators.

Below follows other examples of what’s new in PPAC. There’s much more in there, these are just examples and more to come!

Environment groups, setting rules to be applied to all environments in the group
Decide who should access Copilot in model-driven Power Apps
See how the Power Apps agents are doing, fine tune prompts if needed
Observe and improve on the active usage of apps
Managed on-premises data gateways form here (easier than earlier most likely)
Control who can see the support requests
Sharing limits – set how many times users are allowed to share an app
Purview -> Data catalog – Sensitivity label data sync.

An extra shout-out to my fellow MVP Carl Cookson, who held a theater session, about The Center of Excellence (CoE) Starter Kit, Managed Environments, the Power Platform CLI. He was so up to date that he showed us the new newly released PPAC, Managing Power Platform – Your guide to be an effective Administrator.

There are already new chapters in the official documentation related to the new experience, e.g. Use the new and improved Power Platform admin center (preview).

Azure AI Studio re-branded to Azure AI Foundry

Azure AI Studio has been re-branded to Azure AI Foundry. I have not used Azure AI Studio that much, but I went on a lab at Ignite, all dedicated to Azure AI Foundry and we played around with… Playgrounds! It gave an interesting perspective, thinking about how it could be utilized for my customers. Like to build and fine tune a model in Azure AI Foundry and then utilize it in Copilot Studio, now with the new built-in integration that was announced.

During the lab we learned about how language models work, predicting the next token “The next most likely token. The only thing it does”, how the temperature changes the probability of the next token, lower – more determent. To prompt engineering (concept in natural language processing that involves embedding descriptions of tasks in input to prompt the model to output the desired result). To providing the model additional context with System Message. To play around with models like DALLE-3 (Text to Image), GPT-4o (Text and Image to Text), GPT-4o-Realtime-Preview (Speech to Speach), and remembering to clear the chat with the icon.

Here are a few links that were provided for learning more about Azure AI Foundry.

“Azure AI Foundry portal (formerly Azure AI Studio), now in preview, is a comprehensive visual user interface that will help developers discover and evaluate AI models, services and tools.” – Microsoft Ignite 2024 Book of News

Copilot Studio related news

We got to see new monitoring capabilities, create agents with autonomous capabilities (triggered without user, operate in the background), as well as autonomous triggers. There are improvements to knowledge. New knowledge curation that analyses your knowledge base and gives recommendations for relevant new knowledge. Makers to utilize custom Azure AI Search indexes as a knowledge source and use OOB or Custom Azure AI models from within Copilot Studio “bring-your-own model” via the Azure AI model catalog as already mentioned.

Components are coming – group Copilot Studio stuff together as a reusable component, demoed in session Extending M365 Copilot with Copilot agents: Best practices.

Described as a piece of a custom agent to be used over and over again. Solution aware! Enables topic level import / export, any component within, package, distribute. Settings -> Component collections. Ends up in a new solution, no longer in the current solution. If you create one -> it will be shared asset. Not to be changed for one agent, but for all using the component collection.

Considering that it’s best practice to have one unmanaged solution per development environment, I wonder how this works more in detail, time and some exploration will tell.

It was also brought up – Autonomous agents – Why is this different than Power Automate?! Video is coming they promised… Keep an eye on videos from Dewain Robinson. This might will be an interesting addition to me trying to figure out when to use an agent and when to use…. other technologies.

The below picture is form session What’s new with Copilot Studio and agents and summarizes the Microsoft Ignite news for Microsoft Copilot Studio. I actually did not see this session live – the room was full when we got there! Such a popular topic. Looking at this overview, I’m eager to get a better understanding of the improvements in generative AI answers quality.

Microsoft Purview – sync labels with Dataverse

During the theater session Leveraging rich visibility to secure AI driven business applications Microsoft Purview was highlighted and also in combination with Dataverse.

“Microsoft Purview is a comprehensive set of solutions that can help your organization govern, protect, and manage data, wherever it lives.” – Microsoft Learn/official documentation.

Key takeaways: What can happen? Data leaks, Data oversharing, Non-complience usage. Microsoft Perview is there as a tool to help. In the new Power Platform Admin Center, we have a swith “Sensitivity label sync (Preview)”, i.e. sync labels with Dataverse. I was curious about when to use Microsoft Purview for a customer. One use case could be if we want to delete sensitive data such as social security number, we could have Microsoft Purview help us scan our Dataverse data and delete the sensitive data. Considering the GDPR regulation, we usually create custom solutions for such things. Can you think of other use cases?

The below picture is from session Leveraging rich visibility to secure AI driven business applications Microsoft Purview. Sorry about the quality (perhaps I need a new camera/phone for the next event?!) but you’ll see the overall concept anyway.

Power Apps Start with a Plan

Announced at Power Platform Community Conference and it has been available to try out in an Early Access program. Even so, I wanted to include it since it was all over the place during the Power Apps sessions. We saw this new feature from different angles as the product managers demoed different scenarios. You can watch the demos on-demand in Building intelligent applications with AI and Power Apps.

As well as in the session I mentioned earlier, The Future of Power Platform: Intelligent Apps.

This is a big thing for all makers and solution architects for that matter. Creating a plan from scratch with natural language as a starting point. I got the message. How we build apps is about to change. In a real-world scenario today, we live in Azure DevOps, building up user stories, categorizing in features, breaking down in tasks, prioritize user stories, have sprint planning and we do the work.

We also live in Maker portal where we create our data models and add all components needed including automation and apps. We might live a little bit in Visual Studio or Visual Studio Code as well. Some might be using Copilot for building data models and apps, some still create all tables, columns, relations and apps manually, I do. It gives me a sense of control and I know that everything will look perfect even in the background. (Of course it’s an iterative process, requirements changes, I’m not saying all solutions are all happy days from start).

So what about the future then? The overall thought seems to be that we will be making plans directly in Maker Portal/the Studio, starting with natural language and you can add screenshots of a legacy app while building your plan, describe what previous apps were used and attach ERDs. Then have Copilot to build the data model for us, build suggested apps and suggest security roles. Even get sample data. Then in an iterative process, work your way towards a solution that will meet the requirements.

This is an interesting area. I wonder if we could also connect it back to Azure DevOps somehow to track the work that we do with Copilot and the plan to our user stories in Azure DevOps. ”Power Apps Start with a plan” will enter in Public preview shortly, it’s currently in it’s Early Access days. Below is from the theater session Building complete solutions with Copilot and Power Apps.

I always think about the ALM story when new features are released, as I understand, components will end up in the ”preferred solution”, which can be set for the solution in Maker Portal. Sounds like a good start! I also wonder about all the OOB tables, like Account and Contact, which are frequently used. Hopefully Copilot will utilize the OOB tables when suitable.

That goes down also to columns, I’m thinking of all the small details that we have in mind when we build solutions from scratch, like, should we use an OOB column or a new column, no re-purposing etc. But I assume we will start making use of these plans and see it as our overall planning and that we can adjust on detailed level manually when needed. Will we be able to make any kind of documentation out of these plans, be able to export it, time will tell.

See you next year?

Don’t miss the Microsoft Ignite 2024 Book of News. Chapter 7 is all about Power Platform, however I would recommend looking at chapters 2 (AI Opportunities) and 3 (Azure) as well. I like it how the Book of News contains links to where you can read further or find on-demand sessions.

One of the official blog posts says – Embrace the era of Copilot-first development. So they are looking at YOU and me and I assume we all need to work on our prompts and fine tune how we describe what we want AI to do for us. Are you doing that yet (where suitable)?