Connecting a website to business tools: when to automate without creating a mess?

Automation is about saving time. But can your business really benefit? To find out, look at what you do several times a day: copying a request, finding a client's contact details, sending the same message, or checking a slot. Which of these actions are repetitive? And which could you delegate without losing control over the result?

For a small business, the goal isn't to hand everything over to a machine. It's about identifying a specific workflow, understanding the decisions within it, and then letting the website handle the steps that follow clear rules. The time recovered can then be used to respond to cases that truly deserve your attention.

Starting from what you already do

Imagine someone requests a quote from your form. Today, you read their message, copy their details into a spreadsheet, send an acknowledgement, and then look for a slot to call them back. Each action is simple. It's their repetition, between two other tasks, that ends up being a burden.

Before choosing a solution, I recommend following a request from start to finish. Where does it arrive? What information do you copy manually? Who decides what happens next? And what happens when information is missing? We often discover that the problem isn't one big complicated operation, but a series of small actions that are easy to forget.

A task can be automated when its trigger, information, and expected result are sufficiently defined. If every situation requires a different judgment, you can still prepare the work without automating the decision itself.

A website that also becomes your business tool

A traditional site presents your business and collects requests. Then, you open another software to track prospects, a calendar for appointments, and sometimes a third tool for quotes. This workflow can work, but it also forces you to switch between spaces and move information around.

At Cadarsir, I offer a different approach: integrating the business functions you need directly into the site. The visible part welcomes the visitor and collects their request; the management part allows your team to track it and act according to the rules defined for your business. Both work together as a single unit.

For example, instead of copying a form into a Customer Relationship Management (CRM) software, we can set up a lead sheet within the site's interface. Useful information and actions are then in the same place. Depending on the project, a function previously handled by external software can be taken over; this doesn't mean you have to replace all your tools at once.

Centralising can also allow you to do away with certain subscriptions that have become unnecessary. It's not a guaranteed saving: you have to compare the costs of implementation and maintenance with those of existing solutions, without forgetting the functions you actually need. The right choice depends on how you work, not on the number of software packages you can eliminate.

What if you want to keep your tools?

You might already have software that suits your team, or an existing site that you don't want to turn into a full management platform. In this case, we can look at more targeted automation and, if necessary, a connection with existing tools.

An API is an interface that allows two software programs to exchange information according to rules set by their designers. If the external tool allows it and access is available, the site can, for example, retrieve certain data to display in your interface, or send it a request. The software continues to exist outside the site: we haven't replaced it, we've organised its dialogue with the site.

For Miraj Paris, the idea was to keep Mindbody for bookings while retrieving information about classes, teachers, and other staff members to display them in real-time on the site. The team keeps their booking tool, and visitors find this information online without having to consult a separate interface.

It's necessary to specify what information flows, in which direction, and which tool is the source of truth when data changes. A connection also depends on the possibilities and terms of the external tool. On an existing site, we first check what it allows before promising an integration.

From form to follow-up: a concrete example

Let's take a fictional scenario. Someone fills out your form to request an appointment. Instead of letting their request sit in an inbox, the site can create or update their lead sheet with the necessary information, then send an acknowledgement message. You find the request and its history in your interface.

From this same sheet, buttons can help you offer an appointment. Depending on the chosen rules, part of the preparation can be automatic while you retain control over the slot or the message. The idea isn't to send a confirmation for availability that hasn't been checked: it's to save you from re-keying data and to make the next action easy to find.

The hospitality sector offers another example, this time around events. For Le Coq Noir, weekly evenings can repeat automatically: the site continues to display the schedule without the team having to recreate each appointment. If an evening needs to be cancelled, an exception allows that occurrence to be removed without stopping the entire recurrence.

Automation and AI agents: two different things

In the previous examples, automation follows a pre-defined sequence: when a request arrives, record the useful information, send a message, and then flag what requires intervention. It applies the rules set for these steps. There's no need for artificial intelligence to move a form to a sheet.

An AI agent, on the other hand, can help with more open-ended requests. For example, it can analyse a freely written message, identify missing information, and prepare a response. To act on your business, it also needs tools it can access: consulting a sheet, preparing a message, or searching for a slot doesn't happen by magic.

The two approaches can complement each other, but they don't yield the same results or guarantees. An agent can make mistakes or misinterpret a request; its access must therefore be limited and its actions adapted to the desired level of control. To explore this example further, I invite you to read my article on AI agents for small business customer service.

Reviewing data before linking it

Whether functions are gathered within the site or external tools are kept, you need to know what information is useful. For an appointment, a name, contact method, and desired slot might be enough to start. A quote request might require other elements. There's no point in multiplying fields just because they can be added: every piece of information must serve to process the request.

Next, let's look at where this data lives and who can modify it. If a contact already exists, how do we avoid creating a duplicate sheet? If a slot has just been taken, which information is the source of truth? If an external tool doesn't respond, where can the request be found and who receives the alert? These questions help build a workflow that can be understood and managed, rather than a series of invisible connections.

Start with a workflow you know well. Note the repeated actions, the necessary data, and the decision you want to keep. We can then choose what deserves to be integrated into the site, what stays in an existing tool, and what can be automated without adding unnecessary complexity.