The constraint every company has lived with
In 1937 the economist Ronald Coase asked why firms exist at all. His answer: coordinating people inside a company is cheaper than coordinating them through a market (The Nature of the Firm). Every company since has been shaped by that cost. Hiring, training, managing, handing work from one desk to the next: that is the price of thinking, paid in people.
AI is collapsing that price. A system that can read, write, call, decide and follow up, at any hour and for cents, changes what a company of ten, fifty or five hundred people can do.
When the cost of coordination falls this far, the boundaries of the firm move. Some companies will become far larger than their headcount. Others will find that most of what they do no longer needs doing the old way.
The basic math
Take an owner-led company of 50 people. Suppose a fifth of its hours go to work software can now do: chasing leads, re-keying data, writing the same quote again. That is ten people’s capacity locked into tasks instead of growth (illustrative). Free it, and the company grows without hiring, or prices in a way its competitors can’t follow.
Small businesses make up about 98% of employer businesses in Canada (ISED, Key Small Business Statistics), and most of them are owner-led. Across all of them, this shift is as large as the move to the internet. It just arrives one bottleneck at a time.
Three levels
Everyone can rent the same models. What separates companies is what they keep.
- Level 1, AI-assisted. People use rented models. Nothing persists, so the company improves exactly as fast as everyone else’s tools do. Most companies are here.
- Level 2, AI-operated. The workflows at the bottleneck run on AI, and the business produces more. But if the prompts are the only asset, every model change resets the gain.
- Level 3, AI Forward. Every important decision is recorded with the facts it used, what a person overrode and how it turned out. Those records are tested and fed back into how the work is done, in systems the company controls. The company re-finds its moving bottleneck each cycle, and it gets better with every model release.
Level 2 feels ahead until the next model ships. The gap between Level 1 and Level 3 widens with every release, and it will be the defining competitive gap of the next decade. (Which level is your company?)
Start at the bottleneck
A company moves at the speed of its slowest step. AI applied anywhere else trims costs at best; it doesn’t make the business produce more. So we never start with the technology. We find where the company leaks time and money and put AI there first. In owner-led companies that is usually revenue: the lead nobody called back, the follow-up nobody sent.
Then the bottleneck moves. Once AI clears one, the next shows up where AI can’t reach yet: judgment, checking the work, trust, the data it needs. So the job isn’t finding the bottleneck once. It’s re-finding it every time the models change, and the decision ledger is what makes that fast.
The flywheel
Advice is cheap; operating is not. So we put our own companies first.
- Build AI Forward companies from day one. Shake co-founded Canadian Funding Partners, an independent company, AI-first from the start: an AI agent works the sales floor alongside the human team. Running it is our forcing function. Every system has to work in production, with real customers, before we recommend it to anyone.
- Take what works into other companies. Find the bottleneck, fix the process, put AI on it, and stay accountable for the result.
- Publish what we learn, so the playbook compounds.
- Acquire owner-led businesses that can’t make the jump alone, and run the playbook as owners. That comes only after the playbook is proven in several companies. Manufacturing and healthcare come after that.
Each step funds the next and makes it better.
A prediction
This is a prediction, and we will score it. Within five years, in most industries, the lowest-cost and fastest-growing operator will be an AI Forward company. Many of them don’t exist yet. We intend to build some of them.
The responsibility
Moving fast with AI is easy. Doing it inside the rules, with people on the judgment calls and customers who trust what they hear, is the actual work. Everything we ship is designed to run within the law (CASL, the CRTC’s telemarketing rules, privacy law), logs its decisions, and has a named person accountable for its overrides.
Who is building this
Shake Dewan started in applied AI as a data scientist at Statistics Canada. He was national lead of Deloitte’s Language AI (NLP) Centre of Excellence. Since founding AI Forward in 2023 he has led more than 30 AI and automation projects, and he runs one of those systems on a live sales floor.
If you run a company that should be AI Forward, or want to build one with us, let’s talk.
— Shake Dewan, Founder & CEO, AI Forward