The Knife and the Cook: Adopting AI Without Losing Control

A practical guide for leaders of small and medium-sized businesses and nonprofits.

We used it before we adopted it.

In many small and medium-sized businesses, artificial intelligence arrived quietly, by being useful. An email that needed rewriting. A meeting that needed summarising. A proposal due before the end of the day. A long document handed to the machine, with the faint hope that it might have more patience than we do.

The shift has been fast. According to Statistics Canada, 12.2% of Canadian businesses reported using AI to produce goods or deliver services in the second quarter of 2025, double the 6.1% recorded one year earlier. A Microsoft Canada study published the same year estimates that 71% of Canadian small and medium-sized businesses already use AI tools in at least some part of their operations.

The distinction matters. A company can believe it has no AI strategy while its teams have already built their own habits, because work is urgent, days are full, and a tool that answers quickly finds a place on the desk.

People are using AI. The real issue is that they often use it without rules, without training, without much thought about the data they feed it, and with the slightly excessive trust we tend to place in sentences that sound well written.

This has become institutional common sense. Development banks and business advisors now repeat the same two-sided message: AI can automate repetitive work and support decisions, while carrying real risks, chief among them the confidentiality of the data we feed it and the reliability of what it hands back. Both warnings are correct, and neither is new anymore.

Why AI is already in your business, with or without a strategy

In 2026, the question has shifted. AI is no longer something to discover. It is already in the tools, the software, the habits, and sometimes in the blind spots of the business.

The question now is how to use it without losing control.

What separates AI use from professional AI adoption

AI is a professional tool. In a kitchen, a sharp knife changes the pace of the work. It still leaves everything to the cook. You need to know how to use it, follow hygiene rules, check what comes out, train the people doing the work, and never confuse the speed of a gesture with the quality of a dish.

AI can help write, summarise, compare, prepare. It can save time and, sometimes, improve quality. On its own, it remains blind to what is confidential, what is accurate, and what deserves to be sent to a client.

For that, you still need a method, rules, and trained people.

Where to start: the four questions before buying any AI tool

This is where many businesses risk getting it wrong. They sometimes think the AI question begins with buying a tool. It begins earlier, in choosing the right uses, defining the limits, and training the teams.

The useful questions are plain ones. What are we trying to solve. What data should never go into a public tool. Which outputs need checking before they leave the building. Who signs off on the final document.

These questions seem modest next to an impressive demonstration, and they are worth far more.

If you want a structured way to answer those four questions for your own organisation, MHA Studio AI offers a focused AI briefing for executive teams and boards designed for SMEs and nonprofits.

How to train employees to use AI responsibly

Training employees deserves real attention. You do not hand someone a professional knife and simply tell them to cut faster. You teach the gesture, the risk, the precision, the moment to slow down, and the moment to stop.

Training people in AI goes well beyond handing out three prompts as if they were magic recipes. It means learning to ask a clear question, to protect sensitive information, and to tell a draft from a document ready to circulate.

It is less dazzling than a big announcement. It is where AI use becomes professional.

Where AI delivers real value first in SMEs and nonprofits

In SMEs, the first gains are rarely hidden in big projects. They are often found in the small tasks that eat away at the day: meeting notes, sales follow-ups, service proposals, procedures, summaries, emails postponed because the right tone has not yet shown up.

AI can help precisely there, in those modest places where a great deal of time is lost, provided we ask it only for what it can give.

It can prepare a first draft. It can bring order to scattered notes. It can give a team a little breathing room in an overfilled day.

The company's strategy, the relationship with the client, the nuances of a file, the responsibilities that come with a decision: all of this stays out of its reach.

It is the tool on the worktop. You still need to know what you are preparing, for whom, with what standards, and with what level of acceptable risk.

Experiment or method: building real organisational capability

A company that uses AI without a plan accumulates experiments. The proposal drafted in a hurry, the summary no one checked, the clever prompt that worked once and was never written down. Some of it will prove useful. Some will be forgotten. A few will create risks no one chose.

A company that uses AI with method builds something else: a collective capability. It teaches its teams to work faster without neglecting quality, to produce more efficiently without exposing data, to use powerful tools without surrendering judgment to them.

The Business Development Bank of Canada estimates that the 30% of SMEs already using AI in 2025 were on average 24% more productive than those who were not. The gap widens every quarter. Capability, not curiosity, is what drives that gap.

This is the quiet part of AI, and probably the most important.

Because in the real life of businesses, value is not measured by the number of tools tested. It is measured by what actually changes in daily work: an hour recovered, a clearer document, a team that needs less supervision.

AI increases the capacity of those who know how to direct it. The craft itself remains intact.

As in a kitchen, the quality of the result never depends only on the tool. It depends on the person using it, the gesture they have mastered, and the judgment they keep.

Frequently asked questions: Adopting AI in an SME or nonprofit without losing control

Does every company or nonprofit need to use AI?

Not everywhere, and not for everything. Every SME and nonprofit should understand where AI is already entering its operations. In many organisations, employees are already using AI tools to write, summarise, translate, prepare documents, analyse information, or save time on routine tasks. The real question concerns whether these existing uses are known, useful, and properly managed and adopting AI without losing control.

Where should an SME or nonprofit start with AI?

Start with repetitive, time-consuming, low-risk tasks: meeting notes, document summaries, email drafts, internal procedures, first versions of content, market monitoring, information organisation, or preparation for meetings. The right starting point is the problem to solve, not the tool itself.

Will AI replace employees in small and medium-sized businesses?

In most SMEs, AI primarily supports work rather than replacing it. According to Statistics Canada, 89.4% of Canadian businesses using AI reported no change to employment levels after implementation. The stronger approach uses AI to help people save time, clarify documents, reduce repetitive work, prepare decisions, and structure information, while increasing the capacity and quality of their work.

What are the main risks of using AI in a small business?

The most immediate risks are data confidentiality, undetected errors, fabricated or approximate answers, generic content, copyright issues, bias, dependency on external tools, and unclear responsibility. One of the most underestimated risks is invisible use: employees using AI without shared rules, without validation, and without management knowing what data is being shared.

Is it safe for employees to use ChatGPT, Claude, or Copilot with company data?

Yes, with limits and clear rules. A company should distinguish between public, internal, confidential, personal, and strategic data. It should define what can be entered into an AI tool, what must never be entered, and which uses require a professional, secured, or company-approved environment.

Does an SME or nonprofit need a written AI policy from the start?

Yes, and it can be simple. An SME does not need a heavy twenty-page policy. Clear rules are enough: authorised tools, prohibited data, mandatory human review, acceptable uses, responsibility for final outputs, treatment of client information, and document retention. A good AI policy should be understood and used, not merely stored in a folder.

How do we know if an AI use case is relevant?

A good AI use case should meet three conditions: it solves a real operational problem, it carries a manageable level of risk, and the gain can be observed. That gain may be time saved, improved quality, better consistency, fewer errors, or stronger analytical capacity. If no one can clearly explain what problem AI is meant to solve, the use case is probably not ready.

Do employees really need AI training?

Yes. Training people in AI goes beyond giving them a list of prompts. They need to understand the limits of the tools, how to frame a request, how to check an answer, how to protect data, how to recognise weak output, how to adapt the result to the organisation's context, and when not to use AI at all.

How do we avoid generic AI-generated content?

The tool needs clear direction: target audience, purpose, tone, constraints, examples, level of detail expected, and elements to avoid. The result also needs to be edited. Good AI use requires directing, correcting, challenging, and finalising the work, never simply accepting the first answer.

Is AI mainly a technology issue?

No. AI is a management issue. It touches processes, data, roles, quality, training, confidentiality, client relationships, and decision-making. The technology matters, while the real challenge is integrating it into daily work without losing control.

How should we measure the benefits of AI in an organisation?

Use simple indicators: time saved, fewer revisions, document quality, response speed, consistency of deliverables, team satisfaction, and number of tasks standardised. Without measurement, AI remains a feeling of productivity. With measurement, it becomes a management tool.

What is the right approach for an SME or nonprofit?

Start small, with intention. Identify current uses. Choose two or three concrete use cases. Train the team. Set rules for data and verification. Test. Measure. Adjust. Then expand if the results are there. The goal is targeted use, in places where AI genuinely improves work without weakening judgment, quality, or trust.

──────────────────────────────────────────────────

Marie Horodecki-Aymes, Adm.A., is the founder and CEO of MHA Insights Inc., a Montréal-based consulting firm helping SMEs, nonprofits and boards adopt AI responsibly through its MHA Studio AI practice. She is an Administratrice agréée (Ordre des Adm.A. du Québec) and a registered ÉcoLeader Expert with FAQDD.

To explore an AI briefing for your executive team or board, or a structured AI adoption mandate, contact MHA Studio AI at www.mhainsights.com.

Sources cited in this article

·         Statistics Canada, Analysis on artificial intelligence use by businesses in Canada, second quarter of 2025.

·         Microsoft Canada, Majority of Canadian Small and Medium-Sized Businesses Embrace AI, June 2025.

·         Business Development Bank of Canada (BDC), LIFT initiative on SME AI adoption, 2025.