77% of Small Businesses Use AI. Finance Is Last in Line.

Seventy-seven percent of U.S. small and midsize businesses report using AI regularly as of January 2026, up from 48% just 18 months earlier. That number comes from the 2026 Intuit QuickBooks AI Impact Report, drawing on over 34,000 survey responses across seven quarterly waves, combined with anonymised payment data from more than 5.3 million businesses.

It is a real shift. But the headline hides a gap. When you break adoption down by function, marketing leads at 45%, customer service sits at 37%, and bookkeeping trails at 35%. Finance is last. That is strange, because finance is the function where time lost to manual work is easiest to measure in dollars. Every hour spent keying invoice data has a known cost. Every mistyped figure has a known consequence.

So where are ai tools for accounting actually delivering results, and where is the data weaker than it looks?

The productivity numbers, and what they leave out

The headline figures sound convincing. 78% of U.S. businesses using AI say it improved their productivity. Businesses reporting revenue increases outnumber those reporting decreases by more than 20 to 1 (43% vs. 2%). Roughly one in four say AI has shortened their workday.

But Forbes dug into the methodology and found a problem. When researchers asked businesses how they measured these improvements, more than 50% stated only a general feeling that their business was better. Less than half tracked specific metrics. The productivity figure was based on self-reporting, not time studies. The revenue attributed to AI was based on correlation, not controlled measurement.

That does not mean the gains are fictional. It means most businesses cannot prove them. For finance teams, where every process has a measurable input and output, that is a solvable problem. More on that below.

Three accounting tasks AI handles today

The useful question is not “should we adopt AI?” but “which tasks can AI take over right now, with minimal setup?” For accounting, three stand out.

Invoice and document data extraction is the highest-impact starting point. AI-powered OCR reads invoices, receipts, and statements, then extracts structured data, including vendor name, amounts, dates, VAT, and line items, without manual keying. The data flows straight into your accounting software. This is the task that eliminates manual data entry in finance, and it is the easiest to baseline: count how many invoices your team processes per week, time the process before and after, and the ROI calculation writes itself. Zerentry's document processing handles this extraction for vendor details, amounts, VAT, and line items.

Bank reconciliation is pattern-heavy, repetitive work. AI handles it end-to-end for standard transactions, flagging exceptions for human review. The human still reviews the flagged items. The difference is that the bulk of transactions that match cleanly no longer need a person at all.

Expense categorisation and GL coding round out the list. AI handles expense categorisation, receipt matching, and inconsistency flagging. It learns transaction patterns over time, suggests smart categorisations, and speeds up month-end reconciliation. Tools in this space include Botkeeper, Vic.ai, Dext, and QuickBooks Intuit Assist. These three categories share a trait: they are mechanical, high-volume, and error-prone when done by hand. That combination is where ai tools for accounting pay off fastest, because the manual alternative has a known, measurable cost.

Why finance teams are last, and why that is changing

The barriers are not unique to finance. Across all four countries tracked in the report, the top barriers are concerns about data privacy, fear of errors, and limited knowledge of what AI can actually do.

But the pressure underneath is structural. The Bureau of Labor Statistics projects a 6% decline in bookkeeping clerk employment from 2024 to 2034, directly attributed to software automating those tasks. In the same period, accountant and auditor roles are projected to grow 5%, with automation making advisory duties more prominent. The profession is splitting. Data entry work is shrinking. Advisory work is growing. AI is the mechanism driving the split.

Among tax and accounting professionals specifically, adoption is already higher than the small-business average. Thomson Reuters' 2026 AI in Professional Services Report found 69% AI adoption among tax and accounting professionals. Generative AI adoption at the organisational level rose from 22% to 40% between the 2025 and 2026 surveys. Among current GenAI users, 57% use AI for document summarisation and 55% for document review. Finance teams that have not started are increasingly the exception, not the norm.

How to measure whether AI is working

Here is where the 77% stat breaks down. Gartner predicts global AI spending will reach $2.52 trillion by 2026, a large portion of it spent with no real method to measure the return. Forrester predicts 55% of AI projects will not meet their intended goals, not because the technology failed but because there were no pre-defined success metrics.

For accounting, this is avoidable. Unlike marketing or customer service, finance tasks have countable inputs and outputs. A useful measurement framework comes down to three numbers:

• Time per task. How long does it take to process one invoice before AI, and after? If your team processes 200 invoices a month and AI cuts per-invoice time from 8 minutes to 2, that is 20 hours reclaimed.
• Error rate. How many invoices require correction after initial entry? Manual keying has a known error rate. AI extraction has a different (usually lower) one. Track both.
• Output volume. With the same team size, how many more invoices or reconciliations can you complete per period?

Document processing is the ideal first measurement target because it is time-boxed, high-frequency, and has clear error signals. You can run a baseline in a single week. That is why it is the recommended entry point for automating bookkeeping tasks rather than trying to adopt AI across the entire finance function at once.

The wage signal

One last data point worth noting. PwC's Global AI Jobs Barometer 2025 found that AI-skilled workers in business and finance roles earn a 56% wage premium compared to peers without AI skills. That is not a round number chosen for a headline. It is a specific finding from an analysis of nearly a billion job ads. Finance professionals who learn to work alongside AI tools are being paid significantly more than those who do not.

The 77% adoption figure tells you that most small businesses are already in. The 35% bookkeeping figure tells you that finance teams have room to catch up. And the task list, document extraction, bank reconciliation, expense categorisation, tells you exactly where to start.

FAQ