Agentic Control Plane for SaaS

As you likely know, a SaaS application consists of two parts: the 'application plane' (also known as the 'data plane') and the 'control plane'. The application plane is a part of a SaaS application that delivers direct value to the customers, whereas the control plane is essentially the brain of any SaaS product.

The application plane is usually unique for any SaaS product, but the control plane has many similarities across different products. The typical functions are tenant lifecycle management, user management, release management, tenant-aware observability, billing management, and tenant analytics. They work together to create a comprehensive operational framework that simplifies SaaS product operations and gives the customers more autonomy.

The control plane itself also often consists of two parts: customer-facing and internal-facing. The customer-facing control plane provides self-service capabilities for tenants to manage settings, integrations, and analytics without escalating to operation team. Meanwhile, the internal-facing control plane serves as an operational dashboard for support and operation teams to manage tenants' lifecycle, adjust configurations, and review telemetry without requiring development team intervention.

This dual approach ensures both customers and operation team have the visibility and control they need to keep the SaaS application running smoothly and achieve the business objectives.

As you see, the control plane is an important part of a SaaS product which, however, doesn't generate direct revenue, unlike the application plane. This is why businesses tend to underinvest in the control plane, which sometimes undermines long-term success.

For example, underdeveloped tenant lifecycle management, such as onboarding, can limit the ability to scale customer base. A lack of tenant-aware observability extends issue resolution time, reducing customers satisfaction. And weak tenant analytics limits your ability to identify upsell and churn signals, which prevents the growth potential from being fully realised.

When implemented correctly, the control plane unlocks the ability to scale, allowing businesses to grow efficiently without inflating headcount and operational costs. However, implementing it right requires significant engineering effort, and this is where AI agents can help.

Unlike the application plane, the control plane has many similarities across different SaaS products. All SaaS products have tenants. Tenants have lifecycles, such as onboarding and offboarding. Tenants also usually have configurations which customise your SaaS product for a specific customer. And tenants obviously have activity logs and metrics that can be used for troubleshooting and analytics. A good control plane should have tools to manage all the above.

To do the job, the control plane usually interacts with multiple APIs and data sources, and very often these resources are already in place. In this case, all that is required of the control plane is to orchestrate them to provide an efficient operational framework.

With the traditional approach to software development, you need to build these orchestrations yourself, but there is an alternative. You can leverage AI agents to orchestrate your data and workflows, which allows you to shift your control plane development from the level of code to the level of configuration. All you need is to give your agent a Standard Operating Procedure (SOP) and connect it to your data sources and APIs.

The typical objection is that building and running AI agents may be even more challenging than building the control plane in a traditional way. This was somewhat the reality in the past, but now there are multiple ways to build and run your own chat agents in a no-code way. One example of this is Amazon Quick chat agents .

When you create a chat agent in Amazon Quick, you first specify a persona to give it high-level guidance on behaviour and goals, instruct how to leverage reference documents, and provide general guidance on how to use the provided resources.

Then you attach reference documents. These reference documents are part of the chat agent's permanent context. They can include complex workflows with multi-step processes, such as SOPs, which your agent must follow.

Finally, you give the agent access to your data sources and APIs, allowing it to orchestrate your workflows. There are multiple types of supported integrations, including the MCP protocol and REST API, which gives you a significant level of flexibility.

Amazon Quick's permission system makes chat agents respect existing permission boundaries. This means that when users interact with a chat agent, they can only access data and APIs they are already authorised to access.

The icing on the cake is that you can now embed Quick chat agents directly into your own web applications, allowing you to easily augment your existing control planes with agentic AI capabilities. Read this blog post to learn more about Amazon Quick embedded chats .

The control plane is a key part of any SaaS product that helps businesses scale effectively. It manages tenant lifecycles, configurations, and operations. Without a well-designed control plane, realising a product's full growth potential becomes challenging. As a product scales, the control plane enables automation and consistent operations across all customers and environments without proportionally growing your operations team.