As Satya Nadella, Microsoft CEO, pointed out during the Bg2 podcast, the future of work may be moving towards Copilot that communicates with business applications via agents. This suggests that in the AI-, agentic-driven future, users will primarily perform tasks through AI interfaces like Copilot, which interact with one or more SaaS (software as a service) applications behind the scenes. If we follow this logic, SaaS apps might evolve into back-end utilities, storing data and providing APIs for AI agents to orchestrate workflows. Satya also mentioned that specialized interfaces will remain for complex tasks, but most workflows will be streamlined via natural language queries using an AI-based interface.
This might sound visionary, right? Now these scenarios are already possible or will be soon possible, and the following examples illustrate them:
Scenario 1: User performs a task through an AI agent
- Today: Users sign in to a specific SaaS application (e.g., Dynamics CRM) to perform tasks like data entry, report generation, or customer management.
- "Future", via AI agent: Instead of signing in to Dynamics to update a lead, the user might simply ask their AI agent via Copilot: "Update the lead John Smith to 'Closed Won'". The AI agent handles the interaction with Dynamics behind the scenes.
Scenario 2: AI agent-to-AI agent communication
AI agents will orchestrate workflows that span multiple SaaS applications seamlessly:
- Today: A user manually exports data from one app (e.g., HubSpot), uploads it to another (e.g., Dynamics), and generates a report.
- "Future", via AI agents talking to each other: The AI agent will automate this entire process. A request like "Generate a financial report based on Q4 sales data" triggers the AI to pull data from HubSpot (AI agent 1), process it in Dynamics (AI agent 2), and present a ready-made report, all without direct user intervention in either app.
Scenario 3: AI agent as the jump-off point
Example: User starts content creation in Copilot, then uses "Edit in Pages" (similar to Canvas in ChatGPT Plus), and is redirected to Word for finishing the content work.
Scenario 4: Complex tasks
For highly specialized or visual tasks (e.g., creating a model, designing a presentation, going from draft to publishing mode in a piece of content), users still enter SaaS applications.
Example: Excel as an interface for advanced data modeling, with Copilot acting as a powerful assistant within the application. The same can also apply to PowerPoint, Word, or news creation in SharePoint.
Is SaaS dead?
After Satya’s appearance on the Bg2 podcast, I saw many videos and posts on social media of people referring to this interview with a statement such as “The death of SaaS” applications. In fact, this wasn’t the message that Satya wanted to convey, and this interview with Varun Mayya provides more details on this discussion: Watch the interview.
Now, agents provide something very special. They go beyond just information retrieval, performing tasks on a user’s behalf. This applies even when a task might span multiple SaaS applications, which results in a huge productivity gain for end users as they don’t need to access different applications individually. From the point of view of an end user, it doesn't matter which business applications operate behind the scenes because their main point of contact is their AI agent, which, in the case of Microsoft 365, is accessed from Copilot Chat.
A colleague of mine, Vardhaman Deshpande, shared a recent post on LinkedIn where he provides a very tangible example of a leave request agent that shows how the agent orchestrates a leave request involving SharePoint and Workday, with Copilot Chat as the interface for interaction. For more details, see his LinkedIn post.
I think a lot still needs to change if we want to experience a digital workplace where agents serve as the main interface for interacting with business logic. The use cases that agents provide need to be clear to end users and work reliably so they can move from using graphical interfaces to performing their daily tasks via chat-based interactions with AI agents.
Excitement meets reality
I joined the Microsoft AI Tour in London this month, where I had the opportunity to talk to people working at organizations from different sectors and sizes. Indeed, many organizations have already started their AI journey. Some have already rolled out Copilot to all their employees, some are piloting Copilot, and others are already exploring custom AI development using Azure OpenAI.
Now, some organizations are still trying to solve basic or budgetary challenges, like moving from an on-premises intranet to SharePoint Online or consolidating different third-party applications under Microsoft 365. While AI is interesting in those discussions, it’s not their primary or even secondary goal, and an agentic AI world feels far from reach.
This doesn’t mean that such organizations don’t want to adopt AI. I believe that the adoption of Copilot is a matter of time. In the end, moving to Microsoft 365 provides organizations with the baseline to start their Copilot journey. If you're looking for ideas to plan your Copilot launch, see How Viva Amplify supports your Microsoft 365 Copilot launch.
Conclusion
An agentic world brings many possibilities, and apparently, it comes with huge productivity gains for end users. Some organizations haven’t started their AI journey yet or are still at the very beginning, but rest assured that AI is an important topic for many IT directors and CIOs out there. It's a matter of when, rather than if, Copilot (or any other generative AI option) will be adopted.
As a software vendor, I believe that companies building applications need to work on multiple gears. You need to understand the future but also not forget the present. As Satya said in his recent interview with Varun Mayya, one of these gears is understanding the impossible thing that I can make possible with what's coming, and the other aspect is optimizing the present. You can't forget the current requirements of clients, as the problems they are trying to solve today haven't disappeared just because of AI.
We're going through a platform shift that will significantly impact how we experience work in the future. Right now, we see software vendors launching their first agents that integrate with their SaaS solutions. As a next step, we might see applications that are optimized for use via AI agents, which. So, this future doesn't seem very far away. Now, for end users to really support this platform shift, they need to experience a huge productivity gain. A gain that is so significant that one no longer needs to access the "traditional" applications because the AI agents seamlessly handle the tasks.
Thanks for reading,
Jarbas
