Small firms using AI now build their own, and pay in review hours

Pax8, a Denver company whose online marketplace supplies cloud and AI software to IT service providers serving small and mid-sized businesses (SMBs), published its Q3 2026 Pax8 SMB AI Pulse Report on 12 October. The online survey reached 1,150 owners, executives, functional leaders and managers involved in technology decision-making at companies employing between five and 499 people, split between 400 in the US and 750 in three European markets, including Germany and Sweden with 200 each. It is Pax8’s own research, and the first edition of its quarterly series to survey firms in Europe as well as the US.

Two-thirds (67%) of AI-engaged firms in the sample have built at least one AI solution now in regular use, and another 19% are prototyping. Only 13% rely exclusively on off-the-shelf tools. Of the firms that build, 52% are developing automated workflows that connect AI to other business systems, 46% are building or testing agents that complete multi-step tasks on their own and 39% have built custom integrations using an AI provider’s API.

According to Pax8’s release, a business that deploys an agent or connects AI to its finance system “takes on work that used to belong to software vendors: maintaining what it built, securing it and paying for the compute it consumes”.

Read across the four markets, the survey records three costs of building AI that a license price does not show: the hours decision-makers spend checking its output, bills that arrive above forecast and builds that run without formal review.

Hours spent checking the work

Pax8 found that 69% of AI-engaged firms say AI has freed up at least some capacity. The same survey found that 55% of decision-makers personally spend three or more hours a week reviewing, editing or fact-checking AI-generated work, and 7% spend 10 hours or more. The people carrying that load are the owners, executives and managers the survey questioned.

The release puts no value on that time, and the capacity figure is the respondents’ own assessment. Three hours a week still comes to more than a full working day a month for each decision-maker who spends it, and none of it appears in the price of an AI license or subscription.

Bills that arrive above forecast

Among firms that paid for AI in the past year, 49% received at least one bill or usage charge higher than they expected. The figure rose to 57% in the US sample and 65% in the Swedish one. Pax8 said usage-based charges are proving the hardest to forecast, and that they grow as businesses build more agents and workflows.

A per-seat license costs the same each month however much it is used. An agent running multi-step tasks, or a workflow calling a provider’s API, is generally charged on what it consumes, so the more a firm builds, the more of its AI spend moves from a fixed line to a variable one. Respondents also put a price on data control: 70% said they would pay at least 6% more for greater control over how their data is stored and used, and data security or privacy exposure was the most-cited risk of employee-built AI in every market, at between 30% and 35%.

Builds nobody signed off

Over the past six months, 35% of AI-engaged firms found ‘shadow AI’ – tools, workflows or agents operating without official approval. Formal, documented review of internally built AI ranges from 23% in the lowest-scoring market to 45% in Sweden. In the US, where 81% of AI-engaged firms have built solutions now in use, 36% of builders have formal, documented review and another 41% rely on consistent but undocumented review.

Sweden had the highest formal review rate, and, as The Next Web reported, also the highest rate of shadow AI discovery, at 55%, against 42% in the US. Firms with formal review in place may simply be better at finding what was never approved. Pax8 gives a margin of error of about 6.9 percentage points for each of the German and Swedish samples and describes results for smaller subgroups within markets as directional, so the country gaps carry less weight than the combined figures.

Nick Heddy, Pax8’s president and chief commerce officer, said: “SMBs are moving from using AI tools to building AI into the way their businesses operate.” He added: “The opportunity is significant, but so are the new demands around governance, security, cost, and accountability.” The findings, he said, reinforce why SMBs need “trusted partners who can help them manage agents, workflows, data, and tools as a connected intelligence layer”. Pax8 says more than 47,000 IT partners rely on it, so that conclusion also describes its own business.

On Pax8’s own numbers, the license price is no longer a reliable measure of what AI costs a small or mid-sized firm: almost half of those paying for AI have had a bill higher than they expected, more than half of decision-makers spend three hours or more a week checking its output and no market surveyed puts formal review of internal builds above 45%.