Practical Guide / 2026
AI & SMEs
AI in your SME.
Before you begin
Setting the context
You may already be using artificial intelligence to write an email, prepare a presentation, analyze data, create content, respond to customers, or save time on certain tasks.
Your employees may also be using it, sometimes without a clearly defined framework.
Artificial intelligence can be a real asset for your business: improving productivity, simplifying tasks, supporting decision-making, developing new services, or allowing your teams to spend more time on high-value activities.
But its use raises very practical questions
- Which tasks can we entrust to AI?
- What information can we share with these tools?
- How do we choose a solution that fits our needs?
- How do we verify the reliability of the results?
- How do we regulate its use by our employees?
- How do we prepare our teams for the changes it brings?
- How do we know if AI is truly adding value to our business?
This guide was designed to help you answer these questions simply.
It does not ask you to become an AI expert. Instead, it offers you reflexes, questions to ask yourself, and practical tools to move forward gradually, in a responsible, secure, and SME-appropriate way.
The goal
To help you use AI where it can truly be useful, while protecting your business, your data, your employees, and your customers.
Table of contents
Understand. Choose. Act.
A document to read in order or to open directly based on your needs.
- 01 Setting the context Why this guide and the questions it answers
- 02 The approach Fieldwork, diagnostic tool, survey, and research
- 03 How to use this guide Five ways to move forward based on your needs
- 04 Before you start: five reflexes Need, data, verification, team, benefit
- 05 Taking action A first trial in six steps
- 06 Measuring the results Real time, quality, cost, and net gain
- 07 Protecting people and data Human resources and incidents
- 08 The checklist to complete 14 points, one choice, and one justification per point
For a first use
Read the five reflexes, choose a task, and fill out the checklist with the relevant people. Clarify any outstanding points and set a review date.
The approach
Guidelines based on fieldwork and research.
A shared framework to help AI improve the work and performance of SMEs, while respecting people and their rights.
The Practical AI & SME Charter is the result of a field-based approach aimed at better understanding the needs, concerns, and level of preparedness of SMEs regarding artificial intelligence.
This approach was based on the development of a diagnostic tool, on information gathering carried out notably during the HR Forum in Montreal, as well as on several exchange and awareness activities conducted directly with SMEs, particularly in Châteauguay and Quebec City, in collaboration with Capitale Numérique, and also online.
This field-based approach was reinforced by a targeted review of scientific and institutional publications, to cross-reference the realities observed in businesses with existing knowledge, recommendations, and reference frameworks. All this work made it possible to align the Charter’s principles with the concrete realities of SMEs and to make it a practical, accessible tool rooted in their day-to-day challenges.
A Quebec-based approach / three survey priorities
-
76 %
Risk cited
Data confidentiality
-
62 %
Barrier cited
Uncertainty about the right tools
-
52 %
Tool deemed useful
Short training
Note: multiple-choice; these percentages do not represent parts of a whole and do not add up.
What the literature provides
Studies show potential gains, which vary depending on the tasks and the experience of the people involved. They encourage measuring actual results. Guides recommend protecting data, defining responsibilities, and monitoring usage over time. These findings underpin the five reflexes proposed.
Methodological context
126 responses, from May 3 to August 26, 2026. Percentages rounded; multiple-choice. The charter also covers AI integrated into software. The examples focus mainly on generative AI, which creates text, images, or other content.
Start of the charter
Practical Charter for the Responsible and Productive Adoption of Artificial Intelligence in SMEs
How to use this charter?
This charter supports you at different stages of your AI adoption process.
- 01 Before you begin The five reflexes help you define your need, think about data, plan responsibilities, prepare your team, and determine the desired benefit.
- 02 Taking action The six steps guide you in preparing, framing, and testing a first use of AI.
- 03 Measuring and securing Benchmarks help you assess the real gain, manage risks, and protect people and data.
- 04 Validating your approach The practical checklist allows you to verify the main points to consider before continuing or deploying a use.
Core of the charter
Before you begin: five reflexes for your SME.
These five reflexes support you before choosing a tool, launching a first trial, or reviewing an existing use of AI. They help you start with a concrete need, protect your data, plan for human verification, prepare your team, and measure results.
Use them as a starting point to think about your project and identify the questions to clarify before taking action. These benchmarks also apply to AI functions already integrated into your software.
-
01
Which problem to solve.
Start with a specific task: reduce a delay, improve quality or service. Note the starting situation and check if a simpler solution suffices.
To avoid: Automating a process you do not understand.
-
02
What data to share.
Use an approved tool and only the necessary data. Check the provider’s terms. Personal and confidential information requires authorization and appropriate safeguards.
To avoid: Uploading a client list to an unapproved tool.
-
03
Who verifies and decides.
A convincing answer may be wrong. Assign a competent person, with time to verify facts, sources and biases, and the authority to refuse. Inform people about the role of AI in important decisions.
To avoid: Copying an unverified answer into a submission.
-
04
How to prepare the team.
Involve people in choosing uses and train them for their tasks. Allow time to learn and discuss difficulties. Maintain the skills needed to work without the tool.
To avoid: Letting everyone learn on their own, without common rules.
-
05
What benefit to obtain.
Compare total time, quality and costs for comparable tasks, with and without AI. Count the corrections. Reinvest the gains in service, innovation or team well-being. Reassess after a change.
To avoid: Rolling out a tool based solely on the impression of saving time.
Principle of proportionality
Adjust precautions to the consequences of an error. A reviewed draft allows for a simple test. A decision affecting rights, health, safety or employment requires a thorough review before any use.
This charter is a voluntary tool to adapt to your business. It does not replace your legal, regulatory and contractual obligations, particularly regarding the protection of personal information (Law 25).
Implementation
Take action.
A specific need, a limited trial, then an informed decision.
What you will do
Start with a concrete task, set the rules, test on a small scale, then measure the results before deciding.
-
01 Identify.
Choose a recurring task. Note how long it takes, the difficulties encountered and the result to improve.
-
02 Verify.
Compare solutions and costs. Set an expected gain, a quality level and unacceptable errors. Fill out the grid.
-
03 Frame.
In a sheet, note permitted uses, authorized data, the responsible person and controls. If working alone, you assume these roles.
-
04 Prepare the team.
Show how to give clear instructions to the AI, verify its answers and protect data. Discuss tasks that will evolve.
-
05 Test.
Try on common and difficult cases, for example over two to four weeks. Initially, validate before any sending, payment or modification by the AI.
-
06 Measure and decide.
In the end, continue if the benefits justify the costs and risks. Otherwise, adjust or stop. Suspend immediately in case of a serious incident.
Measuring the results
The real gain is measured after verification.
AI can produce a result quickly. But it is not the generation time that matters: it is the total time required to obtain a usable result of satisfactory quality.
Compare the same task, with and without AI, including preparation, verification and corrections.
Example · Responding to a customer request
Net gain: 12 minutes per task
30 − (10 + 8) = 12 minutes saved
That’s a 40% time saving, with at least equivalent quality.
If this task is performed 20 times a month, the gain represents roughly 4 hours of work per month.
Don’t just measure time
-
Time
How many minutes are actually saved, including verification?
-
Quality
How many corrections are needed? Is the result as good or better?
-
Cost
Does the gain justify the cost of the tool, training and oversight?
A result produced twice as fast is not a gain if you then have to spend as much time correcting it.
Heightened vigilance
Protect people and data.
AI and human resources: distinguish between preparing work and decisions that affect a person.
For preparing work
Draft a job offer, interview questions, a welcome document or training materials. Use fictional examples when individual data is not required.
For decisions that affect a person
Preselection, hiring, performance, promotion, compensation, discipline or termination. A qualified person reviews the context, identifies biases, makes the final decision and can explain and revisit it.
The principle of this charter
No major HR decision is left to AI alone. Inform people of its role and provide a human contact. Pay special attention to CVs, medical files and employee records.
The right reflex for data
Start with fictional or public non-personal information. Removing a name is not always enough to prevent identifying a person. Never enter passwords or access keys into the AI.
In case of serious error or leak
- Suspend use,
- immediately alert the responsible person
- and keep useful elements in a secure space.
They will assess corrections and notification obligations.
Practical grid / 1 sur 3
Before using AI.
A grid to fill out for each new use or major change.