Learn With Nathan
Sign in
Full sample chapter · AI Foundations
What AI Can Do for a Small Business Today
A practical way to find useful AI work without buying technology first or gambling with customer trust.
Chapter 02
What AI Can Do for a Small Business Today
A practical way to find useful AI work without buying technology first or gambling with customer trust.
The owner of a growing service business does not wake up wanting “an AI strategy.” She wants the unanswered enquiries cleared, the weekly numbers understood, the proposal finished, and enough time left to speak with customers. AI becomes useful when it shortens one of those paths without creating a larger risk somewhere else.
Small businesses rarely have spare people, clean data, or months for experimentation. That constraint can be an advantage. It forces the question that large AI programmes sometimes avoid: which specific job will improve, for whom, and how will we know?
This chapter gives you a use-case filter. You will learn where today’s tools are genuinely helpful, which tasks should remain human-led, and how to run a small pilot before you buy, integrate, or automate anything.
Start with friction, not technology
Find the work before choosing the tool
Begin with a repeated moment of friction: copying the same information between systems, turning notes into follow-up, searching several documents for one answer, or producing the first version of familiar content. That moment is a potential .
A useful use case names five things: the person doing the work, the starting input, the desired output, the decision that follows, and the cost of a mistake. “Use AI for marketing” is not a use case. “Help the owner turn an approved product brief into three draft social posts, each reviewed before publishing” is.
Canadian business adoption is growing, but adoption is not universal and relevance varies by industry. Statistics Canada reported that 19.2% of businesses used AI to produce goods or deliver services in the twelve months preceding its second-quarter 2026 survey, while 40% said AI was not relevant to their business. The lesson is not “everyone must adopt.” It is “test relevance against actual work.”
Four jobs AI can help with today
Draft, transform, extract, and explore
Most useful beginner applications fall into four roles. They are broad enough to matter across industries and narrow enough to review.
Draft
Create a starting version of emails, proposals, job aids, product descriptions, agendas, and follow-up messages.
Transform
Shorten, translate, restructure, change tone, or turn unstructured notes into a checklist or table.
Extract
Pull dates, topics, risks, questions, or action items from information you are permitted to use.
These roles make AI a : it accelerates parts of the work while a person retains the goal, context, judgment, and responsibility. This is different from handing an entire business process to an autonomous system.
Climb the use-case ladder deliberately
Value and control should rise together
Not every use case belongs on the same first day. As an AI system gets closer to customers, sensitive data, important decisions, or direct action, the required controls rise. Start low on the ladder and earn your way upward with evidence.
The small-business AI use-case ladder
Move upward only when evidence and controls are ready
Audiobook description
Four steps rise from lower left to upper right. Step one is a private draft that is easy for one person to review. Step two is team assistance with shared practices. Step three is a governed workflow with permissions, rules, logs, and monitoring. Step four is direct customer-facing or operational action with approval gates and recovery plans. An arrow beneath the steps shows that consequence and required control rise together.
The OECD reports that AI adoption remains lower among small and medium-sized enterprises than among large firms, and highlights skills, data, finance, and infrastructure as recurring constraints. The ladder is a way to work with those constraints instead of pretending they do not exist.
Choose the first use case with four filters
Frequency, time, reviewability, and consequence
Score each candidate task from one to five on four questions. How often does it happen? How much time does it consume? How quickly can a knowledgeable person review the output? How serious would an error be?
| Filter | Prefer | Avoid first |
|---|---|---|
| Frequency | Weekly or daily work with a stable pattern. | A rare exception you do not understand well. |
| Time | A clear bottleneck with a measurable baseline. | A task already completed efficiently. |
| Reviewability | An output an experienced person can check quickly. | An answer that requires unavailable expertise to verify. |
| Consequence | A reversible draft or internal aid. | A legal, medical, financial, employment, or safety decision. |
The strongest first candidate is frequent, time-consuming, easy to review, and low consequence. It may not be the most exciting idea. That is precisely why it is a good place to learn.
Run a small pilot and measure the whole result
Time saved is only one part of value
A pilot compares the existing process with an AI-assisted version. Record the original time, quality, error rate, waiting time, and frustration. Then repeat the same task with a clear instruction, approved information, and named reviewer.
Measure total effort—not just generation time. If the tool creates a draft in one minute but requires twenty minutes of correction, the business result is not a one-minute task. Include setup, checking, rework, subscription cost, and any new risk.
Privacy belongs in the use-case decision, not in a policy added later. Canada’s privacy authority advises businesses to limit personal, sensitive, and confidential information, provide appropriate transparency, and build privacy safeguards into AI use. See for the operating rule.
Pause and apply
Reflection questions
- Which repeated task in your business has the clearest input and most reviewable output?
- Where would a mistake be inexpensive and reversible—and where would it affect a person’s rights, money, safety, or trust?
- What baseline will let you decide whether a five-example pilot actually improved the work?
Sources and further reading
Chapter 2 endnotes
- Statistics Canada. Analysis on Artificial Intelligence Use by Businesses in Canada, Second Quarter of 2026, 2026. Official report.
- OECD. AI Adoption by Small and Medium-Sized Enterprises: OECD Discussion Paper for the G7, 2025. Official OECD report.
- Office of the Privacy Commissioner of Canada. AI, Privacy, and Your Business, 2025. Official guidance.