# How to automate a service business in Quebec: the complete guide

> Bookings, dispatch, follow-up, reminders, Law 25 and CASL: what a Quebec service business can automate, what should stay human, and where to start.

Updated on October 15, 2026 · https://app.pilotiko.com/en/guides/automate-a-service-business

Automating a service business means handing repetitive, predictable work to software: online booking, reminders, proposed dispatch, follow-up and reporting. A person stays in charge of anything that touches a customer, money or an outgoing message. Start by measuring the hours you lose today, then automate one task at a time.

## In short
- Best candidates: frequent, predictable tasks you can check easily, such as booking and confirmations.
- Keep human: decisions that commit you to a customer, a price, a refund or a message sent in your name.
- Quebec's Law 25 covers automated decisions, out-of-province vendors and privacy incidents; CASL covers commercial emails and texts.
- Start small: one task, one metric, four weeks of measurement, then the next task.

## What does automation actually mean for a service business?

Automation means letting software run any step that always follows the same rule, so people only handle the steps that need judgment.

In a service business, the value is created at the customer's place: the repair, the cleaning, the consultation, the inspection. Everything around that moment is coordination. Someone has to answer the request, find a slot, pick the right person, confirm, remind, record what was done, invoice and chase payment. Each step is easy on its own. The sheer number of them becomes a burden once you pass a handful of jobs a day.

Automating does not mean letting a machine decide. It means writing the rule once and letting it run on every request. A typical rule: send a Laval service call to the least busy technician in the north zone unless they are on leave. Software applies that rule just as well at 4:45 on a Friday as at 9 on a Monday. A person checks the result when something is at stake.

This guide uses three levels: full automation (software does it all), preparation (software drafts, a person approves) and assisted human work (the person decides, software supplies the data). Most of the gains come from the middle level.

## Why are so many small businesses starting now?

Because the tools now fit small teams, and the share of Canadian businesses planning to use AI has more than doubled in two years.

Statistics Canada released its analysis of the Canadian Survey on Business Conditions on October 1, 2026. In the third quarter of 2026, 25.2% of businesses in Canada planned to use AI to produce goods or deliver services over the next 12 months, up from 14.5% a year earlier and 10.6% in the third quarter of 2024 [1].

The jump is sharpest among mid-sized firms. For businesses with 20 to 99 employees, the share went from 15.0% to 32.8% in a single year [1]. Many of them are service companies juggling several crews on the road or in the office.

These figures describe plans, not results. What they really show is that the question has shifted from whether to automate to what, in which order and with which safeguards. That is what the rest of this guide covers.

| Group | Q3 2025 | Q3 2026 |
|---|---|---|
| All businesses | 14.5% | 25.2% |
| 20 to 99 employees | 15.0% | 32.8% |
| 100 employees or more | 20.5% | 37.5% |

Source: Statistics Canada, Canadian Survey on Business Conditions, released October 1, 2026 [1].

## Which tasks automate well, and which should stay human?

A task automates well when it is frequent, follows a clear rule and is easy to check. It should stay human when a mistake is costly for the customer or when it calls for empathy.

Use the table below as a starting point and drop your own tasks into it. If a task is frequent and rule-based, it is a good candidate. If it involves money, a complaint or lots of exceptions, keep a person in the loop.

| Task | Suggested level | Why |
|---|---|---|
| Showing open slots and taking a booking | Automate | Clear rule (calendar, duration, zone), very frequent |
| Booking confirmation and reminder | Automate | Transactional message, same template every time |
| Choosing the technician for a call | Prepare, then approve | Rules help, but a person knows the exceptions |
| Drafting a quote | Prepare, then approve | Commits to a price, so a person reviews it |
| Handling a complaint | Keep human | Needs judgment and empathy |
| Granting a refund or credit | Keep human | Involves the customer's money |
| Sending a promotional newsletter | Prepare, then approve | CASL applies: consent and unsubscribe |
| Producing the weekly report | Automate | Data is already in the system, nothing goes out |

Indicative grid suggested by Pilotiko. Adapt it to your trade and your appetite for risk.

## How do you automate booking?

Publish your real open slots online, calculated from calendars, service durations and travel time, so customers can book without waiting for a call back.

Booking is the most worthwhile task to automate because it happens dozens of times a week and everything else waits on it. A customer who has to leave a voicemail and wait for a call back is shopping around in the meantime.

Three things make online booking reliable. First, realistic durations per service type, travel included. Second, calendars that are actually up to date: time off, training, existing appointments. Third, a conflict rule that blocks two bookings in the same slot. Miss any of these and the software offers slots that do not exist, and you spend your days calling people to move their appointment.

In Pilotiko, online booking runs on your team's real open slots and avoids conflicts. See Bookings and scheduling.

## How do you automate technician dispatch?

Let the software propose the right person based on each calendar and workload, then have whoever runs dispatch confirm the proposal.

Dispatch is the heart of a service business, and it is where unwritten rules pile up. One technician does not do roofs, another has the only equipped van, a customer asked for the same person as last time. Software only knows those rules if you write them down.

The right approach is the proposal. The software suggests someone based on calendars and workload, and the dispatcher approves or corrects it in one click. Over time, the corrections show you which rules to add. Pilotiko proposes assignments from calendars and workload; route optimisation does not exist yet.

> Coming soon in Pilotiko: route optimisation and service demand forecasting. Today, proposed assignments account for calendars and workload, not for the best order of stops.

## How do reminders cut down on no-shows?

Send a confirmation at booking and one or two reminders before the visit, with an easy way to reschedule. Text reminders have a measured effect, mostly documented in health care.

The best public evidence comes from health care. A 2013 Cochrane review of 8 randomized trials with 6,615 participants found attendance of 78.6% with text reminders versus 67.8% with no reminder [6]. The authors rate the evidence as low to moderate quality, and a clinic is not a plumbing company. The order of magnitude still helps you estimate your own gain.

The guide [Reduce no-shows](https://app.pilotiko.com/en/guides/reduce-no-shows) walks through the cost of a missed visit and the right timing for each reminder. In Pilotiko, the confirmation and the 24-hour and 1-hour reminders are **being commissioned: nothing is sent yet**, because the sending provider is not connected.

## What should you automate in post-visit follow-up?

Visit notes, customer record updates and follow-ups on unanswered quotes. These slip through because they land when the crew is already on the next call.

Follow-up is where small businesses lose the most money without noticing: a quote sent and never chased, a warranty never recorded, a yearly maintenance visit nobody books. The useful automation here is mostly a list. The software knows what is pending and puts it in front of the right person at the right time.

Pilotiko's virtual office shows urgent items in the right place: each department is a room, and whatever needs a decision shows up there. AI agents can draft a follow-up or a visit summary from the real data in the file. Sending it to the customer always waits for a person's approval.

## Why keep human approval on AI agents?

Because an AI agent can be confidently wrong. Requiring a person to approve anything that touches a customer, money or an outgoing message limits the cost of a mistake to a few seconds of review.

An AI agent is software that reads your data and prepares an action: a reply, a quote, a proposed schedule. It is fast and never gets tired. It can also invent a plausible detail or misread an exception. The simplest rule is to rank actions by risk and require approval for the sensitive ones.

In Canada, the voluntary code of conduct on generative AI published by Innovation, Science and Economic Development Canada in September 2023 includes a human oversight and monitoring principle: system use is monitored after deployment [7]. In Quebec, section 12.1 of the private-sector privacy act sets specific duties when a decision is based exclusively on automated processing [2].

The guide [Human approval for AI agents](https://app.pilotiko.com/en/guides/human-approval-for-ai-agents) offers a grid by type of action. In Pilotiko it is the baseline rule for AI agents: they prepare, a person approves.

## What does CASL say about automated emails and texts?

CASL bans sending a commercial electronic message without express or implied consent. Messages confirming a transaction the customer already agreed to are exempt from consent but must still identify the sender.

Canada's Anti-Spam Legislation (CASL) covers emails and text messages alike. Section 6 prohibits sending a commercial electronic message without consent and requires every message to identify the sender, give contact information and include an unsubscribe mechanism [4].

Subsection 6(6) lists exceptions to consent. A message that confirms or completes a commercial transaction the recipient previously agreed to, or that provides a quote or estimate they asked for, does not need prior consent [4]. A booking confirmation generally fits that logic. A promotion slipped into the same message does not.

For promotional messages, implied consent exists in some cases: a purchase within the previous two years or an inquiry within the previous six months [5]. An unsubscribe request must be honoured without delay, and no later than 10 business days after it is sent [5].

> Coming soon in Pilotiko: email marketing and automated sequences. Consent and unsubscribe handling are already built; real sending is not. See [Email marketing](https://app.pilotiko.com/en/product/email-marketing).

## Which parts of Law 25 affect automation projects?

Four in particular: a privacy impact assessment before adopting a system, another before sending data outside Quebec, informing people about automated decisions, and handling privacy incidents.

Quebec's Act respecting the protection of personal information in the private sector, as amended by Law 25, applies to any business that holds information about its customers [2]. For an automation project, keep these points in mind.

- **Section 3.3**: a privacy impact assessment (PIA) for any project to acquire, develop or overhaul an information system involving personal information [2].
- **Section 17**: a PIA and a written agreement before communicating personal information outside Quebec, for instance to a cloud vendor [2].
- **Section 12.1**: tell people when a decision about them is based exclusively on automated processing, and let them submit observations to someone who can review it [2].
- **Section 3.5**: if a confidentiality incident presents a risk of serious injury, notify the Commission d'accès à l'information and the people concerned [2].
- **Portability**: since September 22, 2024, provide computerized personal information on request in a structured, commonly used format [3].

## What data should you clean up before automating?

A service list with durations, current calendars, service areas, clean customer records and written dispatch rules. Software only automates well what is already clear.

Automation is only as good as the data behind it. Before connecting anything, spend half a day on an inventory. How many service types do you offer? What is the real duration of each one, travel included? Who can do what? Which towns or neighbourhoods do you cover?

Use the exercise to do the clean-up the law already expects. Section 23 of the private-sector act requires you to destroy or anonymize personal information once the purposes for collecting it are achieved, subject to any legal retention period [2]. A ten-year-old customer file with card numbers typed into the notes is not a healthy base for an AI agent.

## How much time can automation free up?

It depends on volume and on the office time each booking takes today. The calculation below shows the method with assumptions; swap in your own numbers.

**Example: a hypothetical business with 6 technicians**

Assumptions: 6 technicians, 4 visits a day each, 5 days a week. Today every booking takes 8 minutes of office work (taking the call, data entry, picking the technician, confirming). With online booking, 40% of customers book themselves and the proposed assignment takes 2 minutes to approve.

- Bookings per week (6 × 4 × 5) : 120
- Current office time (120 × 8 min) : 960 min, or 16 h
- Online bookings (40% of 120) : 48
- Approving those 48 bookings (48 × 2 min) : 96 min
- Other 72 bookings handled as before (72 × 8 min) : 576 min
- New office time (96 + 576) : 672 min, or 11.2 h
- **= About 4.8 office hours freed per week under these assumptions**

## Where should you start automating?

With one frequent, measurable task, usually booking or confirmations. Measure before, automate, measure after, then move on to the next one.

1. **Measure a normal week** — For five days, count calls, bookings, reschedules and no-shows. Without a baseline you cannot show any gain.
2. **Pick a single task** — Choose the most frequent task that follows a clear rule. For most service businesses that is booking.
3. **Write the rule down** — Describe on one page how the task is done today, exceptions included. That page is what you use to configure the software.
4. **Run the privacy assessment** — Before adopting the tool, document what data it handles, where it is hosted and who can access it, as section 3.3 of the private-sector act requires.
5. **Decide what needs approval** — List what the software may do on its own and what waits for a person: every outgoing message, every price, every refund.
6. **Run it for four weeks** — Keep the old method running in parallel at first, then compare against the metrics from step 1.
7. **Move to the next task** — Once the first task is stable, add the next one from the grid. Gains stack up without adding chaos.

## How do you pick an automation tool?

Compare tools on your real tasks, data quality, hosting, portability and human control, rather than on the length of the feature list.

A slick demo proves very little. Ask to see your typical week inside the tool: your services, your zones, your exceptions. Check where data is hosted, how to export it and who can approve what. The guide [Choosing business management software](https://app.pilotiko.com/en/guides/choose-business-management-software) offers a weighted scoring grid.

Also compare total cost against the alternatives: agency, hiring or software. The guide [What marketing and sales cost a small business](https://app.pilotiko.com/en/guides/marketing-and-sales-cost-smb) explains the method, and the [24/7 coverage calculator](https://app.pilotiko.com/en/tools/staff-a-position-24-7) shows why round-the-clock staffing gets expensive.

## What does Pilotiko do today, and what is still to come?

Pilotiko handles online booking, proposes dispatch, produces reports and has AI agents prepare work with human approval. Reminders are being commissioned; the phone receptionist and email marketing are still to come.

| Capability | Status |
|---|---|
| Online booking, real open slots, conflicts avoided | Available |
| Assignment proposed from calendars and workload | Available |
| AI agents that prepare work, with human approval | Available |
| Virtual office, dashboards and internal reports | Available |
| Email and SMS confirmations and reminders | Being commissioned: nothing is sent yet |
| AI phone receptionist | Coming soon |
| Email marketing and sequences | Coming soon |
| Route optimisation, demand forecasting | Coming soon |

Statuses come from Pilotiko's capability register. Internal reports cover bookings, revenue and workload.

> Coming soon: the [AI receptionist](https://app.pilotiko.com/en/product/ai-receptionist) to answer calls, qualify leads and book appointments. The code exists but has not been verified with a real phone number yet. Meanwhile, size up what unanswered calls cost you with the [missed-call tool](https://app.pilotiko.com/en/tools/missed-call-cost).

## What mistakes should you avoid when automating?

Automating a fuzzy process, launching everything at once, letting messages go out unreviewed and forgetting to measure. Each one wipes out part of the gain.

- **Automating the mess.** If two people dispatch calls with two different logics, the software will just reproduce the confusion faster.
- **Changing everything at once.** A team absorbs one change at a time. Three new tools in the same month is a reliable way back to paper.
- **Letting messages go out without approval.** One wrong text sent to 300 customers costs far more than a minute of review.
- **Forgetting legal duties.** Privacy assessments, unsubscribe handling, notice of automated decisions: see [Law 25 for service businesses](https://app.pilotiko.com/en/guides/quebec-law-25-for-service-businesses).
- **Not measuring.** Without before-and-after numbers you cannot tell whether a tool is worth its price. To compare offers, see [Pilotiko pricing](https://app.pilotiko.com/en/pricing).

## Frequently asked questions

### Which task should a small service business automate first?

Start with the most frequent task that follows a clear rule, usually booking. It happens dozens of times a week and its effect is easy to measure. Once it is stable, move on to reminders and then follow-up.

### Will automation cost my employees their jobs?

In a service business, automation mostly removes repetitive coordination work. People keep the tasks that need judgment, customer contact and trade skills. Pilotiko talks about equivalent capacity, not eliminated positions.

### Can I let an AI agent answer customers on its own?

It is not advisable for anything that commits you to a price, a refund or a promise. An agent can draft the reply and a person approves it in seconds. In Pilotiko, anything touching a customer, money or an outgoing message waits for that approval.

### Do I need consent to send a customer an appointment reminder?

Under subsection 6(6) of CASL, a message that confirms or completes a transaction the customer already agreed to does not need prior consent. It must still identify the sender and give contact information. Adding a promotion to the reminder would take it out of that exception.

### Does Law 25 ban automated decisions?

No. Section 12.1 requires you to tell people when a decision is based exclusively on automated processing and to let them reach someone who can review it. Systematic human approval means, in practice, that the decision is not exclusively automated.

### My software is hosted in the United States. Is that a problem?

It is not prohibited, but section 17 of the private-sector act requires a privacy impact assessment and a written agreement before communicating personal information outside Quebec. Ask the vendor where the data sits and which protections apply.

### How long does it take to automate a first task?

Plan on one week of measurement, a few days of setup and four weeks of break-in for a task like booking. The slowest part is usually writing down the rules your team was applying without ever naming them.

### Does Pilotiko already send text reminders?

Not yet. The code for confirmations and the 24-hour and 1-hour reminders is ready, but the feature is being commissioned: nothing is sent until the sending provider is connected.

### Is automation worth it for a three-person business?

It can be if booking volume is high or if the owner answers the phone in the middle of jobs. Run the numbers with your real time per booking. Below a few bookings a week, a shared calendar is often enough.

## Sources

1. Statistics Canada, « Analysis on planned use of artificial intelligence by businesses in Canada, third quarter of 2026 », https://www150.statcan.gc.ca/n1/pub/11-621-m/11-621-m2026013-eng.htm (accessed on 2026-10-15)
2. LégisQuébec, « Act respecting the protection of personal information in the private sector (CQLR, c. P-39.1) », https://www.legisquebec.gouv.qc.ca/en/document/cs/P-39.1 (accessed on 2026-10-15)
3. Commission d'accès à l'information du Québec, « Principaux changements apportés par la Loi 25 », https://www.cai.gouv.qc.ca/protection-renseignements-personnels/sujets-et-domaines-dinteret/principaux-changements-loi-25 (accessed on 2026-10-15)
4. Justice Laws (Justice Canada), « Canada's Anti-Spam Legislation (S.C. 2010, c. 23), sections 1 to 9 », https://laws-lois.justice.gc.ca/eng/acts/E-1.6/page-1.html (accessed on 2026-10-15)
5. Justice Laws (Justice Canada), « Canada's Anti-Spam Legislation, sections 10 and 11 (implied consent, unsubscribe) », https://laws-lois.justice.gc.ca/eng/acts/E-1.6/page-2.html (accessed on 2026-10-15)
6. Cochrane, « Mobile phone messaging reminders for attendance at healthcare appointments (CD007458, 2013) », https://www.cochrane.org/evidence/CD007458_mobile-phone-messaging-reminders-attendance-healthcare-appointments (accessed on 2026-10-15)
7. Innovation, Science and Economic Development Canada, « Voluntary Code of Conduct on the Responsible Development and Management of Advanced Generative AI Systems (septembre 2023) », https://ised-isde.canada.ca/site/ised/en/voluntary-code-conduct-responsible-development-and-management-advanced-generative-ai-systems (accessed on 2026-10-15)