SaaS, Software as a Service, is a model in which we use software over the internet rather than installing and managing it on our computers. We design with Canva, for example, and hold online meetings with Teams. Instead of buying the software, we usually access it through a monthly or annual subscription. People open the application, enter information, weigh their options, and carry out the work themselves using the software's features.
GaaS, Agentic AI as a Service, provides AI agents that can use different tools, software, and data sources to reach a particular outcome. The user describes the desired result, and the agent organizes the steps needed to get there. It uses permitted tools and returns the completed work to the user. Situations it cannot resolve, or that fall outside its authority, are left for human review.
The distinction between the two models becomes clear when we look at who manages the work. Many SaaS products now write text, produce summaries, and offer suggestions. People still manage the work, though. In the GaaS model, AI carries out a defined part of the work within agreed boundaries.
How does the approach within organizations change?
When a new need arises in an organization, a common response is to choose suitable software and give employees access. Work is then carried out using the platform's features. The SaaS model has long been part of this approach.
AI agents make it possible to look at that need differently. Some steps in employees' existing work can be carried out on their behalf, rather than giving them another tool. This changes the intrapreneur's role. Alongside choosing software, intrapreneurs examine how the work is done. Designing which steps people will handle, which the agent will carry out, and where control will remain becomes part of the role.
| Comparison criterion | SaaS | GaaS |
|---|---|---|
| What does it offer? | Software to use | A task to carry out |
| Who carries out the work? | The employee using the software | An AI agent under human control |
| What is the human's role? | Performs the operations | Sets the goal and boundaries, evaluates the result |
| How is success measured? | Usage and use of features | Completed and accepted work outcome |
| What is the main risk? | Insufficient use of the software | The wrong work being carried out without oversight |
SaaS and GaaS comparison
Where should intrapreneurs look?
AI-Native intrapreneurs do not need to design a new agent for every process. It makes more sense to start with how people work day to day. Repeated tasks, information moved between several applications, time-consuming comparisons, and requests repeatedly assessed against the same criteria can all be useful areas to examine.
Imagine a salesperson who regularly checks customer records. They identify missing conversations, prepare follow-up messages, and enter the results into the system. Developing a new reporting screen could be a SaaS solution. An agent that reviews records, identifies missing follow-ups, drafts messages, and adds approved actions to the system is closer to the GaaS approach.
How tasks are shared matters in this example. The steps the agent will take on and the decisions it will leave to the employee need to be defined at the outset.
When does GaaS make sense?
GaaS can help with work where the outcome can be described but the path to it sometimes changes. Processes that gather information from several sources, require some choices, and allow uncertain situations to be referred to an employee are particularly suited to this model.
There is little point in handing an agent simple, unchanging work that existing software already handles reliably. For some needs, a good SaaS product remains the best solution. GaaS makes more sense when work calls for different steps depending on the circumstances and involves several tools used together.
As Feluda's comparison of GaaS and SaaS also emphasizes, the two models do not have to replace each other. AI agents can carry out their tasks using existing SaaS applications. SaaS can remain the system that holds records and core functions, while GaaS becomes a layer that carries out work across those systems.
Testing it through a small task
Trying to change an entire department in the first experiment carries unnecessary risk. A single task with clear boundaries can be selected instead. First, examine how the work is done today, then identify the part AI could take on. Comparing SaaS and GaaS options for the chosen task in the same table helps put in writing which steps will go to the agent and which decisions will stay with people.
Five things should be clear in a small pilot:
-
- The business problem to be solved
-
- The steps in the current process
-
- The tasks AI can take on
-
- The decisions requiring human approval
-
- The criteria that will demonstrate success
The pilot's evaluation should track employees' time and output quality as well as whether the system works. It can also reveal where human intervention is needed.
The capabilities intrapreneurs will need
Assessing an opportunity to use an agent begins with understanding the problem behind the apparent request. Knowledge of tools is part of that investigation. Intrapreneurs also need to examine the process, question the reliability of the data, and develop different ways to solve the problem. Problem definition and solving, critical thinking, creativity, data literacy, decision-making, and adaptation are among the core capabilities of AI-Native intrapreneurship.
In the SaaS era, intrapreneurs brought new tools into their organizations. GaaS adds the work of designing, together, what people will do, which tasks AI will carry out, and where control will remain.
Ask yourself
- Which recurring tasks in our organization have so far been addressed only by seeking new software?
- When setting the boundaries of the agent's tasks, which decisions need to remain subject to human approval?
- Which measures will we use to track time savings and output quality in the small pilot?
