On 21 Oct 2026 we spoke at an RBC evening for business owners at the Old House Hotel in Courtenay. The talk was called AI for Better Business Operations, and it opened with a question that had nothing to do with AI: what do you want from your business in the next three years?
The room split four ways. More margin, more capacity, better service, and more value when you sell. Every slide after that was tagged with the goal it served, and the last of the four was the reason RBC was in the room. Operations is where the money gets made. Wealth is what you do with it, and what the business is worth when you hand it over.
Most of what we showed that night is already on this site: how agentic orchestration works, how AI turns your words into meaning it can search, where the human checkpoint goes. This article covers what isn’t. It follows the money from the system that runs your work to the three places it shows up in your wealth.
Most businesses use AI. Few get paid for it.
Business use of AI in Canada went from 6.1% in 2024 to 19.2% in 2026, tripled in two years. Use is not the same as payoff. McKinsey’s 2026 State of AI survey, where about a third of respondents work at billion-dollar companies, found close to nine in ten organizations using AI, 37% seeing any profit impact from it, and 6% earning 5% or more of their profit from it.
What set the 6% apart was how the work flows. About three in four of them redesigned their workflows, against about one in four of everyone else. It shows up where an owner looks: BCG reported on 30 Sep 2026 that companies leading on AI grew revenue 2.4 times faster and EBITDA 2.8 times faster than the laggards.
Redesigned work, in practice, means a system that runs the job, checks itself, and gets better. Here is what that looked like on the night.
A system that catches its own mistake, then writes the rule
We built a demo trades company called Harbourline the way we build for clients. It’s fictional. Its agents and its records are real, 2,099 notes covering orders, clients, quotes, purchase orders, meetings, decisions and procedures.
In the walkthrough, a hotel calls at 7:42 on a Tuesday with no hot water. The orchestrator reads the call and hands out the work: scheduling, parts, invoicing, all at once. Scheduling books the nearest free tech. He’s a plumber, and the job is a gas boiler. A second agent that didn’t make the booking reviews it, rejects it, and rebooks the gas fitter in under two minutes. A person approves the part, because anything over the spending limit waits for one. The invoice goes out at 2:12 the same afternoon, and the office touched the job once.
Eleven seconds after the invoice, the dispatch agent wrote the mistake into the scheduling procedure as a rule: gas appliance work goes to a Class B gas fitter only. Then we ran the real agents on a new email, a different job for the same customer, with nobody scripting it. Dispatch read the rule it had written itself and ruled the plumber out without being told. The reviewer still stopped four of six outputs for a person, and people settled them in four decisions. The invoice, $2,235.45, was recomputed to the cent before it left.
That last paragraph is the part that matters for wealth. The system did the work, and it also left behind a written procedure your next hire, or your buyer, can read.
The same pattern at freight scale
C.H. Robinson, one of the largest freight brokers in North America, runs the pattern we drew on the iPad at a size no Courtenay business needs. In March 2026 the company described hundreds of connected agents working across the shipment lifecycle, orchestrated by one planner, and called that orchestration “the true competitive moat.”
The earlier numbers are specific. Its quoting agent returns a customer-specific price in 32 seconds. Its orders agent builds an order from an emailed tender in about 90 seconds and saves 600 hours of labour a day. Automated LTL orders went from 50% to over 75%.
Closer to home, one of our retail clients measured its own result: work that took three weeks now takes about ten minutes, 75% of the total staff time on that work gone. That client is where the investment numbers below come from.
Wealth, part one: cash in your pocket every year
Treat the system like any other investment, because that’s what it is. Compare what the work costs done by hand against what the system costs to build and run, year by year.
For that retail client, year one came in 40% cheaper than doing the same work by hand, with the build included. That figure is their actual, rounded down. Years two to five are our estimate, built on three assumptions: wages rise 3.5% a year, the Care plan rises 3%, and AI compute grows 15% a year as usage grows. Wage increases cost more than compute increases, so the gap widens each year and the yearly saving grows about 10% from year two to year five.
On those assumptions, the projected five-year saving is about 79%, and 87% in year five alone. In dollars, that’s roughly $61,000 for the system over five years against about $288,000 to keep doing it by hand. Your numbers will differ, and the method transfers.
Try it with your own job. Take the hours a week one task eats and multiply by what that person costs you an hour, wages and benefits, times 50 weeks. Ten hours at $40 is $20,000 a year and $100,000 over five. If your job behaved like our client’s, about $79,000 of that would stay in the business. That’s an illustration, and whether your job is the kind that pays back is arithmetic worth doing before anyone builds.
Wealth, part two: a business that runs without you
This is the one owners undercount. M&A advisors are blunt about it: founder dependence is a classic discount driver. If pricing rules, supplier history and the right person for each job live in your head, the buyer is paying for something that walks out the door.
The timing makes it urgent for a lot of the people who were in that room. A CFIB survey found 76% of Canadian small business owners plan to exit within a decade, with over $2 trillion in assets changing hands. Only about one in ten had a formal succession plan, and 43% struggled to measure what their business was worth. The succession specialists quoted in that release put the problem plainly: most owners aren’t ready on paper, and that’s why most deals break down in negotiation.
Documentation sits under all of it. When you sell, the first thing a buyer asks for is the documents. When you grow, a new hire learns from them. When you’re away, they’re what the business knows without you.
Most owners never get it written down because there’s no time to sit and write it. A system built this way writes it as the work happens. Every job is logged, every decision is recorded, and the gas-ticket rule went into the procedure without anyone stopping to type it. The record is plain text your team can read, and you own it. If you change AI providers, the knowledge comes with you.
Wealth, part three: your time back
The third return is the simplest. The hours the system takes off your plate go somewhere, either into the work that grows the business or into stepping further back from it. Both raise what the business is worth, because a business that grows without the owner’s hours is the same business a buyer wants.
A chat subscription helps one person with one question at a time: a person asks, checks and copies. The 6% got paid because they redesigned the work, and AI engineering is where that redesign gets built.
Before anyone builds, ask four questions
The talk ended on building it right, and the questions travel well. Ask where your data is stored, and separately, which region runs the model, because data stored in Canada can still be sent to a model abroad on every request (more on that). Ask which agents can touch what, and confirm that money and customers wait for a person. Ask whether the encryption can be swapped without a rebuild, since post-quantum standards are already published (more on that). And ask who owns the architecture end to end.
What RBC covered next
We handed the room to RBC’s team for the part we don’t do: what to do with the cash once it’s in your pocket, how to protect it, and how to plan the exit. That’s the right order. Operations puts the money in your hands, and the advisors who manage wealth decide where it goes.
Margin today, value at exit, and an AI system that does the work and writes the business down is how you get both.