Small firms are holding
the better hand.
For a century, new technology paid off most for the big firms. Generative AI ended that streak, and few have noticed.
Most major waves of new technology have followed the same rule: the bigger the firm, the better its return on the tool. Big firms came out ahead in two ways:
- They got the tool first.
- They got more out of it.
Consider law firms. In 1973, a service called LEXIS let lawyers search court decisions from a terminal instead of a library shelf. Its first subscribers were exclusively large New York law firms. Six years later, a competitor, Westlaw, made online research a bit more accessible with dial-up, but it was still largely limited to big firms and law schools. Westlaw had launched at $1,200 USD a month plus $2.50 for every search, more than $7,000 CAD a month in today's money before you'd typed a single query. The small firms had to keep pulling books off shelves. Every hour the big firm saved on research was an hour the small firm couldn't.
The same rule has generally held across technological innovation. Steam engines went into the big plants first. Business computers took 25 years to get from a General Electric plant in 1954 to a spreadsheet small shops could afford. Online legal research took about two decades to reach a solo lawyer's budget.
The big firms could also get more out of the tool. A large upfront cost pays off best for the firm with the most work to spread it across. We can see this play out in France, for example, which kept unusually detailed records on its companies. Economists who studied those records from 1995 to 2007 found that, within the same industry, bigger firms used far more software per worker than small ones. Put simply, the bigger companies were extracting more value from computing.
For a century, innovation meant big firms winning twice.
How the pattern broke
The adoption gap between big and small firms has closed whenever the upfront cost disappeared for the small buyer. Either the tool got cheap enough to own, or someone started renting it out.
Buying electricity from the local utility did it for small factories in the 1900s and 1910s. Cloud software did it again a century later, renting small companies the kind of systems the giants ran on. And once small firms had the same access, the payoff tilted their way: cloud adoption has been linked to faster sales growth, more strongly for small firms than large ones, at least where someone in the firm knew their way around IT.
But every earlier crack came quite late, and even AI followed the old rule at first. Before ChatGPT, putting AI to work in a business meant classic machine learning: building your own models, with data scientists, labelled data and months of engineering. In 2017, according to a US Census Bureau survey sent to 850,000 firms, about 5% of those with 1 to 4 employees were using AI. Among firms with 10,000 or more employees, it was about 62%.
Source: US Census Bureau, 2018 Annual Business Survey (data for 2017), via McElheran et al. (2024). "Using" means at least one of five AI technologies in production, mostly machine learning.
Then large language models arrived, and for once a brand-new capability showed up as a utility from the start. When ChatGPT launched in November 2022, a one-person firm and a 10,000-person bank could use the same model: free at first, then $20 USD a month. Business versions reached all firms at roughly the same time. There was no upfront cost for size to spread.
Today a seat on a leading platform costs a solo lawyer only about $30 CAD a month (or $140 CAD for a pro-tier subscription). A bank on the enterprise tier pays more, but not for a smarter tool: it runs the same models, and the extra money buys security, administration and contract terms. Size can still buy volume discounts and early access to some models, but generally there's little difference in raw capability.
So technology won't decide who wins market share in this next phase. Speed of adapting will.
The bottleneck moved
If reorganizing is the real bottleneck, the advantage tilts toward the smaller firm.
In 2003, Nicholas Carr argued in Harvard Business Review that IT doesn't matter: once everyone can buy the same technology, nobody wins on it. There's good reason to think the same about AI. In one influential account of factory electricity, motors sat in plants for decades with little gain, until owners stopped bolting them onto the old layout and redesigned the floor around them.
When the tool stops being the advantage, the advantage moves to everything around it: which tasks change hands, how roles get redesigned, how work flows between people, and how you charge for the result. This is how the big consulting firms size up AI transformation today. BCG's rule of thumb puts 10% of the effort on algorithms, 20% on technology and data, and 70% on people and processes. PwC's version: technology delivers about 20% of an initiative's value, and redesigning work delivers the other 80%. Ignore that 70 or 80% and the tool barely pays off. Get it right and that's where the advantage lives.
Why small firms should win this race
Reorganizing is the thing small firms are built to do fast and big firms are built to resist.
At Rotman, I built AI strategy programs for big financial institutions. In that role I sat in the room with C-suite executives and senior leaders from some of Canada's largest organizations, and I watched this play out. We'd work through the principles of a proper AI rollout, one aspect at a time.
In one session, we spent two hours on what makes a restructuring succeed. By the end, the room agreed on the answer: the work itself had to be reorganized around what AI could now do, starting with the people closest to it. The executives understood the theory, and even supported it. Then the talk turned to how they would actually begin, and the mood in the room changed.
One executive pointed to a multi-year initiative already under way that pulled in the opposite direction (one that, I suspect, had a few promotions riding on it). Another explained that changing a single job description meant reopening all of them, and the last time HR tried that, it took two years and a consultant. A third suggested the sensible first step was to brief the eight committees whose input would be needed before anyone could approve forming a new committee, whose mandate would be to find ways to formalize the approval process.
Okay, I made that last one up. But not by much.
Each roadblock narrowed the path, until the room had talked itself into a corner. Minutes earlier, the way forward was reorganizing the work around the people doing it. Now the conversation was about creating new roles, hiring people to fill them, and possibly letting go of the people whose jobs had become redundant. That was the kind of change they were built to make. I remember someone letting out a long sigh, then silence. They knew where they wanted to go. They just couldn't see how their organization would let them get there.
A 20-person firm can make those calls in one meeting. A 100-person firm, maybe over a few. Sociologists call the difference "structural inertia": the routines and chains of authority that make a large organization reliable also make it hard to change, and the effect grows with size.
Cutting back is harder than building fresh. A big firm has to unwind roles, systems and habits that already exist, while a small firm can build on a blank slate, and a blank slate gathers momentum fast. You can see it even at the scale of just one person: in a field experiment with 776 professionals at Procter & Gamble, one person with AI matched the quality of a two-person team without it. A blank slate has never been worth more.
The advantage nobody is using
Small firms hold the better hand, but most aren't playing it yet. The large firms are the ones moving.
Make no mistake, large firms understand what's at stake. This isn't only the tech giants building data centres. Big companies in nearly every industry see both the opportunity and the existential risk of standing still, and they're spending heavily to capture every bit of growth they can. Money and motivation aren't what hold them back. Reorganizing is.
Even so, they're pulling ahead for now. When Clio surveyed US law firms this year, most small firms were already using AI, yet fewer than 33% had grown revenue from it, against nearly 60% of firms with 200 or more staff. Numbers like these make it look as if generative AI plays by the old rules, and the big firm wins again. Look closer and the big firms are winning on the easy changes.
Clio points to pricing: 86% of solos hadn't changed how they charge, so when AI made the work faster, they simply billed fewer hours. More than half of the largest firms had changed their pricing, and the time AI saved became revenue instead of a smaller bill. Pricing is a change a big firm can make with one decision at the top, so it moved first. The harder change, reorganizing the work itself, is still wide open, and that's the one small firms are built to make.
Many haven't even started. When Statistics Canada asked businesses with no plans to use AI why, 79% said it wasn't relevant to what they do. Fewer than 5% said it was too expensive. That's an understanding gap, not a capital gap, and it's far cheaper to close.
And it's a gap with a clock on it. The old gaps ran for decades: about 20 years for online legal research, 25 for business computers. This one won't last that long. Whether it's two years or ten, even a slow-turning cargo ship eventually comes around, and some big firms are already turning. Or the competitor your size across town gets there first. If you run a small firm, now is the time to get your house in order.
What to do this week
- Pick one role in your firm and ask one question: if we were starting this business today, with these tools, would this job look the same? If the answer is no, that's where your advantage is.
- Write down what would stop you. List every approval, system and habit that would have to change to redesign that one role. If the list is short, you're the small firm in this story. If it's long, you now know what to fix first.
For a century, size multiplied what technology was worth. Generative AI took that away. Now it's a race to reorganize, and this time, small has the head start.