CodeGPT Insights: AI-Powered Coding Tips & Tutorials

Agentic SaaS: Are AI Agents Killing Traditional SaaS?

Written by Daniel Avila | 12/27/24 2:37 PM

Every now and then, I see headlines proclaiming, "SaaS is Dead" It's a bold and provocative claim. I firmly believe that the SaaS model, as we know it, will continue to operate for quite some time. What's changing is the way we interact with these platforms.

To fully understand the bigger picture, we first need to grasp what SaaS is. A simple analogy often used to explain this is the Pizza as a Service or PizzaaS 🍕 diagram.

The PizzaaS diagram illustrates the different ways to implement these solutions, starting with On-Premises, followed by Infrastructure as a Service (IaaS), then Platform as a Service (PaaS), and finally culminating in the well-known Software as a Service (SaaS).

Pizza as a Service (PizzaaS)

Let me explain how service models can be understood through the analogy of making and serving pizza, highlighting the different responsibilities between providers and users in each model:

Traditional On-Premises: Make pizza completely from scratch at home

  • You Manage Everything: Since you're making pizza at home, you have complete control and responsibility over every aspect, just like running your own datacenter where you're responsible for hardware, software, maintenance, and operations.

Infrastructure as a Service (IaaS): Rent a kitchen to make your pizza

  • Split Management: The vendor only provides the basic infrastructure (like a rented kitchen), because their role is limited to maintaining the physical space and utilities. You manage everything else because you need to customize and control your specific implementation, similar to how AWS provides servers but you manage what runs on them.

Platform as a Service (PaaS): Use a pre-setup kitchen with basic ingredients

  • Mostly Vendor Managed: The vendor handles most of the underlying components (kitchen and basic ingredients) because their goal is to provide a ready-to-use development platform. You only manage your specific application elements (toppings) since that's your unique value proposition, similar to how Heroku or Vercel manages the runtime environment while you focus on your code.

Software as a Service (SaaS): Order a ready-made pizza for delivery

  • Vendor Manages Almost Everything: The vendor controls all aspects except consumption because they're delivering a complete, ready-to-use solution. You only manage how you use it, just like how you only decide when to eat the delivered pizza. This is similar to how Salesforce manages the entire CRM platform while you just use it.

Agentic Pizza as a Service (APizzaaS)

The traditional Pizza as a Service model is evolving to accommodate a new layer of autonomous interactions.

Agentic Pizza as a Service (APizzaaS) reimagines how services are delivered and consumed by introducing intelligent agents on both sides of the transaction. This enhanced model demonstrates how Your Agent and the Vendor's Agent collaborate to automate and personalize the entire service experience.

Just as the original PizzaaS model helped us understand cloud service delivery through the familiar concept of pizza-making, APizzaaS illustrates how AI agents are transforming service delivery from a human-centric process to an agent-mediated interaction, while still maintaining the core principles of service customization and delivery.

This evolution represents not just a technological shift, but a fundamental change in how we think about service automation and user experience.

In this model, two groundbreaking actors emerge to transform the traditional service paradigm: Your Agent and the Vendor Agent.

These AI-powered entities represent a fundamental shift in how services are requested, delivered, and managed. Your Agent acts as your personal digital representative, understanding your preferences, managing interactions, and making decisions on your behalf, while the Vendor Agent serves as the service provider's autonomous system, handling service delivery, customization, and optimization.

Together, these agents create a new layer of intelligent automation that promises to revolutionize how we interact with and consume services, moving us from direct human-to-human or human-to-system interactions to sophisticated agent-to-agent communications.

  • Your Agent: A personalized AI assistant configured by you to represent your interests and preferences. It operates within your environment and systems, managing interactions with various services according to your specified parameters.
  • Vendor Agent: An AI system configured and controlled by the service provider (vendor) that operates within their infrastructure. It manages service delivery, handles interactions with user agents, and implements the vendor's business logic and policies.

These agents will be the ones interacting with, controlling, and managing new generations of SaaS, which is why we must add 3 new concepts to our PizzaaS diagram: Decision Context, Feedback Loop, and Trust Layer.

  • Decision Context: The framework and parameters within which agents make decisions. It includes rules, preferences, constraints, and objectives that guide agent behavior. For example, in PizzaaS, this would include dietary restrictions, budget limits, and delivery timing preferences.
  • Feedback Loop: A continuous learning and improvement mechanism where agents collect, analyze, and incorporate feedback from interactions and outcomes. This enables agents to refine their decision-making and better serve their respective stakeholders (users or vendors).
  • Trust Layer: The security and verification infrastructure that ensures reliable and safe interaction between agents. It includes authentication, authorization, reputation systems, and validation mechanisms to maintain trust between Your Agent and the Vendor Agent during transactions and interactions.

These new layers form the foundation of agent-to-agent interactions in modern service delivery, enabling automated, personalized, and trustworthy service experiences.

So... Is Traditional SaaS Dead?

It's important to understand that traditional SaaS will continue to operate, as agents enter the scenario to handle these existing SaaS platforms. In fact, the widespread adoption of SaaS applications has created the perfect foundation for AI agents to thrive.

These agents can seamlessly interact with SaaS platforms through their existing user interfaces and APIs, essentially acting as highly efficient digital users.

Rather than making SaaS obsolete, AI agents are enhancing its capabilities, automating interactions, and making these services even more valuable by reducing human intervention in routine tasks while maintaining the robust, proven SaaS infrastructure we rely on. Think of it as SaaS getting an AI-powered upgrade rather than a replacement.

The agent can intelligently choose which path to use based on:

  • Task complexity
  • Available API endpoints
  • UI accessibility
  • Performance requirements

This means that traditional SaaS applications don't need radical changes to join this new era, as agents can interact through both user interfaces and APIs, two fundamental capabilities that every SaaS solution already possesses and that agents can effectively leverage.

But what happens when Your Agent interacts with the Vendor Agent?

In the ultimate scenario of an Agentic SaaS, we see the Vendor Agent coming into play, managing service decisions, personalizing, and completing orders without third-party interaction.

We need to establish protocols to make this communication effective and standardized. The interaction between agents represents a new layer of complexity that requires careful consideration of:

  1. Authentication between agents
  2. Trust verification mechanisms
  3. Standardized communication formats
  4. Decision-making boundaries
  5. Error handling and conflict resolution
  6. Data privacy and security measures

I'm currently working on the AIXP protocol, an open-source initiative designed to provide a foundational framework for establishing agent-to-agent interactions.
If you're interested, you're welcome to collaborate directly through our GitHub repository 👇

At CodeGPT, we keep working to ensure the successful implementation of agents across all layers of a "Pizza as a Service"... If you love pizza and software development, create a FREE account on CodeGPT and start working with our agents 🤖.

You can follow me on my social media: