What Has Actually Changed in Bookkeeping Tech Since 2020?
A practical look at the tools and workflows that have reshaped small-firm bookkeeping in five years, from receipt capture to AI-assisted coding and bank feeds.
Bookkeepers who trained five years ago often return to the wider community and find the landscape has shifted. Receipt capture went mainstream, bank feeds got faster, and AI-assisted transaction coding moved from marketing talk into everyday software. This article walks through what has genuinely changed, what has not, and where a solo practitioner or small firm should focus next.
What changed first: document capture and bank feeds
Tools like Dext, Hubdoc, and AutoEntry stopped being a competitive edge and became table stakes. Most firms now push client documents into a capture tool and review the extracted data rather than keying it by hand. The bigger shift is on the bank side. Direct feeds and rules-based matching in QuickBooks Online mean routine bank and credit card transactions largely code themselves, and the bookkeeper’s job has moved toward exception handling.
Where AI actually shows up today
The real change is not a single product but pattern recognition inside the software you already use. QuickBooks Online now suggests categorizations learned from your past coding, and similar machine-suggested matching appears across most cloud platforms. Generative AI assistants have also appeared for drafting client emails and summarizing account activity. These help, but they do not replace review. Suggested categories are only as good as the history behind them, and a wrong suggestion accepted in bulk creates a cleanup job later.
What has not changed
The core work is still the same. Reconciliations still need a human eye. Period-end review, cut-off, and clearing accounts still decide whether the books are trustworthy. Clients still send the wrong documents or nothing at all until the deadline. Firms that cut their task time in half did it mostly through process: standardized chart of accounts, monthly checklists, and a fixed document routine, not through any single tool.
Where the risk sits now
More automation means data problems travel faster. A bad import, a duplicated feed, or a botched migration from another platform can push errors through months of transactions before anyone notices. If your books come out of a conversion or an upgrade looking wrong, our write-up on what QuickBooks Online faults and data damage do to a set of books covers the practical recovery path.
A sensible next step
Pick your slowest recurring monthly task and time it. Then test one change against it: a bank rule set, a capture tool you have not tried, or a checklist that removes the decisions you make from memory. Measure again the following month. Incremental process fixes, tested one at a time, are what actually move the needle, and they compound far better than another tool subscription.