AI agents are one of the hottest topics in technology today. Whether you’re deeply involved in the tech industry or just a casual observer, you’ve likely heard about them. Businesses are racing to develop autonomous AI agents, often marketing them as the next big thing—just as they did with the “our products are AI-powered” buzz which is largely a lot of marketing fluff. But what are AI agents really capable of? Are they just another layer of fluff, or do they represent a fundamental shift in how we interact with business applications?
Microsoft CEO Satya Nadella recently suggested that traditional SaaS models could be displaced by AI agents, an idea that raises big questions about the future of software applications. In this post, we’ll break down the traditional SaaS model and explore how AI agents could transform the way we work, using real-life examples built with Microsoft’s Copilot Studio.
How AI Agents Work
Traditional SaaS 📟 = > SaaS apps have a frontend for users to interact with that then performs some type of business logic to update a backend database. Most modern apps have API capabilities to update the backend programmatically. The data structure is usually unique to the application.
AI Agents 🤖 => Users can chat with an AI agent which can automatically perform the business logic for you leveraging a LLM. Users no longer have to click and interact with a frontend. Business logic is conducted without users having to write any code. The AI agent can interact with a backend database to perform the typically CRUD capabilities
Orchestration ⏩ => This is where things get very powerful. Given the proliferation of SaaS apps, we are typically performing actions across multiple apps for a single task. Imagine that you leverage Microsoft 365 for email and files, Hubspot as a CRM, and MailChimp as a marketing engine. Connecting these apps into an agent would allow for users to chat to orchestrate entire events across these applications. The agent can work agnostically with any data structure.
Think about a typical sales workflow:
You have a meeting with a prospect.
You take notes in Microsoft Teams.
You draft a proposal in Word.
You update the contact in HubSpot.
You tag them in Mailchimp for marketing follow-up.
Each step is siloed and time-consuming. AI agents, however, can automate this entire workflow. You could simply chat with an AI agent, which would:
Analyze your meeting notes.
Generate a proposal document.
Update the CRM with key information.
Schedule the next outreach email.
Instead of manually configuring integrations and workflows, the AI agent does it seamlessly—no code required.
Real-Life Example: AI-Powered Loan Analyzer
To demonstrate the power of AI agents, I built a simple Loan Analyzer Agent in about an hour using Microsoft Copilot Studio. In a traditional setting, a loan officer manually reviews applications, checks financial details, and approves or rejects loans. With an AI agent, this process can be automated:
A user submits a loan application through Microsoft Forms.
The AI agent analyzes the application using predefined risk criteria defined in our companies knowledge base.
It calculates key metrics like debt-to-income ratio (even if it’s not explicitly provided).
Based on predefined approval rules, the AI agent either approves or rejects the loan.
The applicant receives an email with a detailed decision explanation.
This AI agent interacts with multiple applications—Forms, Outlook, SharePoint—and autonomously executes the business logic. Not only does this save time, but it also removes human error and ensures consistency in decision-making.
Expectations vs Reality
I will be coming out with a video tutorial of how I built the Loan Analyzer Agent next week so you can get a better grasp of how this technology works. The reality is that while Copilot studio made this much easier to create, it still had some bugginess and took a lot of tweaking in the prompt to get it in the state I wanted. It’s far from, “let me just type a few sentences and I’ve got a fully working agent.” With that said, its miles ahead of what we have seen in the past where the closest you could have gotten this is manually creating some RPA workflow with Power Automate which is significantly under adopted, not just in the SMB space. It will be interesting to see where this evolves given the rapid pace of innovation and global focus on this technology. At a minimum, I would start trying to leverage the technology in your MSP to find good business use cases. Getting the reps in now will help with customer conversations down the line if they are not already asking about it.