Most of us have used AI to answer a question or help write an email. The next step is giving AI the ability to do the work itself.
That is where AI agents come in.
Think of an AI agent as a digital assistant with a job to complete, instructions to follow and access to the tools it needs. You give it a goal, and it works through the steps required to achieve it.
For example, a parent emails your swim school asking about make-up lessons. An agent could read the email, check your make-up policy, look up the student’s eligibility and available classes, then prepare a reply in Gmail for your team to approve.
Instead of copying information into an AI chatbot, the agent gathers the information through its connections to your systems.
How does it connect?
You may hear terms such as APIs and MCP. In plain language, these provide ways for an agent to communicate with your software.
An API is like a doorway into a platform, allowing authorised tools to retrieve information or perform specific actions. MCP provides a common way for AI agents to discover and use those tools.
With the right connections and permissions, an agent could work across your email, swim school software, website and policy documents.
How does it know what to do?
You give it written instructions explaining its role and your rules. These can be saved in simple text documents called Markdown files, which have a .md extension.
Your instructions might say: “Use our current make-up policy. Be friendly and concise. Draft replies for approval. Refer refund requests to the manager.”
These instructions guide its behaviour. The permissions you grant determine what it can access and change.
What could this mean for swim schools?
Beyond drafting customer replies, an agent could:
- Update your website after you approve changes to a policy document.
- Match vacant class places with suitable families on your waiting list.
- Prepare follow-up messages for students who have missed lessons.
- Review outstanding enquiries and give you a daily list of decisions needing attention.
Initially, your team might approve every action. Over time, you could let agents complete familiar, routine tasks independently, while keeping approval requirements for exceptions and sensitive matters.
Why could this become normal within 1–2 years?
My expectation is that, over the next one to two years, AI agents will become a normal part of running a swim school.
In fact, many businesses are already using agents to take action.
For example, at First Class, we are using agents to review and test our new code and features.
Agents also help draft our release notes before each release. To do this, they need to interact with multiple systems to understand which features are being released and what changes have been made to the code.
This work previously required a human to painstakingly review all the features to be released and document their functionality step by step.
The rate of adoption for swim schools will depend on software connections becoming available and agents proving reliable in everyday use.
Adoption could start with one simple task, such as drafting enquiry replies. As teams gain confidence, they can hand over more of the repetitive work.
People will still need to check results, resolve unusual situations and take responsibility for decisions. But much of the daily administration could happen in the background—with staff directing the work and focusing their time on teachers, students and families.
The First Class API can now be connected to tools like Claude and those from OpenAI to help you start building agentic workflows in your business.
If you would like to discuss the possibilities for your business, please drop us a line at info@memberretentionsystems.com.