
For a long time, a small business had to grow before it could afford the tools of a larger one. AI changes that order.
Imagine an eight-person service business. The team knows its clients well. Requests arrive by email, quotes sit in a shared folder, and most follow-ups live in the owner's head. It works, until the week the owner takes a vacation.
A familiar path follows: buy one tool for clients, another for quotes, and another for invoicing. Copy information between them. Later, once the business is big enough, put someone in charge of coordinating it all, then launch a project to connect the systems. Each step waits until the business has the size and budget to take it.
I think AI now makes it possible to skip some of those steps.
Skipping steps is nothing new
Before cloud computing, hosting your own system often meant buying a server, finding room for it, and paying someone to maintain it. Today, a business can start with online services.
This is called leapfrogging: using a new technology to skip stages that others had to work through one at a time.
For a small business, the stage you can skip is often the years spent copying information between disconnected tools while waiting to afford a way to connect them.
What makes the leap possible
Two things have changed.
First, AI now helps build the software itself. The team describes its work, a first version is built with AI's help, and the team tries it and asks for changes to anything that slows them down. A tool built around an eight-person team becomes an option worth considering.
You can keep the tools that work and build what's missing between them, without waiting for your business to double in size.
Second, AI can handle some of the recurring work: matching information from two sources, drafting a follow-up, or finding the right procedure. A person checks the result and decides what happens next.
An established small business starts with some advantages: clients, knowledge of its field, and experience. It also often has fewer layers of decision-making than a large company. The people dealing with a problem can speak directly to the person who makes the decision.
Small businesses already using AI report some of these benefits. In an OECD survey of more than 5,000 small and medium-sized businesses across seven countries, including Canada, 29% of those using generative AI said it had helped them compete with larger companies. Read the OECD report.
Ask the right question
You can use AI to do familiar work faster: summarize an email, write a proposal, or edit a document. That's useful. But the leap starts with a different question:
If we were organizing this work today, with today's tools, would we still do it this way?
Would we copy the same information three times? Would every follow-up go through the owner? Would we search five folders for a procedure?
What it can look like
Let's return to that eight-person business.
It keeps the tools that are useful and uses AI to help build a small CRM, a tool for keeping track of clients. It brings requests and quotes together in one record per client, shows who owns each file, and flags when a follow-up is due. The team tries a first version with test data. Employees explain what helps and what makes their work harder. The tool is adjusted before it handles real files.
Next, a feature prepares a summary of the exchanges or a draft follow-up, which the person responsible checks before sending.
The business hasn't waited for a major IT project to connect its client follow-up. Someone can take over when the owner is away, without digging through all the owner's emails.
I did it for Nord Paradigm
I'm not a programmer. Since May, I've had several tools built with Orion, the AI agent I use as my assistant: a CRM that brings together my business information, a knowledge base my assistants consult before working, a booking page connected to my calendars, billing monitoring, and follow-up preparation.
These tools build on Gmail, Notion, Stripe, and Google Calendar rather than replacing them.
The leap is real. There is still groundwork to do. My most useful corrections have been about rules: recognizing that two records refer to the same client, distinguishing a draft from a sent message, and making each invoice's status clear.
Skip steps without skipping the basics
Define the rules. Write down what the terms in your tracking system mean: a closed file, a completed follow-up, an active client. AI can write the code, but it can't make those decisions for you.
Keep people in charge of decisions. Decide what the tool can prepare and what a person must approve: a message, a payment, a commitment.
Check real cases. Compare the tool's results with what actually happened, not just with test scenarios.
Plan for maintenance. Someone needs to understand the tool, manage access and backups, and know what to do when something stops working. Compare it with existing software, too: sometimes that is enough.
Protect the data. Choose what the tool can access. In Québec, an information system project involving personal information may require a privacy impact assessment. The Commission d'accès à l'information explains when one is required (in French).
Where to start
Bring your team together around a recurring task that really gets in their way. Ask: if we were organizing it today, how would we do it? Then run a small first trial with results you can check.
Want to learn with your team? A custom AI workshop uses your team's tasks to practise and learn how to check the results.
Want support to move forward? AI Partnership helps you bring AI into your work, with regular follow-up.
