Two pieces of research published in the past week point in opposite directions, and the gap between them is the most useful thing a small firm will read this month.
The first is a study from the consultancy BearingPoint, reported on 1 October, covering 1,050 C-suite and senior leaders across thirteen countries in Europe, the United States and China. Only 13 percent of those organizations are scaling their AI work as originally planned. Roughly three-quarters either cut the scope or achieved less than they intended. Deep integration rose from 7 percent last year to 11 percent this year, which is progress, but slow. More than a third are still experimenting.
The barriers they named are revealing. Forty percent cited legal regulation. Thirty-four percent cited integration with their existing IT systems. Fifty-four percent said trusted, high-quality data was the critical factor.
Yet nearly three-quarters reported measurable business impact from the work they did manage to do. The returns are there. The scaling is not.
Now the other number
On 28 September, Federal Reserve Governor Lisa Cook gave a speech on AI and the economy. Citing the Fed Small Business Credit Survey, she noted that nearly half of small employer firms now use AI, and 71 percent report increased productivity. She called AI “poised to become the most significant technological shift of our lifetime,” and was careful to add that the effects arrive with long and variable lags.
Hold those two findings next to each other. Large organizations with budgets, data teams, and procurement departments are stuck at 13 percent on-plan. Small firms with none of those things are at roughly half adoption, and seven in ten of the adopters say it made them more productive.
Why the small firm wins this one
It is not because owners are cleverer. It is because the three things blocking large companies are things a small firm doesn’t have.
Integration debt. A twenty-year-old ERP with four bolted-on systems and nobody left who understands the middle layer. You have a handful of tools, and you know them all.
Governance by committee. Decisions need sign-off from legal, IT, risk, and a steering group that meets once a month. You decide on Tuesday and start on Wednesday.
Data nobody trusts. Fifty-four percent of those executives said trusted data was the critical factor, and in a large organization, departments have to negotiate that trust. In a five-person firm, you know where the numbers come from because you entered most of them.
That is a real advantage, and it has a shelf life. It lasts as long as it takes the large firms to work through their integration problems, which, on the evidence of a 7 to 11 percent move in a year, is a few years rather than a few months.
What to do with the opening
Use it on something narrow. The firms in that study did not fail for lack of ambition; they failed trying to scale. Pick one process, automate it properly, and measure what came back before you pick the second.
And measure honestly. Nearly three-quarters of those large organizations reported measurable impact, which means they bothered to measure. A small firm that automates something and never checks the hours has no idea whether it won. Knowing how many hours you got back, and what it cost to set up, also tells you whether the thing is ready to carry more volume as you grow.
AI Today covers where to start, how to find automatable hours, and how to measure the return without fooling yourself. This week’s research says the door is open. It does not say it stays open.