Questions about AI and automation

A system that carries out a repetitive task from start to finish. The automation follows the steps. The AI handles what requires reading or understanding, like an email or an invoice.

An automation follows a path set in advance. An AI agent chooses its next steps to reach a goal, within set limits. Many tasks do not need an agent.

ChatGPT is often fine for writing or summarizing. It only knows what it is given, and it does not work in the business’s tools without being connected to them. Custom work is useful when a task has to get done on its own, with the business’s data.

Size matters less than repetition. A task that comes back every week and takes hours is worth looking at, even in a very small business.

Frequent tasks that follow the same rules: following up on quotes, sending invoices, answering missed calls, entering expenses, putting together a report. See the services.

What relies on judgment or on the relationship, like a delicate decision or an unhappy customer. AI can prepare the ground, and the decision stays with a person. A rare task does not justify the cost of an automation either.

A single place where the AI finds the business’s decisions, ways of doing things, customers and automations. That way, it starts from the business with every question. See the page and its guided tour.

The tools will keep improving, but a system’s strength comes from the context built up about the business, and that context is built over time. The earlier you start, the stronger the system is. A single process is enough to start.

According to the Institut de la statistique du Québec, 12.7% of businesses in Quebec used AI for production purposes in the 12 months before the second quarter of 2025. This figure covers production only.

The cost depends on the project: the work to build it, and what the system may cost each month to run. See what makes the cost vary.

Often, yes. It can be hosting, APIs, AI usage, text messages or a tool subscription. The amount varies with the project.

An API lets two pieces of software talk to each other. Some are included in a subscription. Others are billed by usage, for example for each question sent to an AI or each text message sent.

One-time needs and small integrations are billed by the hour: $150 plus taxes. A larger project gets a proposal.

The basic calculation compares what the task costs each month, in time and errors, to the cost of building and running the system.

Yes. The call is scheduled after the request has been read, when the project is suited to it.

It depends on the product used. Anthropic and OpenAI state that data sent through their APIs is not used for training by default, unless the customer agrees. Consumer versions have other rules: with ChatGPT, training is turned on by default and can be turned off in the settings.

Security depends on how the system is built, notably on what it can read and change. No system is without risk.

Law 25 does not prohibit automation. It applies as soon as the system touches personal information. The Commission d’accès à l’information asks, among other things, for a privacy impact assessment for an information system project that contains such information. This answer is not legal advice.

If the decision is based only on automated processing of their personal information, the law requires that the person be informed. They must also be able to submit their observations to someone who can review the decision. Act respecting the protection of personal information in the private sector, s. 12.1.

The law first requires a privacy impact assessment. The communication is possible if the assessment shows adequate protection, and it must be covered by a written agreement. Act respecting the protection of personal information in the private sector, s. 17.

Not in itself. The risk is not knowing what gets copied into it. A simple rule helps: which tools are allowed, and what never goes into them, like customer information.

Automation takes away repetitive tasks, and some disappear. The time freed up can go to relationships and judgment. The decision is up to the business.

No. The system must be usable without technical knowledge.

Talk about it early and start with a task the team does not like. The people who do the task know the exceptions, and a system built with them is more likely to be adopted.

Yes, enough to know what to check and who to call if there is a problem.

Yes. That is why what has to be exact, like a calculation, follows fixed rules, with checks. When a result is uncertain, the system can ask a person to confirm.

Only if it is set up that way.

The checks and security measures are adapted to the type of project, its complexity and its risks. See the process.

Any system can break, often because a connected tool has changed. Platform maintenance may be offered if the client needs it.

It happens, and the industry moves fast. When possible, the AI model remains a part that can be replaced without redoing everything.

With what takes the most time each week. Describe your need.

After the request, a free discovery call is scheduled if the project is suited to it. APC.lab then does a feasibility study and proposes an approach to the client. The work is then done with the client to reach their exact need. See the full process.

It depends on the project. A small integration is nothing like a complete piece of software.

Almost nothing: the task, who does it, the time it takes and the tools used. A real example, like a report or an invoice, also helps.

Not necessarily. Tools that can be connected are kept. What cannot be connected gets built.

Not necessarily, but it is part of the work and it can affect the cost.

An existing tool that solves the problem at a reasonable price is often the right choice. Custom work is useful when the ways of doing things do not fit what exists, or when several tools have to be linked together.

Antoine Provencher Chabot, founder of APC.lab, who has automated his own businesses. Learn more.

Depending on the terms of the contract, the code may belong to you or remain the property of APC.lab after launch, according to the needs and the conditions of the contract.

Platform maintenance may be offered if the client needs it.